Give them their moment in the moonlight, it's not like Intel have had a monopoly for the last 10 years or so is it. This is still a huge step forward for AMD, and the price/performance could be just what we need.
Dudes, AT presented the results of AMDs launch and it was their duty to present these caveats. Otherwise, if AMD messed with the benchmark, there will be plenty more people complaining that AT presented some skewed results without even a mention of the caveats.
Striking similarities in their unrelenting Nvidia/Intel shillage lead one to believe that Michael Bay and Chizow might indeed be one and the same sad individual.
He could be right and i wouldn't bet about anandtech honesty, since they talk about credible benchmarks and blender not being reliable, but cinebench is intel favorable and polaris benchmarks are crap.
But he is also wrong if he thinks this article is wrong and AMD is playing with fire, or already burned to the ashes before we realize it.
As anandtech says, AMD moves are weird, marketing is weird.
Ashes of the singularity leaked benchmark show a WEAK zen. ZEN performs there in the leak at 58 FPS core score. 6900k at does it at same setup and settings at 127.
Also Zen with 16 threads is just faster than 4690k (4 threads), but slower than 4790k 8 threads. Adjusting maths by clock per clock, still looks like ZEN is 40% slower than intel core per core, thread per thread. FX 8350 does half what ZEN does, at same clock speed BUT ZEN have exactly twice the threads piledriver has. So ZEN, assuming that leak is true, is just a piledriver chip with a lot less TDP and just more cores. This is what i'm seeing.
Why this AMD presentation came out after that leak, is also interesting.
My question is, who is lying, AMD or that AOTS leak?
what frequency was that zen? remember that in the demo test for rendering, they dropped broadwell-E's frequency to match Zen's. What if that "leaked" thing was on an even lower frequency rates and the 4790k was on full turbo?
There is no "leaked" Ashes of Singularity benchmark because THERE'S IS NO such benchmark done for a Zen CPU, it has NEVER been done. It's all a media shill, click bait.
More like a petulant child pissed off that these overpriced pieces of silicon will be boat racing Intel in almost every category, and meeting at the finish line with their Xeon systems. In short, Ten years of gap have been closed; and that pisses them off.
Fukc 'em. If Anandtech and the rest of the pro-nVidia/Intel cabal can't stand that AMD is rising up to destroy both of these larger companies, they can suck d0nkey balls.
I see anandtech like an intel fanboy who this time is right, just this time.
Sorry guys, anandtech is right this one. And is good, we lose nothing taking precautions. If ZEN is a success, we all and even some intel fanboys, will buy a ZEN chip.
I just think we are getting close to dark ages of tech and gaming.
Not strictly an intel or ms or apple fanboy, they are fanboys of money, and amd just can't afford to pay everyone off because it does so poorly at business.
The entire tech internet is about to be talking about these benchmarks, benchmarks people have been waiting for for what, 5+ years? And you don't see the point in this article...?
The article claims amd speculations on basis of speculations. It is pointless, stupid and hypocritical.
This is not benchmark, just one random test amd made, without making any concrete performance claims relative to intel. What amd said is 40% IPC improvement, which given how bad previous amd designs were, is 100% possible.
Nobody said it will be "2%" faster than broadwell-e, so it brings the questions what's with this article trying to convince us in the opposite without containing a shred of evidence or fact... I know AT standards are fairly low, but that's low even by those standards.
I can't stand all the people who came from the dailytech comment section. That site had trash comment sections towards the end and now they came here to ruin this site's comment sections.
The problem is our expectations are so low for AMD these days that it's hard to believe they will actually ever be competitive with Intel again...but Intel IS getting lazy. As far as I'm concerned they haven't had relevant IPC gains since Sandy Bridge
I've done a lot of reading recently on SB-era Xeon servers (which can be gotten on Ebay for cheap now) with 2 CPUs and tons of RAM to do some highly threaded workloads. Modern Broadwell-E chips (in real workloads) in single chip configurations beat them. Not by a lot, but enough that Xeon's based on Broadwell-E are far more appealing to that market.
So the claim about SB era chips having nearly the same IPC as modern chips is empirically false outside of synthetic benchmarks, like you see on review sites. This is why reviewers like the ones at AT have to do a good job of explaining the results.
He said meaningful improvements not none. Yes for servers or situations where absolute max performance matters you'd rather have a Broadwell. The issue is that for standard uses, including gaming, a Sandy Bridge is nearly as good. If it's slower, it's by a small enough margin that it frankly doesn't matter.
ddriver, I think you overestimate people's ability to be rational, people will fill the void with their hopes and dreams and change what is a essentially a "we got a not-shit CPU" announcement, with "Zen pwns Intel in to the dust!~!~!", because it won a benchmark by a whopping 2% (which looking at the video on youtube, is easily calculated).
Just look at the fallout from No Man's Sky, Some thought it was better to DDoS reviewers websites than face the fact that /someone didn't like the game they liked/.
Nope, I just make a clear distinction between "people" and fanboys/wannabes - and nobody gives a damm about the beliefs of the latter. It is obvious that everyone in the right mind would hope that zen turns out ok for the sake of competition and prices, but the point is amd claimed no concrete performance figures, they just demoed a running chip using a single app against identically clocked direct competitor - only a complete idiot would take it as a concrete indicator of performance, and only a fool would immediately launch an article to "refute" those non-existent claims. It is funny that the bulk of the suckup readers of AT think very highly of their intellect, yet they would need such articles to see through "amd's deception" LOL. And with claims that Blender being open source being bad and such other absurdities. I guess if it is not closed source test that can hide the `if (cpu != intel) doBad();` lines and everyone can browse the source to see what's going on exactly that's just not an OK test.
Sorry you apparently didn't like the article that made it incredibly clear what was going on, which is, as I said, already the topic of much conversation all over the place. By the way, AT has posted many many articles that quote internal numbers over the years and always come with a disclaimer of their source. Including this one! How wacky-wild!
Oh also, if you are so pouty about AT's standards, be a big boy and go someplace else? Nobody cares about whatever stupid brand fanboy angle you are trying to play in 20-goddamn-16.
Yah, we need this hype to build, so they they can sell out of Zen and also sell "founders" editions for a few hundred. Give some infusion to AMD to build some competing products. Getting bored of Intel and Nvidia always winning.
Why don't we speculate why AMD went from a closed source Cinebench R15 (as often seen in AMD's presentations for IPC measurement) to using open source Blender (which as the article stated, can be modified and recompiled)? Thus why didn't AMD show the standard default Cinebench R15 benchmark which can be easily compared with already existing results?
Maybe they were trying to show Zen doing something in an actual production software(sorta, it`s Blender) instead of synthetics. Then again, they should have gone with Maya or something.
I dont see whats wrong with that. Intel has modified and manipulated benchmarks themselves. Anyone remembers the multithreaded Pi and chat programs scandals? where intel pretty much sabotaged the multithreading if it detected a non intel processor.
Our win is not in AMD challenging Intel (in the ultra high end) but AMD challenging Intel in the midrange. Intel currently has two main challengers to their HEDT platform. Old Intel HEDTs (you have to make enough of an improvement to warrant people buying the new model) and Intel's midrange platform (the I can pay $200-$400+ less and still be satisfied platform).
AMD challenging Intel's midrange offerings will add a 3rd competitor to that market... one that isn't Intel.
So regardless of how fast these chips are against modern chips (which will be Kaby Lake by the time this launches), it also could put significant pressure on the upper tier as well (Broadwell-E).
I dont disagree with anything in this article, but it's sort of a given to not listen to AMD's performance claims ahead of launch. Pretty much every new chip design they have released over the past decade has been alot of hype, then delays, then thermal issues forcing them to reduce the final clock speed, then actual products that fail to live up to the claims... Without fail every single time. AMD is capable of a home-run here, and based on the hype it seems like this one is finally going to be competitive this time fore sure right? Either way, none of it means anything until retail units (not engineering samples) are independently tested.
Not this time, they absolutely do have a serious chip on their hands. This is not like the previous few generations. They brought in one of the best designers in the world (in history for that matter) to do this.
AMD are just taking their time and leaking small details because they want to get everything right with this as it is a very significant launch. Maybe the biggest in their history.
Also, let's put something into perspective here, they already have several future generations mapped out for Zen, as per the designers instructions. They are not, and have not, claimed that the first generation will outperform Intel's top chips at this point. They are clearly looking to get back to being competitive from a price/performance standpoint and fix the outstanding issues of TDP/IPC.
I have family/friends who work for the company, and they do truly have a great product, that will be even greater moving forward. They just need to get the launch and marketing right as this has always been their weakest point.
They need to manage expectations which is why they are not talking about numbers for the most part right now.
Unless AMD intends to shake up the Performance/Dollar ratio and throw Intel from its commanding perch, it won't matter how great the product is. Currently, Intel wants $1,700 for its top HEDT chip and $1,100 for the 6900K that AMD used in the benchmark.
The real question is whether AMD will simply "fall in line" with Intel's pricing in order to maximize profit or shake things up and expand market share. I think they need the latter more than the former.
Very true. If it is as good as Intel with the same price, there is no need to choose AMD. It has be at least %5 better than Intel in all the benchmarks or metrics and it has to be at least %20 cheaper than Intel. That would make it competitive. Of course by 2017 Intel has already made all the initial high profits that it would usually make on its latest chips and will be ready for a fight both in terms of new offerings and the price. However, if AMD really can make a good impression with Zen in 2017, the real benefit to the consumers will start arriving in 2018 and 2019. That is when Intel will have to move some schedules forward or in other words start ticking for real.
"It has be at least %5 better than Intel in all the benchmarks or metrics and it has to be at least %20 cheaper than Intel"
I'd have to disagree. While I truly wish they keep pushing prices down and value up, I'm sure many OEMs and informed consumers would prefer the AMD route, especially for ultrabooks. Intel's iGPUs are nothing but brown stuff; Sure they're "OK" for light workloads and some *old* titles with low settings, but they're by no means good satisfactory. Hell, even their HDMI implementation doesn't support 5.1 audio. AMD will BURN them there with integrated Polaris GPUs with MUCH better performance/watt, better multimedia, and MUCH better drivers.
Your statement has some truth to it in regards of consumer perception. "Core i7" does have a ring to it. I seriously hope AMD does marketing right, and reviewers/sites/blogs are truthful about the value (AND performance) the consumer is getting with AMD chips.
"Hell, even their HDMI implementation doesn't support 5.1 audio."
What are you talking about? According to this very site (http://www.anandtech.com/show/2622/3) Intel's GPUs have supported up to 8 channels (aka 7.1) over HDMI for over 10 years (since G965). Before I moved to a RPi3 my HTPC was running on a SNB Celeron and definitely did 5.1 just fine.
Well having plenty of machines do the same exact same feat that Wolrah used his SNB celeron for, I can definitely say it is supported and you obviously just don't know what you're doing. There is such a thing as user error.
This. If Zen is anywhere near Intel in performance it will have them shitting down one pant leg. This is because they won't stand a chance against AMD's Zen APU lineup. My ancient Llano APU has an integrated GPU that's just as fast as a discrete R7 240 SFF. All the interconnect/HSA work is already done. All they needed was a viable X86 core and now they have it.
It's my belief that the massive layoffs by Intel recently were a direct result of Zen hitting sampling stage. Intel knows they're screwed, and there's nothing they can do about it.
It's a bit silly to think Intel is laying off workers because they are so petrified of what AMD is bringing down the pipeline. Far more likely (and documented) is that Intel has been replacing high-cost American workers with significantly cheaper H1-B VISA employees that will work for 1/3rd of the pay. The layoffs had everything to do with maximizing profit and nothing to do with fear of Zen.
It's this kind of baseless speculation that causes AMD's products to not meet expectations. We have no idea how much the 8c/16t Zen will cost, and we might not be able to trust AMD's benchmarks for reasons stated in the article. Setting your expectations so unrealistically high will only result in disappointment.
Except the price of the $1,700 chip is simply because it is "the best". Intel will happily sell you a $300 chip that gets 90% of the performance, but doesn't give you bragging rights of 'the best'. If AMD can't match the $300 chip, then don't expect the profits (the $200 i5[K] gets close to the $300 i7[K] and things go downhill from there (last I saw, the 8320FX was "competing" with the i3).
It gets even worse with the economics of Naples. The big Xeon chips really rake in the money for Intel, but the cost of the systems that they are in are so high that a 90% AMD solution simply isn't worth putting in the expensive system even if it is free. AMD has to completely beat Intel at *something*, *anything*. Preferably something that they can convince about 10% (or whatever GoFlo's capacity is to Intel's) of current customers to abandon Intel and go with AMD. Maybe Blender can do it (although it doesn't seem likely that AMD will beat Intel's clocks, which appears necessary in this benchmark), but I would recommend something like using SSD as main memory (preferably using HBM as "cache', although that largely depends on AMD's experience with the Fury). While this wouldn't appeal to everyone, massive cheap "memory" should at least appeal to enough server buyers to put AMD back in the game. The biggest danger is Intel swooping in with 3dXpoint and re-taking their lunch back, but AMD really doesn't have all that many options.
