Comments Locked

37 Comments

Back to Article

  • MrCommunistGen - Monday, May 2, 2016 - link

    Good. I'd rather both companies spend their resources on innovation instead of lawyers.
  • jasonelmore - Monday, May 2, 2016 - link

    what possible GPU patent could samsung have that warranted the ITC to side with them? Samsung's doesn't even make a gpu do they?

    Nvidia invented the GPU with the Geforce 256, so it's kind of absurd, and makes me question the ITC's motives.
  • MrCommunistGen - Monday, May 2, 2016 - link

    The GeForce 256 is far from the first GPU, from NVIDIA or otherwise so I'd question that they "invented" GPUs.

    I only glanced at some of the linked articles but it looks like some small functional unit(s) within the GPU are what are under contention, not the entire concept of a GPU.
  • Morawka - Monday, May 2, 2016 - link

    Prior GPU's depended on the CPU to do transforms and lighting. Thats why the 256 was so ground breaking at it's released. It was where the term "GPU" was widely used and Nvidia marketed it that way.

    https://en.wikipedia.org/wiki/GeForce_256
  • silverblue - Tuesday, May 3, 2016 - link

    They popularised the term "GPU" but they didn't invent the technology; they were merely the first people to bring a hardware T&L product to the consumer PC space.
  • Arnulf - Tuesday, May 3, 2016 - link

    What about 3DFX Voodoo line?
  • StevoLincolnite - Tuesday, May 3, 2016 - link

    Voodoo never had a dedicated TnL unit.

    However... Today's GPU's also don't have a TnL unit either, that fixed function hardware was abolished and is now done in the pipelines... Thus technically... GPU's today are technically not GPU's anymore. :P
  • Samus - Tuesday, May 3, 2016 - link

    I consider a GPU a programmable FPU, unlike purpose built ASIC's or optimization\arithmetic specific CPU's (ARM, x86)

    All modern GPU's are GPU's. They literally do things a CPU can't. Sure a lot gets offloaded to a CPU now, and certain GPU features have been abandoned such as T&L, physics, etc, but a CPU doesn't take DirectX commands, compile CUDA, and isn't optimized for in-order math.

    But the point is, nVidia opened a can of worms here. They believed their position was stronger than it actually was because they are the father of the modern GPU, and those who argue 3Dfx is, are right, but don't forget who bought 3Dfx's patent portfolio.

    Like the OP said, nVidia needs to stick to engineering, not litigation.
  • tyleeds - Wednesday, May 4, 2016 - link

    It might actually be fairer to say that Imagination technologies was the father of the modern GPU as they pre-date nVidia, battled 3dfx in the original Voodoo and Voodoo2 days and are the driving architecture behind the vast majority of mobile 3d chipsets today.

    Many of the features of modern non-PowerVR GPU's are directly related to tactics originally developed for the tile-based rendering on powerVR chipsets.
  • ET - Tuesday, May 3, 2016 - link

    Your question can be easily answered by going to the linked news report about the Samsung win. To quote (because I'm sure you're too lazy to follow that even with prompting): "The patents cover an implementation of SRAM, a shared computer bus system with arbiter and a memory sub-system with a data strobe buffer."

    In short, not GPU specific.
  • BurntMyBacon - Tuesday, May 3, 2016 - link

    @ET: "..."The patents cover an implementation of SRAM, a shared computer bus system with arbiter and a memory sub-system with a data strobe buffer." ..."

    Samsung owns the patents for 1) an implementation of SRAM, 2) a shared bus with an arbiter, and 3) memory sub-systems with a data strobe buffer? Or is it that they own a patent on using them all in conjunction?

    1) I suppose it makes sense that they could patent an (singular) implementation of SRAM. It is, however, hard to believe that nVidia could accidentally run afoul of this as they could just select an other implementation (probably more suitable to GPUs).

    2) Shared buses and arbiters go hand in hand and are also a common with plenty of prior precedent. They must again be talking about a single implementation that would be hard to accidentally run afoul of.

    3) A memory subsystem with a data strobe buffer seems specific enough (for the US patent system) that they might have something here. If you were aware of it, I can't imagine that you wouldn't just design it a little differently, though you might select an infringing design without knowing.

    Seems more likely that this is a case of using these (mostly) common elements in a specific manor that infringes the overall architecture patent.
  • DarkKnyte - Saturday, May 7, 2016 - link

    Nvidia invented the GPU? Lol. You've got to be like what, 15? I bet you think Nvidia invented SLI too.

    The first dedicated graphics processing units (