You have No idea what you are talking about. AMD only needs a roughly competitive server chip, they do NOT have to beat Intel.
In PCs, AMD's graphics are 50% better than Intel. Because ZEN is roughly equivalent with Intel, AMD's APUs will destroy Intel's integrated graphic PC chips.
The only reason AMD APU's smash Intel graphics is because they devote twice the die area to the GPU, at an obvious cost of performance to the CPU, but who knows if adding more cache, etc, would even help their architecture.
"In PCs, AMD's graphics are 50% better than Intel."
AMD devotes a much higher percentage of the die than Intel does and (counter to what you might expect) Intel has a tighter integration between their CPU and GPU (on die) operations. Basically, Intel started integrating GPUs on die before AMD.
The fact that Intel is even close to AMD is actually amazing, given those constraints. But the reality is that Intel spends something like 3 times what AMD does into R&D.
That said, for GPU optimized builds (that you don't want a discrete card), AMD has the best offering.
CPUs are not GPUs. AMD is already years ahead of Intel when it comes to GPUs and multimedia. AMD should not waste any space for multimedia on Zen than it has to. Just keep it close to what Intel does and that is more than enough. In other words just chase after Intel when it comes to the GPU part and let Intel worry about staying ahead on that front. Later on and in short order AMD can add enough multimedia capabilities for laptops and home entertainment devices on its CPUs. For now the CPU part is the most important part to focus on.
"Not this time, they absolutely do have a serious chip on their hands."
- But they say that every time. Phenom, Bulldozer, piledriver, steamroller, excavator and the others, every single bloody time. Also, where is the retail chips? Already delayed until 2017. I really hope its a great chip, we can all benefit from the competition, but I do not believe anything AMD says. It may well be a great chip, and when retail units are independently tested I will believe it and not a second before that happens. I also would highly recommend no-one buys anything until retail units are independently tested. Too many
This is a new architecture so there's an actual chance it'll be way different/better. Over the last many years we've seen the same cpu's and gpu's with minor yearly tweaks from AMD (and intel to be honest) and minor tweaks aren't moving the needle much. This is a major release and it could be something....
But AMD doesn't seem to be announcing anything new/groundbreaking. With the Athlon 64 they moved the memory controller on-die and went with a shorter pipeline. It was up against the P4 that was a very different design. At the time it was a big deal. Zen doesn't sound like it's doing anything intel isn't doing, so I don't expect it to be faster than intel. I also suspect it'll use more power and clock lower due to fabbing issues and possibly design choices as intel has spent years reducing the power consumption on their chips. I'd love to be wrong but we need to see independently tested retail chips. Equal IPC isn't nearly enough. (And that's assuming the blender results apply to other tests as well).
I don't want Zen to fail, but I'm skeptical it'll be everything people hope.
While I agree with your scepticism, one genuine option for AMD is releasing a cost-effective 8 core (16 thread) CPU. For years now Intel have been dedicating die area from process shrinks to integrated graphics, rather than larger CPU cores, or more cores. If you want more than 4 cores, you need to spend up on the HEDT platforms. Broadwell-E even extended the pricing divide between their 4 and 6+ core CPUs.
Even if AMD can't quite reach the IPC of a Kaby Lake quad core, if they can get within 10-25%, but in a genuine 8 core CPU at a competitive price, that would still be a compelling product IMHO. If they choose to dedicate the die area to additional CPU cores rather than large integrated graphics, perhaps they could price a genuine 8 core 16 thread CPU similarly to Intel's high end quad cores.
"With the Athlon 64 they moved the memory controller on-die and went with a shorter pipeline."
For the sake of historical accuracy, AMD didn't come up with the on-die memory controller. The feature had already been around for about a decade when AMD implemented it.
Secondly, AMD didn't just "go" with the shorter pipelines, their chips were based on the Pentium III (the one Intel had sadly dropped in favor of the Netburst P4), which had shorter pipelines, as designed by Intel.
Agreed. This launch is crucial. They had a minor hiccup with the 400 series launch, but recovered quickly, and well - although I have to say they -had- to know what was going on before they released it.
All in all, as long as they regain a competitive footing against Intel, that's what is really important. Not -beating- them necessarily (although that would really be a kick in the nuts to Intel), but just getting close which is something they haven't been in quite a few years. They don't design bad chips, but they've really missed the boat with thermals/TDP.
Another area where Zen will hopefully pick things up is in the far more lucrative server market. Opterons are almost nowhere to be found these days, and getting back into the server market with a Zen based Opteron again would be a great start.
At least somebody gets it. Keller has designed this chip from the ground up, it is NOT a piledriver rehash. It will at least be competitive with Intel's offerings, and every human being alive should be grateful for that. A lack of competition breeds $1700 chips.
Actually, Keller didn't design Zen "from the ground up." AMD stated that he was given free reign to expand upon their existing base for that was already in the R&D stages for Zen...Now Zen+ designs will most likely feature more of Keller's vision..
Either way, I am very excited for this release. I have a 5Ghz (48+ hour torture test stable) 3770K coupled with a high end Z77 board...I have plenty of IPC, but I would like to have double the cores, and I would really love to have the platform updates (more PCI-E lanes, M.2, etc)..I think AMD is going to have a winner on their hands if they market and price it right..
All of the fanyboys expecting a $300 8c/16t cpu that is just as fast as a Broadwell-E SKU are going to be in for a rude awakening..I wouldn't expect AMD to price it any less then $600 vs the $1000 asking price from Intel. That would still be a great value, and allow AMD to have some much needed profit margins..
I don't think the design is bad but you do have to remember that they're hugely disadvantaged by their use of Global Foundries to actually make the chips. It's using Samsung's 14nm FinFET but from what I'm hearing this doesn't mean it's competitive with what Intel can do and not everyone is all that confident about GloFo's implementation. AMD is contractually obligated to use Global Foundries for a lot of stuff because GloFo is a spinoff of AMD. Zen might have been better off being made by TSMC or even Samsung themselves.
Honestly I think this time around it's not really the architecture that's the issue but the fact that AMD has to use Global Foundries due to obligations leftover from spinning them off. I've heard from people who've talked to people who really would know whether Global Foundries is doing well with their 14nm process (licensed from Samsung) and the stuff they're saying isn't particularly reassuring. If Zen ends up not being able to clock up it may end up largely being because they're stuck with using GloFo instead of another foundry like TSMC for these CPUs.
Why are we comparing Zen to Broadwell? Broadwell is Intel's tech from 2014. Skylake's IPC is greater than Broadwell, and by the time Zen actually releases we will have Kaby Lake with even more IPC than Skylake.
As a gamer I am interested in buying whatever CPU has the fastest per thread performance with at least 4 cores. Currently that's the 6700K and soon will be the 7700K. So to me and millions of people like me that's what AMD needs to compete against as well.
Not really interested in something with more cores but slower cores since it doesn't help my workloads.
Anyway,if you only need 4 cores you can look at a 120-150$ Zen quad core with no GPU and decide if you pay 350$ for Skylake with a GPU you don't need or get the Zen based on w/e metric matters to you.
Its just a matter of comparing Intel's highest performing consumer CPU. That would be Broadwell E. AMD will have an 8 core consumer CPU that (they say) will compare to Broadwell E, Intels current top of the line.
When was the last time it really mattered? Ten years ago with the core2? Intel cores have pretty much topped out and have had tiny incremental boosts for a long time (although I've noticed that the 8350FX seems to have fallen from "competing with the i5" to "competing with the i3").
If AMD could somehow compete with Sandy Bridge (which I *think* is the basis for all recent cores, there don't seem to be any visible uarch changes since then) they should be able to make some money. I'm not even holding my breath for that, Intel has simply been executing too well since abandoning Netburst and Itanium.
1/3rd? Source please, Intel has been increasing their TDP from 65 watts and are now at what 77 watts? So you are saying that despite the increases to TDP somehow your chip has magically decreased? Impressive to say the least, please share with the world how you pulled off this magnificent feat!
So what? Intel's generational increases have mostly been about i/o and connectivity, not about improvements to the core design. Most people that have SB still have no reason to upgrade, so it doesn't seem to matter much if it's one, two, or even three generations behind. AMD doesn't need to match or exceed KL IPC to make a dent in the market, they just have to be competitive in price/performance/watts.
You're right on this but very few talk about it. How much does the iGPU add to the cost of an Intel chip and do you need it if you already have a discrete GPU in your system ? And then there is the software. I was dumb enough a few years ago to buy a laptop that had a discrete GPU in it. Any time I was running Google Earth the program would crash after a few minutes. It turned out that both graphics drivers were running and conflicts arose. I had to disable ( uninstall the driver) the GPU to get the program running. What a pain in the a--.
A massive majority of i5 and i7 users are business and Soho applications then gamers. All of these applications currently benefit much more from single core speed up to 4 cores and occasionally going up to 6. Meaning a very large majority of users will see 0 benefit or even negative performance from an 8core/16 thread CPU if the per core speed is lower than Intel's 4 core/8 thread, including gamers.
Because it is an 8 core part with 16 threads and if you want 8 cores and 16 threads on intel you have to go x99. Rumour has it that 6 cores will come to the z platform. Where up to now a 4 core with hyperthreading was the maximum.
Because realistically this is the only scenario where Zen will be competitive, on the CPUs with tons of cores they become much more competitive because the intel parts are also low clocked and more concerned about computing power efficiency than raw power per core.
Skylake doesn't have the core count to compare to Zen; they're comparing to Broadwell-E. Broadwell-E was released this May, and its replacement, Skylake-X, isn't due until 3Q 2017.
As everyone here knows and is now anxiously awaiting... The clock speed. If it can come out at ~4Ghz and this leak rough IPC is accurate, it needs upwards of 3.8 base clock. If they can do that they have a winner.
On the other hand, AMD's 3GHz Zen engineering sample demonstration could be in the same mould as 3GHz "Phenom FX" http://www.dailytech.com/AMD+Demonstrates+30+GHz+Q... which never saw the light of day. That is considering 14nm LPP which is primarily low power process node optimized for mobile SoCs with low clockspeeds.
They're comparing Zen to Broadwell-*E*, which was released less than 3 months ago. This comparison was presumably chosen because Zen is an 8C/16T part, and Broadwell-E is Intel's equivalent in that space.
It's being compared to the most comparable Intel part. Intel's current "big core farm" offering is the Broadwell-E, and the 6900K discussed was released this May. Obviously they can't compare to Kaby Lake as that isn't out.
Because if you want to use Blender as a test between different vendors, and the newcomer has 8C/16T, they want to show that their 8C/16T is just as good as a very expensive CPU from the competition.
We don't have a Skylake-X yet, so Broadwell it is. It's also very likely that Skylake is faster than Zen on single-thread IPC, why show that off when you can be as fast as something fairly recent that is still at the high end?
because skylake is barely even better than haswell. there is almost no improvement at all. if a cpu is as fast as haswell or broadwell it is basically as fast as skylake.
It's marketing, we all know it's the absolute best case scenario and performance will be nothing like that in the real world. Disagree, look at AMD's last big bit of performance number marketing of a future product: http://www.anandtech.com/show/9886/amd-reveals-pol...
Hahahahahahaha xD. Sounds like Intel's Comet Lake marketing tbh. It's still better than AMD at 720p gaming (which is completely irrelevant) & that's why you should buy it instead of Ryzen!
Good on Anandtech giving AMD the scrutiny equally that all major technology players receive here. Instead of sugar coating AMD's bold claims which other tech site articles does.
I feel that the point of the benchmark was to show that they have produced a competitive processor. Given if there were optimizations, some cherry picked conditions, they are at least showing that they have overcome significant architectural obstacles and thanks to the choice of foundries they can at least approach the performance of Intel.
It is impressive, at the least, that they are staying in the game being a much smaller organization. Thanks Lisa.
If AMD have learnt from their marketing mistakes and are being conservative and under promising on performance, having Zen beat Broadwell-E by 2% clock-for-clock is pretty damn good.
The cheapest Broadwell-E costs around $400. If we get even 90% of its performance at around $300 from AMD with similar power requirement, Then Hats off to AMD. I am sold
On the memory side it seems that Blender doesn't care much about bandwidth and to be fair lots of consumer apps don't scale with BW. On the TDP side,a roadmap leak and Zauba are showing 95W but we don't really know what clocks they can reach.Still, it would be safe to assume 95W or less for 3GHz and then you could compare with Broadwell-E -not that it matters since perf is unclear for now. On point 3 and 4 maybe you go too far on the negative side and somehow you argue for SPEC? That's too much about the compiler and Intel puts too much effort into it. You used it to compare Apple's core to Intel and that was unfair.
Also don't make the mistake of assuming that clocks, perf and power can only only surprise on the negative. Intel had no reason to try for more than 5 years, it was all about keeping revenue up in a declining market. We'll also have to see how it scales,i'm not convinced yet that this is aimed at very high perf instead of moderate perf and higher efficiency.
There remains no point in speculation about how Zen will perform. When retail products are subjected to benchmarks and multiple hardware review sites have posted numbers that can be compared among one another, we'll know.
Even if Zen was significantly faster than Broadwell-E I still would NOT even consider the product. New architectures are often full of bugs. Zen will have to demonstrate reliable and trouble free operation for weeks (if not months) before I would even entertain the notion of using AMD in my next system build.
I think it's a legit concern. The original phenom had a serious bug and the bios fix resulted in much lower performance. Obviously intel processors can have bugs to, but it's reasonable to assume that a new architecture is more likely to have serious bugs than an existing one.
It'll be amazing to finally get some competition. I'm still really happy with my Sandybridge though.
It'll be tough for AMD though, Intel has a huge margin/price advantage with their fabs and volume. Also, not clear how many smart competent ppl actually still work @ AMD, but even Intel has laidoff a lot of ppl. Very little money left in the consumer market on the low end and Intel is pretty competitive there.
We're trying to extrapolate on a product that we won't see on shelves for almost a year? Lot can change by then, will see. Obviously there's no trust right now and it'll have to be earned after the previous failures.
I LOVE how AT insinuates that open source benchmarks are somewhat less trustworthy. That's just....well, odd. Typically, people point explain their setups, point to the code they've run, and then TRULY anyone can duplicate it AND, far more importantly, examine the friggin code to see if there is an gaming happening. Let's just relax, and see if amd does the right thing and points to the code (and toolchain, ofc) so it can be examined.
I am pretty sure they were talking about compiler optimizations regarding open source software.
Closed source is compiled for all CPUs. Probably optimized for Intel as we saw with No Man's Sky. But open source it is easy to rebuild it with custom options to make it run really well on a particular CPU.
Compiler optimizations are covered by reference to code (to see if they actually made source code changes that benefit them....and, as I said, they have done this, but it is well known in the oss community and if amd was trying to hide that info, they did a terrible job) and toolchain (which includes any libraries needed, the linker used, and the compiler, along with the versions of each---again, this is all really basic stuff). It goes without saying that compiler flags are listed as part of this.
I think the main point one should take away from AMD's initial look at ZEN is that it will be roughly competitive. And that is all AMD needs to get at least 10% market share ( measured in Revs, and not units) in high end PCs and Servers.
In order to understand what a 10% market share means to AMD please read the following article:
AMD: Zen =10% Market Share in Servers/PCs + VR/AR/MR= $75/share
Nopesiree, not happening yet because we have not seen any pre-announcements from major sever and supercomputer hardware manufacturers on the adoption of AMD's Zen for future upcoming servers and supercomputers (which meant something is amiss with AMD's Zen). Just look at Intel's Knights Landing as an example, many months before launch already many major server and supercomputer hardware manufacturers (e.g. Cray) pre-announced upcoming products that will be using the new chip. And AMD's Zen being so much "closer" to (paper) launch with just few months away, we have yet to see any of those pre-announcements (especially from companies like Cray, SuperMicro, Tyan, Lenovo, Huawei, HP, Dell, etc)...
It's a pity they chose a multithreaded benchmark. Single threaded is more important. Zen only has to get close to Intel's single threaded performance to be successful.
"It's a pity they chose a multithreaded benchmark. Single threaded is more important. Zen only has to get close to Intel's single threaded performance to be successful."
Why? Both products are 8C/16T CPUs so if MT performance is similar their ST performance should also be similar. Nothing suggest that ZEN will gain more from running the second thread than Broadwell. Neither CPU is designed like the Power8 that aims to achieve the best dual-threaded and quad-threaded performance. Power8 gains something like 50% from the second thread but nothing suggests that ZEN can do the same. I think gains from the SMT will be very close between ZEN and Broadwell.
Pro tip: if your article can be summed up as "we have no idea what the answer to the question posed by the headline is", then it's clickbait, not an article.
A lot of articles out there aren't stating the obvious, or getting the facts mixed up, or calling things by the wrong name and extrapolating that into what the results mean. Sometimes the obvious needs to be said.
While you are arguably correct, many people, probably you included, would be yelling at the lack of reports on Zen performance that is available on other tech sites. It's always best to take single results with a grain of sand, that's why AT runs a benchmark suite, not just one. I appreciated the article, thanks Ian.
I have to admit I am a bit "salty" around the tone implied in the article itself. There's no depth to it other than just question blindly a single benchmark not targeted to consumers. Blender is a FOSS project, visible to anyone, but here the piece revolves around the ways you can "cheat" around it, other than exploring how those numbers come to be.
I don't appreciate opinionated pieces when they don't put facts before arguments. It does gives perspective, but only from a biased POV; or that is the conclusion I get. Even if I don't disagree with the statements, they don't seem objective (weirdly enough).
"there could be CPU vendor specific optimizations in either the code or compiler"
how is this related to Open Source? /any/ software program can be setup like this, you could even make the opposite argument, if the code isn't tuned for the specific CPU then you're not taking advantage of the CPU properly.
Considering no one is writing code for Zen yet, it's only logical to use a codebase that can be modified and optimised if performance is biased against it.
As long as AMD did a fair comparison (optimised for Zen on Zen, and optimised for Broadwell-E on Broadwell-E), I don't see how this has /anything/ to do with OSS.
I suspect this benchmark was chosen because it threads well, looks good (name another benchmark that you can actually /show/ in real time) maybe Zen is pulling more out of the SMT implementation, while being slightly slower in the IPC department.
I suspect AMD wants to push multicore, gaming loves the "4 core" CPU, and the industry is too lazy to write code for CPUs that don't really exist yet, though with the PS4/Xbone pushing 8 cores, Vulkan and DX12 here, I suspect the next gen of gaming is scaling past those 4 cores.
The difference is that at AMD they dont say it is actual gameplay or video playback. At the Intel presentstion he's pretending to play the game, fails to play the video. And tries to cover up his bad mimicing that by saying its been played backstage. While you can clearly see it is a video which is also laggy.
Honestly as an AMD fanboi I have to say that I dont see how this article was antagonistic or really biased. It neatly points out problems with the benchmark and essentially says it's not an accurate metric and to not get your hopes up just yet. Wait for more valid benchmarks.
Not sure where the bias is either.. It hints at possibilities.. but nothing verifiable. My takehome on the article is it's possible Amd has a very competitive chip coming out..
I don't like Anandtech. I could say that cinebench also helps intel and it's used in this site to measure wich CPU is faster and also this article praises the cinebench as a reliable source of conclusions.
BUT, besides cinebench and anandtech, this article is OUTSTANDING. I agree in everything as an AMD fan. I'm having a bad feeling about ZEN and i fear it's the same fiasco bulldozer was.
I "upgraded" from an FX8350 to an i5 6600. It boots up faster, for sure, by about 10 seconds. And er, I get like 5 more FPS in a spatter of games.
Do I feel mugged? Yes. Yes I do.
Will I be buying Zen regardless of how competitive it is, yes I will.. I tried the Intel side and found it left a bitter taste in my mouth/wallet. I gave my GF the fx8350, I've offered to trade systems but she said now.
I have zero intention to sit here with a PC that constantly disappoints me. This one will be sold off at a cheap price to anyone looking to get into PC gaming.
I'll use that to fund an AMD Zen and Vega build and go nuts, build my dream pc.
That's some severe invasion by AMD fanboys. The only question is, how many of them are being rewarded by AMD for posting such idiotic attacks and how many do it for free. (Not fist time either...)
To play devils advocate, I disagree with the mention that the lower bound of the high end part is 3GHz for Zen. There is nothing to suggest that. To milk the benchmark AMD could have run it with a stouter cooler and upped the voltage to juice it to 3GHz. Or 3GHz could be the turbo speed, it was a pretty short benchmark. I'd assume that the chip IS capable of 3GHz, for awhile at least, but that might also be with OC'ing. I think that is the most that it suggests.
I'd think AMD wouldn't want to milk things TOO much, so I'd expect the 3GHz is probably at most a turbo speed which it can handle with a stock cooler for the duration of the test without excess thermals. It also could be within the base clock performance. So that means that the base clock is probably somewhere between 2.6-3.1GHz for final chips and turbo performance is probably 3-3.4GHz as a random guess.
As for actually IPC, I don't think it says a thing other than custom benchmarks can be made to make Zen look good. That said, I have some hope here. I don't need it to be equal to or better than Broadwell-E. If it can achieve 80+% of the single thread performance of Broadwell-E and at the same time can beat the performance or at least equal the performance of a HEXAcore Broadwell-E for a Zen 8 core, 16 thread part at somewhat less platform price, AMD has a sale out of me.
I have to agree that I don't know what the point of this article was. It was always clear it was an isolated quick test and didn't necessarily mean anything. That great warranted maybe a paragraph, the rest is just wasted words.
People here should realize that AOTS cant take advantage of More then 8 threads and Zen has double of that. Plus engineering sample Zen has lower clockspeed per Core then full fedget 4790 so no Wonder it loses in single Core performance and thats still what games want more. Now something like 5960x or 6900k would have shown similar performance like Zen engineering example at same clockspeeds. You must realize that currently 4c 8t i7s perform better in gaming vs any 8c 16t simply because having More cores means lower clockspeeds and clockspeed is the käy to single Core performance if IPC is the same. I5 Lost simply because it didnt have enough threads but decent clockspeed. Also Zen had 3,2ghz turbo and I7 had 4ghz or was it even More. So no surprises here Zen is delivers close to skylake depending on the spesific tasks. 1. Zen is not at it's final clockspeeds 2. Zen is More comparable to 5960x or 6900k both in terms of multi threaded tasks and gaming 3. Reqular i7s beats higher end i7s in gaming simply because higher turbo clockspeeds 4. Zens ipc is not 100% confirmed. 5. Zen is going To make a difference By offering similar performance compared to Intel By lower prices. 6. No matters the similarities both architectures have pros and cons.
I would like AMD to do well again, but I can't really trust their marketing. Two things make me suspicious:
1. AMD has been in the red for many quarters in the last few years. I can't really seeing them having the R&D time and money to create a competitive product other than Intel getting lazy. If they did it, hat's off, but I'm sceptical.
2. AMD's benchmarks and marketing have always been selective and dodgy. I remember when the FX series launched and AMD released a bunch of ridiculous gaming benchmarks at high resolution/high AA claiming similar performance to Intel CPUs (because they benchmarks were highly GPU bound, obviously, taking the GPU out of the equation).
I no more believe AMDs numbers at all, they promised than Polaris would much more power efficient than Maxwell, it was 100% lie, sadly all so-called analysts dont care about promises vs. reality checks.
You might want to read that article you linked again. The only time maxwell is mentioned is when they compared what must have been the 460 to the gtx 950. Other than that, it only discusses that Polaris is a big leap in performance per watt over previous gen Radeon GPUs.
Hey it could be worse, it's not like they used wood screws on a fake card with the ceo waving it at investors and claiming it to be real. Wood was fitting however because when Fermi was finally released it ran so hot that you could Bar-B-Q a T-Bone on it. How come you trust nVidia after that fiasco? Or what about that gong show where nVidia renamed the GeForce 9800GTX+ as the GTS 250 and increased the price by 30%? Or wait, remember when they were sodomising their own board partners to the point that BFG died? Is it any wonder that XFX and PowerColor flipped "dear leader" the bird and started making Radeon boards? That was unprecedented as no board partner had ever left ATi. Or maybe the debacle that ensued when Charlie Demerjian exposed the way that nVidia was threatening to deny samples to review sites that didn't use only "nVidia-approved" games in their reviews (read: games that performed better on nVidia hardware) because they tried to do the same thing to theinquirer.net and he wouldn't have it. Either you're too young to have seen all that (around 2009) or you have the memory of a fruit fly. My point is, sure, AMD has done some stupid things but when it comes to all-out dick moves, nVidia has ATi outscored by a score of ten to one at least. Hell, they redid their driver package to disable PhysX and CUDA if ANY ATi graphics products were detected in the system. Ask yourself now why you're so quick to condemn AMDbut not anyone else? "LET HE WHO IS WITHOUT SIN CAST THE FIRST STONE!" (I read that somewhere... Heheheh)
Can't wait to buy a 4C/8T Zen to go with my RX 480. Should be plenty fast enough to drive a single card in DX12/Vulkan/VR. Which is my primary reason for still having a PC.
As much as I like the sound of having 8C/16T, I couldn't really justify it for what I want to do. So as long as the 4C/8T is cheaper than a 4C/8T i7 (and ~as fast) I'll go with the 'mini-me' Zen, and save some money. If I change my mind, or find some use for it, I can always sell the 4C/8T and buy the 8C/16T anyway.
All I really want from *any* CPU is that it doesn't bottleneck my graphics card. So I think a 4C/8T Zen should be fine, as long as they don't want silly money for it. Like $300...
Could you be more of a troll? "primary reason for still having a PC" Like having a PC is such a frikin big deal. It's hilarious really. What else are you STILL clinging to for no real reason? Life?
Going down the optimization for specific CPU microarchitectures leads on to another more philosophical issue?????
Are you serious?! LOL - that's the whole point of an umm, compiler (and compiler optimizations, differing instruction sets between cpus, etc.)
This would be like saying Intel was "cheating" because they used AVX in their processor before AMD did, and a developer used the extensions in their code.
Simply ludicrous to think like this; software can (and sometimes should) be optimized for a specific processor even though the general architecture is the same... That's not "cheating", it's called "using something as designed"
Why do you put a question mark in this article's title when you give NO INFORMATION suggesting otherwise (e.g. AMD's processor is certainly not faster)?
This is almost buzzfeed quality "journalism" - nice job.
An FX-8350 (4 Ghz) is competitive with an i7-2600K (3.2 Ghz) in blender according to Anandtech's benchmarks. Assuming a claimed 40% and 25% measured IPC improvement between generations respectively, this result isn't a surprise or very meaningful outside of Blender users who have cheap electricity.
I appreciate the re-situating of a media event. But isn't the fact that AMD performance tests are being contextualized instead of receiving the same head shaking of the last 8-10 years acknowledging some press release as verification that AMD still has a pulse, albeit not a competitive one, reason for excitement enough? I also heard that the Intel testbed mentioned in this article was farther from the HVAC system and situated at a slightly lower elevation than that of the AMD Zen. Take it as you will... Just saying,
I'd wait to see Zens prices before saying that it isn't competitive ? The cheapest 4C/8T i7 is ~$300 afaia ? If AMD made the 4C/8T Zen nearer $200 then I believe it would be more than competitive. Even if it was only 95% as fast.
General consensus of opinion is that the 4C/8T should be about the same size as a P11 die, ie ~120-130mm^2. Given that you can buy an entire RX 460 card (including P11) for ~$110, then $200 would seem entirely doable ? As going by P11 pricing I doubt a die that size costs more than ~$50 to make, if that ? Assuming that to be the case they could sell it for $200 giving the retailer a 100% markup, and still make 100% gross profit margin.
Would a CPU 95% as fast as a 4C/8T i7 for gaming and VR, but costing ~33% less, be competitive ? I think it would, because most people would probably prefer to spend the extra $100 they saved on a better graphics card ? I know I would, because it would be pretty much guaranteed to be more than 5% faster than one costing $100 less.
There are a lot of things going on with Zen, mostly for the better. First up, most people familiar with AMD love the fact that AMD doesn't replace sockets every time a developer passes gas the way Intel does. This means that a socket AM4 motherboard will support AMD processors for at least the next two years for upgrades, and the socket AM4 chip they might buy today will work in their next motherboard if they go AM4+ in the future. No need to worry about different sockets for APU vs. pure CPU, and you could start with an APU and then upgrade easily if you wanted to.
DDR4 support is a plus.
Now, overall performance. Engineering samples tend to be on the slow side in terms of clock speed. The entire purpose is to let hardware partners develop boards, so clock speeds are a non-issue. They could be 500MHz for all AMD cares when it comes to these engineering samples, because the PURPOSE would be met.
Demonstrations are designed for different purposes. Early demos are more to show "hey, things work, it isn't just vapor products we are talking about". Close to launch, we will hear and demonstrations showing the chip running at 6GHz on liquid nitrogen will come out to hype the product LAUNCH.
Remember, we can't expect to see product available for sale until at least November/December at the very earliest. If AMD were to release to distributors in October(doubtful at this point), it would take a good month for those to make it into the hands of consumers.
Chipsets are a big part of performance testing. Working around problems by motherboard manufacturers tweaking settings should be expected right now, so who knows for sure how much of what we are hearing about is real-world performance. Expect that at the same clockspeed, we will probably see at least 10 percent better performance as debug code is disabled.
So, we are at the end of August. One more month until we find out how yields are, which will determine official launch dates. No matter what, AMD will finally have a new generation core, and that means better performance. For those who are being critical, how many love those dual-core i3 and i5s that are all over the industry that SUCK compared to a quad-core A8 or A10 in the same price point for a system? At the high end, where quad core and oct-core processors are the norm, it is easier to say that AMD has not been competitive, but in the low to mid range, I'd rather go for a quad-core AMD over a dual-core Intel any day.
I don't put much stock into that Blend benchmark win but even if it was optimised for Zen it doesn't matter, an engineering sample did beat a market-grade Broadwell-E in a benchmark. (Regardless of Anandtech's feeble attempt at Intel damage control proven by this worthless excuse for an article made up almost completely of semantics and political-style rhetoric.) At the end of the day, it's really this simple: IF the chip was the POS that all the Intel shills are claiming then NO amount of optimisation would help a mere ENGINEERING SAMPLE narrowly defeat a MARKET-GRADE Broadwell-E with the SAME NUMBER OF CORES at the SAME CLOCK FREQUENCIES in ANY benchmark. It would have to have been completely faked. (The use of caps lock was to emphasise only the objective facts that not even the author of this "article" dares to dispute.)
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
177 Comments
Back to Article
powerarmour - Tuesday, August 23, 2016 - link
Give them their moment in the moonlight, it's not like Intel have had a monopoly for the last 10 years or so is it. This is still a huge step forward for AMD, and the price/performance could be just what we need.ddriver - Tuesday, August 23, 2016 - link
I don't see any point in this article full of speculations besides a loud "intel pays us off" statement. Just wait and see.Michael Bay - Tuesday, August 23, 2016 - link
Hi there, AMD damage control team.powerarmour - Tuesday, August 23, 2016 - link
Hi there, Fanboyus McTrollius.prisonerX - Tuesday, August 23, 2016 - link
If only there was a brain damage control team for you.ddriver - Tuesday, August 23, 2016 - link
For his condition only amputation produces any results.Michael Bay - Tuesday, August 23, 2016 - link
Here`s your pity comment.close - Wednesday, August 24, 2016 - link
Dudes, AT presented the results of AMDs launch and it was their duty to present these caveats. Otherwise, if AMD messed with the benchmark, there will be plenty more people complaining that AT presented some skewed results without even a mention of the caveats.Michael Bay - Wednesday, August 24, 2016 - link
Nah, I`m completely okay with how AT presents things. It`s just that ddriver has terrible paranoia of all things intel.ddriver - Wednesday, August 24, 2016 - link
Oh wow, a suckup fanboy wannabe approves paid publications, that's a new one :DMichael Bay - Thursday, August 25, 2016 - link
Case in point.Ranger1065 - Saturday, August 27, 2016 - link
Striking similarities in their unrelenting Nvidia/Intel shillage lead one to believe that Michael Bay and Chizow might indeed be one and the same sad individual.SageFox - Wednesday, August 24, 2016 - link
He could be right and i wouldn't bet about anandtech honesty, since they talk about credible benchmarks and blender not being reliable, but cinebench is intel favorable and polaris benchmarks are crap.But he is also wrong if he thinks this article is wrong and AMD is playing with fire, or already burned to the ashes before we realize it.
As anandtech says, AMD moves are weird, marketing is weird.
Ashes of the singularity leaked benchmark show a WEAK zen. ZEN performs there in the leak at 58 FPS core score. 6900k at does it at same setup and settings at 127.
Also Zen with 16 threads is just faster than 4690k (4 threads), but slower than 4790k 8 threads. Adjusting maths by clock per clock, still looks like ZEN is 40% slower than intel core per core, thread per thread. FX 8350 does half what ZEN does, at same clock speed BUT ZEN have exactly twice the threads piledriver has. So ZEN, assuming that leak is true, is just a piledriver chip with a lot less TDP and just more cores. This is what i'm seeing.
Why this AMD presentation came out after that leak, is also interesting.
My question is, who is lying, AMD or that AOTS leak?
tamalero - Friday, August 26, 2016 - link
what frequency was that zen?remember that in the demo test for rendering, they dropped broadwell-E's frequency to match Zen's.
What if that "leaked" thing was on an even lower frequency rates and the 4790k was on full turbo?
Gastec - Thursday, September 8, 2016 - link
There is no "leaked" Ashes of Singularity benchmark because THERE'S IS NO such benchmark done for a Zen CPU, it has NEVER been done. It's all a media shill, click bait.Cooe - Sunday, February 28, 2021 - link
The future says - "Hahahahaha look at this idiot! xD"cknobman - Tuesday, August 23, 2016 - link
The tone of the article (until the very end) more or less came off like that to me as well.powerarmour - Tuesday, August 23, 2016 - link
Indeed, the tone is completely antagonistic. Shame.melgross - Tuesday, August 23, 2016 - link
More like realistic.mdriftmeyer - Wednesday, August 24, 2016 - link
More like a petulant child pissed off that these overpriced pieces of silicon will be boat racing Intel in almost every category, and meeting at the finish line with their Xeon systems. In short, Ten years of gap have been closed; and that pisses them off.anubis44 - Wednesday, August 24, 2016 - link
Fukc 'em. If Anandtech and the rest of the pro-nVidia/Intel cabal can't stand that AMD is rising up to destroy both of these larger companies, they can suck d0nkey balls.Ranger1065 - Saturday, August 27, 2016 - link
Well spoken :)Gastec - Thursday, September 8, 2016 - link
I bet you are one rich troll with a Titan X video card ;)SageFox - Wednesday, August 24, 2016 - link
I see anandtech like an intel fanboy who this time is right, just this time.Sorry guys, anandtech is right this one. And is good, we lose nothing taking precautions. If ZEN is a success, we all and even some intel fanboys, will buy a ZEN chip.
I just think we are getting close to dark ages of tech and gaming.
ddriver - Wednesday, August 24, 2016 - link
Not strictly an intel or ms or apple fanboy, they are fanboys of money, and amd just can't afford to pay everyone off because it does so poorly at business.sonicmerlin - Wednesday, August 24, 2016 - link
I'm still waiting for the 2.8x performance/Watt improvement in efficiency that Polaris was supposed to bring. Where is it???tamalero - Friday, August 26, 2016 - link
its there for the 470 in directX12.Cygni - Tuesday, August 23, 2016 - link
The entire tech internet is about to be talking about these benchmarks, benchmarks people have been waiting for for what, 5+ years? And you don't see the point in this article...?ddriver - Tuesday, August 23, 2016 - link
The article claims amd speculations on basis of speculations. It is pointless, stupid and hypocritical.This is not benchmark, just one random test amd made, without making any concrete performance claims relative to intel. What amd said is 40% IPC improvement, which given how bad previous amd designs were, is 100% possible.
Nobody said it will be "2%" faster than broadwell-e, so it brings the questions what's with this article trying to convince us in the opposite without containing a shred of evidence or fact... I know AT standards are fairly low, but that's low even by those standards.
loveroftech - Tuesday, August 23, 2016 - link
I can't stand all the people who came from the dailytech comment section. That site had trash comment sections towards the end and now they came here to ruin this site's comment sections.Samus - Tuesday, August 23, 2016 - link
The problem is our expectations are so low for AMD these days that it's hard to believe they will actually ever be competitive with Intel again...but Intel IS getting lazy. As far as I'm concerned they haven't had relevant IPC gains since Sandy Bridgedescendency - Tuesday, August 23, 2016 - link
I've done a lot of reading recently on SB-era Xeon servers (which can be gotten on Ebay for cheap now) with 2 CPUs and tons of RAM to do some highly threaded workloads. Modern Broadwell-E chips (in real workloads) in single chip configurations beat them. Not by a lot, but enough that Xeon's based on Broadwell-E are far more appealing to that market.So the claim about SB era chips having nearly the same IPC as modern chips is empirically false outside of synthetic benchmarks, like you see on review sites. This is why reviewers like the ones at AT have to do a good job of explaining the results.
Nagorak - Wednesday, August 24, 2016 - link
He said meaningful improvements not none. Yes for servers or situations where absolute max performance matters you'd rather have a Broadwell. The issue is that for standard uses, including gaming, a Sandy Bridge is nearly as good. If it's slower, it's by a small enough margin that it frankly doesn't matter.powerarmour - Tuesday, August 23, 2016 - link
Look at you, complaining about one more turd floating in your cesspool.euskalzabe - Tuesday, August 23, 2016 - link
Yup. AT isn't what it used to be anymore. The quality is still there, but it's much less frequently seen and all the vultures have come to stay... :(YellowOnion - Tuesday, August 23, 2016 - link
ddriver, I think you overestimate people's ability to be rational, people will fill the void with their hopes and dreams and change what is a essentially a "we got a not-shit CPU" announcement, with "Zen pwns Intel in to the dust!~!~!", because it won a benchmark by a whopping 2% (which looking at the video on youtube, is easily calculated).Just look at the fallout from No Man's Sky, Some thought it was better to DDoS reviewers websites than face the fact that /someone didn't like the game they liked/.
ddriver - Wednesday, August 24, 2016 - link
Nope, I just make a clear distinction between "people" and fanboys/wannabes - and nobody gives a damm about the beliefs of the latter. It is obvious that everyone in the right mind would hope that zen turns out ok for the sake of competition and prices, but the point is amd claimed no concrete performance figures, they just demoed a running chip using a single app against identically clocked direct competitor - only a complete idiot would take it as a concrete indicator of performance, and only a fool would immediately launch an article to "refute" those non-existent claims. It is funny that the bulk of the suckup readers of AT think very highly of their intellect, yet they would need such articles to see through "amd's deception" LOL. And with claims that Blender being open source being bad and such other absurdities. I guess if it is not closed source test that can hide the `if (cpu != intel) doBad();` lines and everyone can browse the source to see what's going on exactly that's just not an OK test.Cygni - Wednesday, August 24, 2016 - link
Sorry you apparently didn't like the article that made it incredibly clear what was going on, which is, as I said, already the topic of much conversation all over the place. By the way, AT has posted many many articles that quote internal numbers over the years and always come with a disclaimer of their source. Including this one! How wacky-wild!Oh also, if you are so pouty about AT's standards, be a big boy and go someplace else? Nobody cares about whatever stupid brand fanboy angle you are trying to play in 20-goddamn-16.
Byte - Tuesday, August 23, 2016 - link
Yah, we need this hype to build, so they they can sell out of Zen and also sell "founders" editions for a few hundred. Give some infusion to AMD to build some competing products. Getting bored of Intel and Nvidia always winning.stimudent - Wednesday, August 24, 2016 - link
Yes.American made vehicles still suck!
medi03 - Wednesday, August 24, 2016 - link
Agreed on full of speculations part, but "intel pays" part makes no sense in this context/market situation/etc.BlueBlazer - Wednesday, August 24, 2016 - link
Why don't we speculate why AMD went from a closed source Cinebench R15 (as often seen in AMD's presentations for IPC measurement) to using open source Blender (which as the article stated, can be modified and recompiled)? Thus why didn't AMD show the standard default Cinebench R15 benchmark which can be easily compared with already existing results?Michael Bay - Wednesday, August 24, 2016 - link
Maybe they were trying to show Zen doing something in an actual production software(sorta, it`s Blender) instead of synthetics.Then again, they should have gone with Maya or something.
kn00tcn - Wednesday, August 24, 2016 - link
um, cinema4d is actual production software (even with a realtime asset link to adobe after effects for years), so no cinebench is NOT syntheticblender got good, people are using it, high end pros probably look for gpu rendering instead of cpu
N Shaftoe - Tuesday, August 30, 2016 - link
Blender(Cycles) does excellent GPU rendering with CUDA.tamalero - Friday, August 26, 2016 - link
I dont see whats wrong with that. Intel has modified and manipulated benchmarks themselves.Anyone remembers the multithreaded Pi and chat programs scandals? where intel pretty much sabotaged the multithreading if it detected a non intel processor.
descendency - Tuesday, August 23, 2016 - link
Our win is not in AMD challenging Intel (in the ultra high end) but AMD challenging Intel in the midrange. Intel currently has two main challengers to their HEDT platform. Old Intel HEDTs (you have to make enough of an improvement to warrant people buying the new model) and Intel's midrange platform (the I can pay $200-$400+ less and still be satisfied platform).AMD challenging Intel's midrange offerings will add a 3rd competitor to that market... one that isn't Intel.
So regardless of how fast these chips are against modern chips (which will be Kaby Lake by the time this launches), it also could put significant pressure on the upper tier as well (Broadwell-E).
retrospooty - Tuesday, August 23, 2016 - link
I dont disagree with anything in this article, but it's sort of a given to not listen to AMD's performance claims ahead of launch. Pretty much every new chip design they have released over the past decade has been alot of hype, then delays, then thermal issues forcing them to reduce the final clock speed, then actual products that fail to live up to the claims... Without fail every single time. AMD is capable of a home-run here, and based on the hype it seems like this one is finally going to be competitive this time fore sure right? Either way, none of it means anything until retail units (not engineering samples) are independently tested.novastar78 - Tuesday, August 23, 2016 - link
Not this time, they absolutely do have a serious chip on their hands. This is not like the previous few generations. They brought in one of the best designers in the world (in history for that matter) to do this.AMD are just taking their time and leaking small details because they want to get everything right with this as it is a very significant launch. Maybe the biggest in their history.
Also, let's put something into perspective here, they already have several future generations mapped out for Zen, as per the designers instructions. They are not, and have not, claimed that the first generation will outperform Intel's top chips at this point. They are clearly looking to get back to being competitive from a price/performance standpoint and fix the outstanding issues of TDP/IPC.
I have family/friends who work for the company, and they do truly have a great product, that will be even greater moving forward. They just need to get the launch and marketing right as this has always been their weakest point.
They need to manage expectations which is why they are not talking about numbers for the most part right now.
nathanddrews - Tuesday, August 23, 2016 - link
Unless AMD intends to shake up the Performance/Dollar ratio and throw Intel from its commanding perch, it won't matter how great the product is. Currently, Intel wants $1,700 for its top HEDT chip and $1,100 for the 6900K that AMD used in the benchmark.The real question is whether AMD will simply "fall in line" with Intel's pricing in order to maximize profit or shake things up and expand market share. I think they need the latter more than the former.
versesuvius - Tuesday, August 23, 2016 - link
Very true. If it is as good as Intel with the same price, there is no need to choose AMD. It has be at least %5 better than Intel in all the benchmarks or metrics and it has to be at least %20 cheaper than Intel. That would make it competitive. Of course by 2017 Intel has already made all the initial high profits that it would usually make on its latest chips and will be ready for a fight both in terms of new offerings and the price. However, if AMD really can make a good impression with Zen in 2017, the real benefit to the consumers will start arriving in 2018 and 2019. That is when Intel will have to move some schedules forward or in other words start ticking for real.lilmoe - Tuesday, August 23, 2016 - link
"It has be at least %5 better than Intel in all the benchmarks or metrics and it has to be at least %20 cheaper than Intel"I'd have to disagree. While I truly wish they keep pushing prices down and value up, I'm sure many OEMs and informed consumers would prefer the AMD route, especially for ultrabooks. Intel's iGPUs are nothing but brown stuff; Sure they're "OK" for light workloads and some *old* titles with low settings, but they're by no means good satisfactory. Hell, even their HDMI implementation doesn't support 5.1 audio. AMD will BURN them there with integrated Polaris GPUs with MUCH better performance/watt, better multimedia, and MUCH better drivers.
Your statement has some truth to it in regards of consumer perception. "Core i7" does have a ring to it. I seriously hope AMD does marketing right, and reviewers/sites/blogs are truthful about the value (AND performance) the consumer is getting with AMD chips.
wolrah - Wednesday, August 24, 2016 - link
"Hell, even their HDMI implementation doesn't support 5.1 audio."What are you talking about? According to this very site (http://www.anandtech.com/show/2622/3) Intel's GPUs have supported up to 8 channels (aka 7.1) over HDMI for over 10 years (since G965). Before I moved to a RPi3 my HTPC was running on a SNB Celeron and definitely did 5.1 just fine.
lilmoe - Wednesday, August 24, 2016 - link
Definitely doesn't support 5.1 over HDMI in my experience with various PCs....Schecter1989 - Wednesday, August 24, 2016 - link
Well having plenty of machines do the same exact same feat that Wolrah used his SNB celeron for, I can definitely say it is supported and you obviously just don't know what you're doing. There is such a thing as user error.jengel - Wednesday, August 24, 2016 - link
This. If Zen is anywhere near Intel in performance it will have them shitting down one pant leg. This is because they won't stand a chance against AMD's Zen APU lineup. My ancient Llano APU has an integrated GPU that's just as fast as a discrete R7 240 SFF. All the interconnect/HSA work is already done. All they needed was a viable X86 core and now they have it.It's my belief that the massive layoffs by Intel recently were a direct result of Zen hitting sampling stage. Intel knows they're screwed, and there's nothing they can do about it.
fanofanand - Wednesday, August 24, 2016 - link
It's a bit silly to think Intel is laying off workers because they are so petrified of what AMD is bringing down the pipeline. Far more likely (and documented) is that Intel has been replacing high-cost American workers with significantly cheaper H1-B VISA employees that will work for 1/3rd of the pay. The layoffs had everything to do with maximizing profit and nothing to do with fear of Zen.fallaha56 - Tuesday, August 23, 2016 - link
it will be 50% of the priceand prob 90% of the performance of Skylake-E by the time it gets here (AVX-512 to one side...)
but AMD will make a killing
LancerVI - Tuesday, August 23, 2016 - link
If your statement comes to fruition; then that would be an absolute win for AMD.They would indeed "make a killing." I don't foresee replacing my 5820k anytime soon, but at least when I'm ready, there'd be some options.
IF IF IF. I hope it works out for competitions sake.
thunderwave_2 - Wednesday, August 24, 2016 - link
It's this kind of baseless speculation that causes AMD's products to not meet expectations. We have no idea how much the 8c/16t Zen will cost, and we might not be able to trust AMD's benchmarks for reasons stated in the article. Setting your expectations so unrealistically high will only result in disappointment.Cooe - Sunday, February 28, 2021 - link
Winner winner, chicken dinner. This was about dang on accurate for the OG Ryzen launch.wumpus - Tuesday, August 23, 2016 - link
Except the price of the $1,700 chip is simply because it is "the best". Intel will happily sell you a $300 chip that gets 90% of the performance, but doesn't give you bragging rights of 'the best'. If AMD can't match the $300 chip, then don't expect the profits (the $200 i5[K] gets close to the $300 i7[K] and things go downhill from there (last I saw, the 8320FX was "competing" with the i3).It gets even worse with the economics of Naples. The big Xeon chips really rake in the money for Intel, but the cost of the systems that they are in are so high that a 90% AMD solution simply isn't worth putting in the expensive system even if it is free. AMD has to completely beat Intel at *something*, *anything*. Preferably something that they can convince about 10% (or whatever GoFlo's capacity is to Intel's) of current customers to abandon Intel and go with AMD. Maybe Blender can do it (although it doesn't seem likely that AMD will beat Intel's clocks, which appears necessary in this benchmark), but I would recommend something like using SSD as main memory (preferably using HBM as "cache', although that largely depends on AMD's experience with the Fury). While this wouldn't appeal to everyone, massive cheap "memory" should at least appeal to enough server buyers to put AMD back in the game. The biggest danger is Intel swooping in with 3dXpoint and re-taking their lunch back, but AMD really doesn't have all that many options.
ddriver - Tuesday, August 23, 2016 - link
So which 300$ chip comes within 90% of the performance of the 1700$ chip?KenLuskin - Tuesday, August 23, 2016 - link
You have No idea what you are talking about. AMD only needs a roughly competitive server chip, they do NOT have to beat Intel.In PCs, AMD's graphics are 50% better than Intel.
Because ZEN is roughly equivalent with Intel, AMD's APUs will destroy Intel's integrated graphic PC chips.
Samus - Tuesday, August 23, 2016 - link
The only reason AMD APU's smash Intel graphics is because they devote twice the die area to the GPU, at an obvious cost of performance to the CPU, but who knows if adding more cache, etc, would even help their architecture.descendency - Tuesday, August 23, 2016 - link
"In PCs, AMD's graphics are 50% better than Intel."AMD devotes a much higher percentage of the die than Intel does and (counter to what you might expect) Intel has a tighter integration between their CPU and GPU (on die) operations. Basically, Intel started integrating GPUs on die before AMD.
The fact that Intel is even close to AMD is actually amazing, given those constraints. But the reality is that Intel spends something like 3 times what AMD does into R&D.
That said, for GPU optimized builds (that you don't want a discrete card), AMD has the best offering.
bcronce - Tuesday, August 23, 2016 - link
Correction, Intel spends about 50x more on R&D than AMD makes in revenue.Intel R&D: ~$55bil/year
AMD Revenue: ~$1bil/year
mkaibear - Wednesday, August 24, 2016 - link
Er, no. Intel's full year revenue for 2015 was $55bn. AMD's full year revenue for 2015 was $3.99bn.versesuvius - Wednesday, August 24, 2016 - link
CPUs are not GPUs. AMD is already years ahead of Intel when it comes to GPUs and multimedia. AMD should not waste any space for multimedia on Zen than it has to. Just keep it close to what Intel does and that is more than enough. In other words just chase after Intel when it comes to the GPU part and let Intel worry about staying ahead on that front. Later on and in short order AMD can add enough multimedia capabilities for laptops and home entertainment devices on its CPUs. For now the CPU part is the most important part to focus on.fallaha56 - Tuesday, August 23, 2016 - link
lol absolute gibberish sorryhow many people own $1700 CPUs, not even me
as to $300 -as ddriver said, which chip are you talking about?
AMD wil have a 4-core iGFX Zen that is going to piss on Intel iGFX -have you seen PS4?
retrospooty - Tuesday, August 23, 2016 - link
"Not this time, they absolutely do have a serious chip on their hands."- But they say that every time. Phenom, Bulldozer, piledriver, steamroller, excavator and the others, every single bloody time. Also, where is the retail chips? Already delayed until 2017.
I really hope its a great chip, we can all benefit from the competition, but I do not believe anything AMD says. It may well be a great chip, and when retail units are independently tested I will believe it and not a second before that happens. I also would highly recommend no-one buys anything until retail units are independently tested. Too many
retrospooty - Tuesday, August 23, 2016 - link
/edit too many false claims to take it seriously.andrewaggb - Tuesday, August 23, 2016 - link
This is a new architecture so there's an actual chance it'll be way different/better. Over the last many years we've seen the same cpu's and gpu's with minor yearly tweaks from AMD (and intel to be honest) and minor tweaks aren't moving the needle much. This is a major release and it could be something....But AMD doesn't seem to be announcing anything new/groundbreaking. With the Athlon 64 they moved the memory controller on-die and went with a shorter pipeline. It was up against the P4 that was a very different design. At the time it was a big deal. Zen doesn't sound like it's doing anything intel isn't doing, so I don't expect it to be faster than intel. I also suspect it'll use more power and clock lower due to fabbing issues and possibly design choices as intel has spent years reducing the power consumption on their chips. I'd love to be wrong but we need to see independently tested retail chips. Equal IPC isn't nearly enough. (And that's assuming the blender results apply to other tests as well).
I don't want Zen to fail, but I'm skeptical it'll be everything people hope.
rhysiam - Tuesday, August 23, 2016 - link
While I agree with your scepticism, one genuine option for AMD is releasing a cost-effective 8 core (16 thread) CPU. For years now Intel have been dedicating die area from process shrinks to integrated graphics, rather than larger CPU cores, or more cores. If you want more than 4 cores, you need to spend up on the HEDT platforms. Broadwell-E even extended the pricing divide between their 4 and 6+ core CPUs.Even if AMD can't quite reach the IPC of a Kaby Lake quad core, if they can get within 10-25%, but in a genuine 8 core CPU at a competitive price, that would still be a compelling product IMHO. If they choose to dedicate the die area to additional CPU cores rather than large integrated graphics, perhaps they could price a genuine 8 core 16 thread CPU similarly to Intel's high end quad cores.
For sure we need to wait and see though.
D. Lister - Wednesday, August 24, 2016 - link
"With the Athlon 64 they moved the memory controller on-die and went with a shorter pipeline."For the sake of historical accuracy, AMD didn't come up with the on-die memory controller. The feature had already been around for about a decade when AMD implemented it.
Secondly, AMD didn't just "go" with the shorter pipelines, their chips were based on the Pentium III (the one Intel had sadly dropped in favor of the Netburst P4), which had shorter pipelines, as designed by Intel.
melgross - Tuesday, August 23, 2016 - link
Yeah, yeah, yeah. We hear this every generation of AMD chips since Intel's Yonah. Nothing good ever comes of it.bill.rookard - Tuesday, August 23, 2016 - link
Agreed. This launch is crucial. They had a minor hiccup with the 400 series launch, but recovered quickly, and well - although I have to say they -had- to know what was going on before they released it.All in all, as long as they regain a competitive footing against Intel, that's what is really important. Not -beating- them necessarily (although that would really be a kick in the nuts to Intel), but just getting close which is something they haven't been in quite a few years. They don't design bad chips, but they've really missed the boat with thermals/TDP.
Another area where Zen will hopefully pick things up is in the far more lucrative server market. Opterons are almost nowhere to be found these days, and getting back into the server market with a Zen based Opteron again would be a great start.
KenLuskin - Tuesday, August 23, 2016 - link
Dr. Lisa Su and the ZEN design team ( lead designer was legendary guru Jim Keller) are NOT the same folks who bungled the post 2006 designs.Remember, AMD's success from 2003 to 2006, were based on Jim Keller led innovative designs.
So, if you are looking for some perspective, look to Jim Keller's successes, rather than AMD's NON Jim Keller designs.
fanofanand - Wednesday, August 24, 2016 - link
At least somebody gets it. Keller has designed this chip from the ground up, it is NOT a piledriver rehash. It will at least be competitive with Intel's offerings, and every human being alive should be grateful for that. A lack of competition breeds $1700 chips.C.C. - Wednesday, August 24, 2016 - link
Actually, Keller didn't design Zen "from the ground up." AMD stated that he was given free reign to expand upon their existing base for that was already in the R&D stages for Zen...Now Zen+ designs will most likely feature more of Keller's vision..Either way, I am very excited for this release. I have a 5Ghz (48+ hour torture test stable) 3770K coupled with a high end Z77 board...I have plenty of IPC, but I would like to have double the cores, and I would really love to have the platform updates (more PCI-E lanes, M.2, etc)..I think AMD is going to have a winner on their hands if they market and price it right..
All of the fanyboys expecting a $300 8c/16t cpu that is just as fast as a Broadwell-E SKU are going to be in for a rude awakening..I wouldn't expect AMD to price it any less then $600 vs the $1000 asking price from Intel. That would still be a great value, and allow AMD to have some much needed profit margins..
TekDemon - Saturday, August 27, 2016 - link
I don't think the design is bad but you do have to remember that they're hugely disadvantaged by their use of Global Foundries to actually make the chips. It's using Samsung's 14nm FinFET but from what I'm hearing this doesn't mean it's competitive with what Intel can do and not everyone is all that confident about GloFo's implementation. AMD is contractually obligated to use Global Foundries for a lot of stuff because GloFo is a spinoff of AMD. Zen might have been better off being made by TSMC or even Samsung themselves.TekDemon - Saturday, August 27, 2016 - link
Honestly I think this time around it's not really the architecture that's the issue but the fact that AMD has to use Global Foundries due to obligations leftover from spinning them off. I've heard from people who've talked to people who really would know whether Global Foundries is doing well with their 14nm process (licensed from Samsung) and the stuff they're saying isn't particularly reassuring. If Zen ends up not being able to clock up it may end up largely being because they're stuck with using GloFo instead of another foundry like TSMC for these CPUs.SirMaster - Tuesday, August 23, 2016 - link
Why are we comparing Zen to Broadwell? Broadwell is Intel's tech from 2014. Skylake's IPC is greater than Broadwell, and by the time Zen actually releases we will have Kaby Lake with even more IPC than Skylake.baka_toroi - Tuesday, August 23, 2016 - link
Currently there is no Skylake-E platform.SirMaster - Tuesday, August 23, 2016 - link
Why do we have to compare it to a -E platform?As a gamer I am interested in buying whatever CPU has the fastest per thread performance with at least 4 cores. Currently that's the 6700K and soon will be the 7700K. So to me and millions of people like me that's what AMD needs to compete against as well.
Not really interested in something with more cores but slower cores since it doesn't help my workloads.
r3loaded - Tuesday, August 23, 2016 - link
Because they're comparing workstation CPUs with a workstation benchmark.jjj - Tuesday, August 23, 2016 - link
Why do you think the world revolves around you?Anyway,if you only need 4 cores you can look at a 120-150$ Zen quad core with no GPU and decide if you pay 350$ for Skylake with a GPU you don't need or get the Zen based on w/e metric matters to you.
SirMaster - Tuesday, August 23, 2016 - link
"Why do you think the world revolves around you?"I don't. But I come from a community of millions of gamers with the same/similar interests as me.
I could ask you the same question and why ignore the millions of us?
retrospooty - Tuesday, August 23, 2016 - link
Its just a matter of comparing Intel's highest performing consumer CPU. That would be Broadwell E. AMD will have an 8 core consumer CPU that (they say) will compare to Broadwell E, Intels current top of the line.melgross - Tuesday, August 23, 2016 - link
Yep. As usual, one generation behind.wumpus - Tuesday, August 23, 2016 - link
When was the last time it really mattered? Ten years ago with the core2? Intel cores have pretty much topped out and have had tiny incremental boosts for a long time (although I've noticed that the 8350FX seems to have fallen from "competing with the i5" to "competing with the i3").If AMD could somehow compete with Sandy Bridge (which I *think* is the basis for all recent cores, there don't seem to be any visible uarch changes since then) they should be able to make some money. I'm not even holding my breath for that, Intel has simply been executing too well since abandoning Netburst and Itanium.
Intel999 - Tuesday, August 23, 2016 - link
"one generation behind"You did see that the demonstration, as pointed out by the author, was 2% ahead of Broadwell, yes?
Skylake, depending on the benchmark, struggles to show an IPC gain of 2% over Broadwell.
So on an IPC level, at first blush, one can determine that Zen is within a margin of error to being on the same level as Skylake.
Not to get though, I'm sure Kaby Lake will allow Intel to pull ahead by 2% again at the end of next year.
StrangerGuy - Wednesday, August 24, 2016 - link
A i5-6500 @ stock 3.3GHz with slowass DDR4-2133 beats a OC 4.5GHz 2500K + DDR3-2133 in real-time gaming tests at 1/3 the TDP.But it's OK, the Intel hatedom can keep chanting the "Intel did nothing since Sandy Bridge" memes when its demonstrably false.
fanofanand - Wednesday, August 24, 2016 - link
1/3rd? Source please, Intel has been increasing their TDP from 65 watts and are now at what 77 watts? So you are saying that despite the increases to TDP somehow your chip has magically decreased? Impressive to say the least, please share with the world how you pulled off this magnificent feat!fanofanand - Wednesday, August 24, 2016 - link
So what? Intel's generational increases have mostly been about i/o and connectivity, not about improvements to the core design. Most people that have SB still have no reason to upgrade, so it doesn't seem to matter much if it's one, two, or even three generations behind. AMD doesn't need to match or exceed KL IPC to make a dent in the market, they just have to be competitive in price/performance/watts.cocochanel - Tuesday, August 23, 2016 - link
You're right on this but very few talk about it. How much does the iGPU add to the cost of an Intel chip and do you need it if you already have a discrete GPU in your system ? And then there is the software. I was dumb enough a few years ago to buy a laptop that had a discrete GPU in it. Any time I was running Google Earth the program would crash after a few minutes. It turned out that both graphics drivers were running and conflicts arose. I had to disable ( uninstall the driver) the GPU to get the program running. What a pain in the a--.Audiophizile - Tuesday, August 23, 2016 - link
A massive majority of i5 and i7 users are business and Soho applications then gamers. All of these applications currently benefit much more from single core speed up to 4 cores and occasionally going up to 6. Meaning a very large majority of users will see 0 benefit or even negative performance from an 8core/16 thread CPU if the per core speed is lower than Intel's 4 core/8 thread, including gamers.kmi187 - Tuesday, August 23, 2016 - link
Because it is an 8 core part with 16 threads and if you want 8 cores and 16 threads on intel you have to go x99. Rumour has it that 6 cores will come to the z platform. Where up to now a 4 core with hyperthreading was the maximum.fallaha56 - Tuesday, August 23, 2016 - link
which workload is that?there's hardly a game that would be bottlenecked on a 8-core Broadwell or even a 4-core one at ~4Ghz
TekDemon - Saturday, August 27, 2016 - link
Because realistically this is the only scenario where Zen will be competitive, on the CPUs with tons of cores they become much more competitive because the intel parts are also low clocked and more concerned about computing power efficiency than raw power per core.arayoflight - Tuesday, August 23, 2016 - link
Broadwell-E released in 2016.jensend - Tuesday, August 23, 2016 - link
Skylake doesn't have the core count to compare to Zen; they're comparing to Broadwell-E. Broadwell-E was released this May, and its replacement, Skylake-X, isn't due until 3Q 2017.Audiophizile - Tuesday, August 23, 2016 - link
As everyone here knows and is now anxiously awaiting... The clock speed. If it can come out at ~4Ghz and this leak rough IPC is accurate, it needs upwards of 3.8 base clock. If they can do that they have a winner.BlueBlazer - Wednesday, August 24, 2016 - link
On the other hand, AMD's 3GHz Zen engineering sample demonstration could be in the same mould as 3GHz "Phenom FX" http://www.dailytech.com/AMD+Demonstrates+30+GHz+Q... which never saw the light of day. That is considering 14nm LPP which is primarily low power process node optimized for mobile SoCs with low clockspeeds.Black Obsidian - Tuesday, August 23, 2016 - link
They're comparing Zen to Broadwell-*E*, which was released less than 3 months ago. This comparison was presumably chosen because Zen is an 8C/16T part, and Broadwell-E is Intel's equivalent in that space.Phasenoise - Tuesday, August 23, 2016 - link
It's being compared to the most comparable Intel part. Intel's current "big core farm" offering is the Broadwell-E, and the 6900K discussed was released this May.Obviously they can't compare to Kaby Lake as that isn't out.
hansmuff - Tuesday, August 23, 2016 - link
Because if you want to use Blender as a test between different vendors, and the newcomer has 8C/16T, they want to show that their 8C/16T is just as good as a very expensive CPU from the competition.We don't have a Skylake-X yet, so Broadwell it is.
It's also very likely that Skylake is faster than Zen on single-thread IPC, why show that off when you can be as fast as something fairly recent that is still at the high end?
loveroftech - Tuesday, August 23, 2016 - link
because skylake is barely even better than haswell. there is almost no improvement at all. if a cpu is as fast as haswell or broadwell it is basically as fast as skylake.Dribble - Tuesday, August 23, 2016 - link
It's marketing, we all know it's the absolute best case scenario and performance will be nothing like that in the real world. Disagree, look at AMD's last big bit of performance number marketing of a future product:http://www.anandtech.com/show/9886/amd-reveals-pol...
Cooe - Sunday, February 28, 2021 - link
Hahahahahahaha xD. Sounds like Intel's Comet Lake marketing tbh. It's still better than AMD at 720p gaming (which is completely irrelevant) & that's why you should buy it instead of Ryzen!uefi - Tuesday, August 23, 2016 - link
Good on Anandtech giving AMD the scrutiny equally that all major technology players receive here. Instead of sugar coating AMD's bold claims which other tech site articles does.powerarmour - Tuesday, August 23, 2016 - link
It's not even scrutiny, as there are no facts to base it on. There's no need to be critical or sceptical just yet.brianmerel - Tuesday, August 23, 2016 - link
I feel that the point of the benchmark was to show that they have produced a competitive processor. Given if there were optimizations, some cherry picked conditions, they are at least showing that they have overcome significant architectural obstacles and thanks to the choice of foundries they can at least approach the performance of Intel.It is impressive, at the least, that they are staying in the game being a much smaller organization. Thanks Lisa.
r3loaded - Tuesday, August 23, 2016 - link
If AMD have learnt from their marketing mistakes and are being conservative and under promising on performance, having Zen beat Broadwell-E by 2% clock-for-clock is pretty damn good.If.
patel21 - Tuesday, August 23, 2016 - link
The cheapest Broadwell-E costs around $400. If we get even 90% of its performance at around $300 from AMD with similar power requirement, Then Hats off to AMD. I am soldjjj - Tuesday, August 23, 2016 - link
On the memory side it seems that Blender doesn't care much about bandwidth and to be fair lots of consumer apps don't scale with BW.On the TDP side,a roadmap leak and Zauba are showing 95W but we don't really know what clocks they can reach.Still, it would be safe to assume 95W or less for 3GHz and then you could compare with Broadwell-E -not that it matters since perf is unclear for now.
On point 3 and 4 maybe you go too far on the negative side and somehow you argue for SPEC? That's too much about the compiler and Intel puts too much effort into it. You used it to compare Apple's core to Intel and that was unfair.
Also don't make the mistake of assuming that clocks, perf and power can only only surprise on the negative. Intel had no reason to try for more than 5 years, it was all about keeping revenue up in a declining market.
We'll also have to see how it scales,i'm not convinced yet that this is aimed at very high perf instead of moderate perf and higher efficiency.
BrokenCrayons - Tuesday, August 23, 2016 - link
There remains no point in speculation about how Zen will perform. When retail products are subjected to benchmarks and multiple hardware review sites have posted numbers that can be compared among one another, we'll know.Elsote - Tuesday, August 23, 2016 - link
“Reality is the only truth” (Aristóteles)TEAMSWITCHER - Tuesday, August 23, 2016 - link
Even if Zen was significantly faster than Broadwell-E I still would NOT even consider the product. New architectures are often full of bugs. Zen will have to demonstrate reliable and trouble free operation for weeks (if not months) before I would even entertain the notion of using AMD in my next system build.SydneyBlue120d - Tuesday, August 23, 2016 - link
Thanks for sharing your opinion, we will deal with it.While waiting for Zen to became reliable I have a small tip for You:
http://arstechnica.com/gadgets/2016/01/intel-skyla...
andrewaggb - Tuesday, August 23, 2016 - link
I think it's a legit concern. The original phenom had a serious bug and the bios fix resulted in much lower performance. Obviously intel processors can have bugs to, but it's reasonable to assume that a new architecture is more likely to have serious bugs than an existing one.DigitalFreak - Tuesday, August 23, 2016 - link
"Bug discovered while using Prime95 to find Mersenne primes" Hardly a major bug.webdoctors - Tuesday, August 23, 2016 - link
It'll be amazing to finally get some competition. I'm still really happy with my Sandybridge though.It'll be tough for AMD though, Intel has a huge margin/price advantage with their fabs and volume. Also, not clear how many smart competent ppl actually still work @ AMD, but even Intel has laidoff a lot of ppl. Very little money left in the consumer market on the low end and Intel is pretty competitive there.
We're trying to extrapolate on a product that we won't see on shelves for almost a year? Lot can change by then, will see. Obviously there's no trust right now and it'll have to be earned after the previous failures.
Gothmoth - Tuesday, August 23, 2016 - link
given the fact that AMD has practically lied about the performance of it´s CPU´s in the past, i hope this time it´s not just hype.if AMD can not deliver this time they are just a sad joke.
we need competition, badly.
let´s hope AMD can at least catch up to intel.
tuxRoller - Tuesday, August 23, 2016 - link
Since you folks don't really follow oss news I'm gonna let you in on a little secret:AMD has OPENLY spent time getting blender to work well with their hsa hardware.
No secret.
http://developer.amd.com/community/blog/2015/07/10...
https://wiki.blender.org/index.php/OpenCL
I LOVE how AT insinuates that open source benchmarks are somewhat less trustworthy.
That's just....well, odd. Typically, people point explain their setups, point to the code they've run, and then TRULY anyone can duplicate it AND, far more importantly, examine the friggin code to see if there is an gaming happening.
Let's just relax, and see if amd does the right thing and points to the code (and toolchain, ofc) so it can be examined.
Zan Lynx - Tuesday, August 23, 2016 - link
I am pretty sure they were talking about compiler optimizations regarding open source software.Closed source is compiled for all CPUs. Probably optimized for Intel as we saw with No Man's Sky. But open source it is easy to rebuild it with custom options to make it run really well on a particular CPU.
tuxRoller - Tuesday, August 23, 2016 - link
Compiler optimizations are covered by reference to code (to see if they actually made source code changes that benefit them....and, as I said, they have done this, but it is well known in the oss community and if amd was trying to hide that info, they did a terrible job) and toolchain (which includes any libraries needed, the linker used, and the compiler, along with the versions of each---again, this is all really basic stuff).It goes without saying that compiler flags are listed as part of this.
Achaios - Tuesday, August 23, 2016 - link
Ι believe these are the best news for the end user we have had since February 1st, 2004, when Intel released Prescott.The last time an AMD CPU had the advantage over an Intel CPU was...12 years ago.
I sincerely hope the reported Zen advantage is not an advertising gimmick. Intel really need a slap in the face, let's hope Zen will give it to them.
KenLuskin - Tuesday, August 23, 2016 - link
I think the main point one should take away from AMD's initial look at ZEN is that it will be roughly competitive. And that is all AMD needs to get at least 10% market share ( measured in Revs, and not units) in high end PCs and Servers.In order to understand what a 10% market share means to AMD please read the following article:
AMD: Zen =10% Market Share in Servers/PCs + VR/AR/MR= $75/share
BlueBlazer - Wednesday, August 24, 2016 - link
Nopesiree, not happening yet because we have not seen any pre-announcements from major sever and supercomputer hardware manufacturers on the adoption of AMD's Zen for future upcoming servers and supercomputers (which meant something is amiss with AMD's Zen). Just look at Intel's Knights Landing as an example, many months before launch already many major server and supercomputer hardware manufacturers (e.g. Cray) pre-announced upcoming products that will be using the new chip. And AMD's Zen being so much "closer" to (paper) launch with just few months away, we have yet to see any of those pre-announcements (especially from companies like Cray, SuperMicro, Tyan, Lenovo, Huawei, HP, Dell, etc)...lefty2 - Tuesday, August 23, 2016 - link
It's a pity they chose a multithreaded benchmark. Single threaded is more important. Zen only has to get close to Intel's single threaded performance to be successful.LeptonX - Tuesday, August 23, 2016 - link
"It's a pity they chose a multithreaded benchmark. Single threaded is more important. Zen only has to get close to Intel's single threaded performance to be successful."Why? Both products are 8C/16T CPUs so if MT performance is similar their ST performance should also be similar. Nothing suggest that ZEN will gain more from running the second thread than Broadwell. Neither CPU is designed like the Power8 that aims to achieve the best dual-threaded and quad-threaded performance. Power8 gains something like 50% from the second thread but nothing suggests that ZEN can do the same. I think gains from the SMT will be very close between ZEN and Broadwell.
lefty2 - Tuesday, August 23, 2016 - link
Actually, there are potentially lots of things. Zen has more cache per core for instance.benzosaurus - Tuesday, August 23, 2016 - link
Pro tip: if your article can be summed up as "we have no idea what the answer to the question posed by the headline is", then it's clickbait, not an article.Alexey291 - Tuesday, August 23, 2016 - link
Spot on sireuskalzabe - Tuesday, August 23, 2016 - link
Sadly correct.Alexey291 - Tuesday, August 23, 2016 - link
Another clickbait article.All of it is literally stating the obvious.
Ian Cutress - Tuesday, August 23, 2016 - link
A lot of articles out there aren't stating the obvious, or getting the facts mixed up, or calling things by the wrong name and extrapolating that into what the results mean. Sometimes the obvious needs to be said.Ranger1065 - Saturday, August 27, 2016 - link
Lol.YukaKun - Tuesday, August 23, 2016 - link
I can still see the flames coming out of the keyboard that wrote this heavily opinionated piece of article.I don't think I've ever seen one of these for any other vendor out there from Anands, but please prove me wrong.
Cheers!
Tchamber - Tuesday, August 23, 2016 - link
While you are arguably correct, many people, probably you included, would be yelling at the lack of reports on Zen performance that is available on other tech sites. It's always best to take single results with a grain of sand, that's why AT runs a benchmark suite, not just one. I appreciated the article, thanks Ian.YukaKun - Wednesday, August 24, 2016 - link
I probably would, I won't argue that.I have to admit I am a bit "salty" around the tone implied in the article itself. There's no depth to it other than just question blindly a single benchmark not targeted to consumers. Blender is a FOSS project, visible to anyone, but here the piece revolves around the ways you can "cheat" around it, other than exploring how those numbers come to be.
I don't appreciate opinionated pieces when they don't put facts before arguments. It does gives perspective, but only from a biased POV; or that is the conclusion I get. Even if I don't disagree with the statements, they don't seem objective (weirdly enough).
Cheers!
YellowOnion - Tuesday, August 23, 2016 - link
"there could be CPU vendor specific optimizations in either the code or compiler"how is this related to Open Source? /any/ software program can be setup like this, you could even make the opposite argument, if the code isn't tuned for the specific CPU then you're not taking advantage of the CPU properly.
Considering no one is writing code for Zen yet, it's only logical to use a codebase that can be modified and optimised if performance is biased against it.
As long as AMD did a fair comparison (optimised for Zen on Zen, and optimised for Broadwell-E on Broadwell-E), I don't see how this has /anything/ to do with OSS.
I suspect this benchmark was chosen because it threads well, looks good (name another benchmark that you can actually /show/ in real time) maybe Zen is pulling more out of the SMT implementation, while being slightly slower in the IPC department.
I suspect AMD wants to push multicore, gaming loves the "4 core" CPU, and the industry is too lazy to write code for CPUs that don't really exist yet, though with the PS4/Xbone pushing 8 cores, Vulkan and DX12 here, I suspect the next gen of gaming is scaling past those 4 cores.
wiak - Tuesday, August 23, 2016 - link
remember kids dont do like intel and playback a video of a game while doing a presentationhttps://www.youtube.com/watch?v=Otcge1cn8Os
atleast amd showed it playing a REAL game on a REAL graphics card... that they built them self in house
wiak - Tuesday, August 23, 2016 - link
https://youtu.be/oQS8s7TOXsE?t=64Crp - Thursday, August 25, 2016 - link
The difference is that at AMD they dont say it is actual gameplay or video playback. At the Intel presentstion he's pretending to play the game, fails to play the video. And tries to cover up his bad mimicing that by saying its been played backstage. While you can clearly see it is a video which is also laggy.nexellinAlpha - Tuesday, August 23, 2016 - link
Honestly as an AMD fanboi I have to say that I dont see how this article was antagonistic or really biased. It neatly points out problems with the benchmark and essentially says it's not an accurate metric and to not get your hopes up just yet. Wait for more valid benchmarks.just4U - Wednesday, August 24, 2016 - link
Not sure where the bias is either.. It hints at possibilities.. but nothing verifiable. My takehome on the article is it's possible Amd has a very competitive chip coming out..DifferentFrom - Wednesday, August 24, 2016 - link
AMD's GPU division have been working fairly heavily on Blender Cycles support for their devices. They would know the code quite intimately.SageFox - Wednesday, August 24, 2016 - link
I don't like Anandtech. I could say that cinebench also helps intel and it's used in this site to measure wich CPU is faster and also this article praises the cinebench as a reliable source of conclusions.BUT, besides cinebench and anandtech, this article is OUTSTANDING. I agree in everything as an AMD fan. I'm having a bad feeling about ZEN and i fear it's the same fiasco bulldozer was.
If it is the case, AMD is done, AMD is ashes.
Mugur - Wednesday, August 24, 2016 - link
I don't think it will be the same fiasco... more like Radeon RX 480.Cooe - Sunday, February 28, 2021 - link
Hahahahaha xD. How's that working out for you today? Current AMD stock price ≈ $85.Fukkit - Wednesday, August 24, 2016 - link
I "upgraded" from an FX8350 to an i5 6600. It boots up faster, for sure, by about 10 seconds. And er, I get like 5 more FPS in a spatter of games.Do I feel mugged? Yes. Yes I do.
Will I be buying Zen regardless of how competitive it is, yes I will.. I tried the Intel side and found it left a bitter taste in my mouth/wallet. I gave my GF the fx8350, I've offered to trade systems but she said now.
willis936 - Wednesday, August 24, 2016 - link
If you're buying a new build when your current one is less than two years old then the only one mugging you is yourself.Fukkit - Wednesday, August 24, 2016 - link
I have zero intention to sit here with a PC that constantly disappoints me. This one will be sold off at a cheap price to anyone looking to get into PC gaming.I'll use that to fund an AMD Zen and Vega build and go nuts, build my dream pc.
Klimax - Wednesday, August 24, 2016 - link
That's some severe invasion by AMD fanboys. The only question is, how many of them are being rewarded by AMD for posting such idiotic attacks and how many do it for free. (Not fist time either...)Fukkit - Wednesday, August 24, 2016 - link
Because it's so bad to support the underdog...I've currently got an Intel Nvidia build and I can't wait to get back to team red.
Michael Bay - Wednesday, August 24, 2016 - link
AMD doesn`t even have money to properly pay off Demerjan, what do you expect.Ranger1065 - Saturday, August 27, 2016 - link
Shill.zodiacfml - Wednesday, August 24, 2016 - link
whatever! regardless of any performance, it is the price that matters!azazel1024 - Wednesday, August 24, 2016 - link
To play devils advocate, I disagree with the mention that the lower bound of the high end part is 3GHz for Zen. There is nothing to suggest that. To milk the benchmark AMD could have run it with a stouter cooler and upped the voltage to juice it to 3GHz. Or 3GHz could be the turbo speed, it was a pretty short benchmark. I'd assume that the chip IS capable of 3GHz, for awhile at least, but that might also be with OC'ing. I think that is the most that it suggests.I'd think AMD wouldn't want to milk things TOO much, so I'd expect the 3GHz is probably at most a turbo speed which it can handle with a stock cooler for the duration of the test without excess thermals. It also could be within the base clock performance. So that means that the base clock is probably somewhere between 2.6-3.1GHz for final chips and turbo performance is probably 3-3.4GHz as a random guess.
As for actually IPC, I don't think it says a thing other than custom benchmarks can be made to make Zen look good. That said, I have some hope here. I don't need it to be equal to or better than Broadwell-E. If it can achieve 80+% of the single thread performance of Broadwell-E and at the same time can beat the performance or at least equal the performance of a HEXAcore Broadwell-E for a Zen 8 core, 16 thread part at somewhat less platform price, AMD has a sale out of me.
Nagorak - Wednesday, August 24, 2016 - link
I have to agree that I don't know what the point of this article was. It was always clear it was an isolated quick test and didn't necessarily mean anything. That great warranted maybe a paragraph, the rest is just wasted words.diglo1 - Wednesday, August 24, 2016 - link
People here should realize that AOTS cant take advantage of More then 8 threads and Zen has double of that. Plus engineering sample Zen has lower clockspeed per Core then full fedget 4790 so no Wonder it loses in single Core performance and thats still what games want more. Now something like 5960x or 6900k would have shown similar performance like Zen engineering example at same clockspeeds. You must realize that currently 4c 8t i7s perform better in gaming vs any 8c 16t simply because having More cores means lower clockspeeds and clockspeed is the käy to single Core performance if IPC is the same. I5 Lost simply because it didnt have enough threads but decent clockspeed. Also Zen had 3,2ghz turbo and I7 had 4ghz or was it even More. So no surprises here Zen is delivers close to skylake depending on the spesific tasks.1. Zen is not at it's final clockspeeds
2. Zen is More comparable to 5960x or 6900k both in terms of multi threaded tasks and gaming
3. Reqular i7s beats higher end i7s in gaming simply because higher turbo clockspeeds
4. Zens ipc is not 100% confirmed.
5. Zen is going To make a difference By offering similar performance compared to Intel By lower prices.
6. No matters the similarities both architectures have pros and cons.
darkfalz - Wednesday, August 24, 2016 - link
I would like AMD to do well again, but I can't really trust their marketing. Two things make me suspicious:1. AMD has been in the red for many quarters in the last few years. I can't really seeing them having the R&D time and money to create a competitive product other than Intel getting lazy. If they did it, hat's off, but I'm sceptical.
2. AMD's benchmarks and marketing have always been selective and dodgy. I remember when the FX series launched and AMD released a bunch of ridiculous gaming benchmarks at high resolution/high AA claiming similar performance to Intel CPUs (because they benchmarks were highly GPU bound, obviously, taking the GPU out of the equation).
ruthan - Wednesday, August 24, 2016 - link
I no more believe AMDs numbers at all, they promised than Polaris would much more power efficient than Maxwell, it was 100% lie, sadly all so-called analysts dont care about promises vs. reality checks.Here are these old lies slides, youtube videos are gone:
http://www.pcper.com/reviews/Graphics-Cards/AMD-Ra...
Morg72 - Saturday, August 27, 2016 - link
You might want to read that article you linked again. The only time maxwell is mentioned is when they compared what must have been the 460 to the gtx 950. Other than that, it only discusses that Polaris is a big leap in performance per watt over previous gen Radeon GPUs.Alouette Radeon - Thursday, September 15, 2016 - link
Hey it could be worse, it's not like they used wood screws on a fake card with the ceo waving it at investors and claiming it to be real. Wood was fitting however because when Fermi was finally released it ran so hot that you could Bar-B-Q a T-Bone on it. How come you trust nVidia after that fiasco? Or what about that gong show where nVidia renamed the GeForce 9800GTX+ as the GTS 250 and increased the price by 30%? Or wait, remember when they were sodomising their own board partners to the point that BFG died? Is it any wonder that XFX and PowerColor flipped "dear leader" the bird and started making Radeon boards? That was unprecedented as no board partner had ever left ATi. Or maybe the debacle that ensued when Charlie Demerjian exposed the way that nVidia was threatening to deny samples to review sites that didn't use only "nVidia-approved" games in their reviews (read: games that performed better on nVidia hardware) because they tried to do the same thing to theinquirer.net and he wouldn't have it. Either you're too young to have seen all that (around 2009) or you have the memory of a fruit fly. My point is, sure, AMD has done some stupid things but when it comes to all-out dick moves, nVidia has ATi outscored by a score of ten to one at least. Hell, they redid their driver package to disable PhysX and CUDA if ANY ATi graphics products were detected in the system. Ask yourself now why you're so quick to condemn AMDbut not anyone else? "LET HE WHO IS WITHOUT SIN CAST THE FIRST STONE!" (I read that somewhere... Heheheh)Haawser - Thursday, August 25, 2016 - link
Can't wait to buy a 4C/8T Zen to go with my RX 480. Should be plenty fast enough to drive a single card in DX12/Vulkan/VR. Which is my primary reason for still having a PC.As much as I like the sound of having 8C/16T, I couldn't really justify it for what I want to do. So as long as the 4C/8T is cheaper than a 4C/8T i7 (and ~as fast) I'll go with the 'mini-me' Zen, and save some money. If I change my mind, or find some use for it, I can always sell the 4C/8T and buy the 8C/16T anyway.
All I really want from *any* CPU is that it doesn't bottleneck my graphics card. So I think a 4C/8T Zen should be fine, as long as they don't want silly money for it. Like $300...
Gastec - Thursday, September 8, 2016 - link
Could you be more of a troll? "primary reason for still having a PC" Like having a PC is such a frikin big deal. It's hilarious really. What else are you STILL clinging to for no real reason? Life?lolipopman - Sunday, September 25, 2016 - link
Why are your feelings hurt?Not everyone has the time or money to shell out 1000s of dollars for a rehashed CPU just to gaze at a few renders.
chunkRank123 - Thursday, August 25, 2016 - link
Going down the optimization for specific CPU microarchitectures leads on to another more philosophical issue?????Are you serious?! LOL - that's the whole point of an umm, compiler (and compiler optimizations, differing instruction sets between cpus, etc.)
This would be like saying Intel was "cheating" because they used AVX in their processor before AMD did, and a developer used the extensions in their code.
Simply ludicrous to think like this; software can (and sometimes should) be optimized for a specific processor even though the general architecture is the same... That's not "cheating", it's called "using something as designed"
Why do you put a question mark in this article's title when you give NO INFORMATION suggesting otherwise (e.g. AMD's processor is certainly not faster)?
This is almost buzzfeed quality "journalism" - nice job.
stardude82 - Friday, August 26, 2016 - link
An FX-8350 (4 Ghz) is competitive with an i7-2600K (3.2 Ghz) in blender according to Anandtech's benchmarks. Assuming a claimed 40% and 25% measured IPC improvement between generations respectively, this result isn't a surprise or very meaningful outside of Blender users who have cheap electricity.Comdrpopnfresh - Saturday, August 27, 2016 - link
I appreciate the re-situating of a media event. But isn't the fact that AMD performance tests are being contextualized instead of receiving the same head shaking of the last 8-10 years acknowledging some press release as verification that AMD still has a pulse, albeit not a competitive one, reason for excitement enough? I also heard that the Intel testbed mentioned in this article was farther from the HVAC system and situated at a slightly lower elevation than that of the AMD Zen. Take it as you will... Just saying,Haawser - Saturday, August 27, 2016 - link
I'd wait to see Zens prices before saying that it isn't competitive ? The cheapest 4C/8T i7 is ~$300 afaia ? If AMD made the 4C/8T Zen nearer $200 then I believe it would be more than competitive. Even if it was only 95% as fast.
General consensus of opinion is that the 4C/8T should be about the same size as a P11 die, ie ~120-130mm^2. Given that you can buy an entire RX 460 card (including P11) for ~$110, then $200 would seem entirely doable ? As going by P11 pricing I doubt a die that size costs more than ~$50 to make, if that ? Assuming that to be the case they could sell it for $200 giving the retailer a 100% markup, and still make 100% gross profit margin.
Would a CPU 95% as fast as a 4C/8T i7 for gaming and VR, but costing ~33% less, be competitive ? I think it would, because most people would probably prefer to spend the extra $100 they saved on a better graphics card ? I know I would, because it would be pretty much guaranteed to be more than 5% faster than one costing $100 less.
Targon - Tuesday, August 30, 2016 - link
There are a lot of things going on with Zen, mostly for the better. First up, most people familiar with AMD love the fact that AMD doesn't replace sockets every time a developer passes gas the way Intel does. This means that a socket AM4 motherboard will support AMD processors for at least the next two years for upgrades, and the socket AM4 chip they might buy today will work in their next motherboard if they go AM4+ in the future. No need to worry about different sockets for APU vs. pure CPU, and you could start with an APU and then upgrade easily if you wanted to.DDR4 support is a plus.
Now, overall performance. Engineering samples tend to be on the slow side in terms of clock speed. The entire purpose is to let hardware partners develop boards, so clock speeds are a non-issue. They could be 500MHz for all AMD cares when it comes to these engineering samples, because the PURPOSE would be met.
Demonstrations are designed for different purposes. Early demos are more to show "hey, things work, it isn't just vapor products we are talking about". Close to launch, we will hear and demonstrations showing the chip running at 6GHz on liquid nitrogen will come out to hype the product LAUNCH.
Remember, we can't expect to see product available for sale until at least November/December at the very earliest. If AMD were to release to distributors in October(doubtful at this point), it would take a good month for those to make it into the hands of consumers.
Chipsets are a big part of performance testing. Working around problems by motherboard manufacturers tweaking settings should be expected right now, so who knows for sure how much of what we are hearing about is real-world performance. Expect that at the same clockspeed, we will probably see at least 10 percent better performance as debug code is disabled.
So, we are at the end of August. One more month until we find out how yields are, which will determine official launch dates. No matter what, AMD will finally have a new generation core, and that means better performance. For those who are being critical, how many love those dual-core i3 and i5s that are all over the industry that SUCK compared to a quad-core A8 or A10 in the same price point for a system? At the high end, where quad core and oct-core processors are the norm, it is easier to say that AMD has not been competitive, but in the low to mid range, I'd rather go for a quad-core AMD over a dual-core Intel any day.
Alouette Radeon - Thursday, September 15, 2016 - link
I don't put much stock into that Blend benchmark win but even if it was optimised for Zen it doesn't matter, an engineering sample did beat a market-grade Broadwell-E in a benchmark. (Regardless of Anandtech's feeble attempt at Intel damage control proven by this worthless excuse for an article made up almost completely of semantics and political-style rhetoric.) At the end of the day, it's really this simple: IF the chip was the POS that all the Intel shills are claiming then NO amount of optimisation would help a mere ENGINEERING SAMPLE narrowly defeat a MARKET-GRADE Broadwell-E with the SAME NUMBER OF CORES at the SAME CLOCK FREQUENCIES in ANY benchmark. It would have to have been completely faked. (The use of caps lock was to emphasise only the objective facts that not even the author of this "article" dares to dispute.)kyaaaaaaaaaa - Friday, December 2, 2016 - link
I do not even hope it would match haswell in IPC. I do expect it outpace Kabylake in terms of c/p ratio though, at least on some model.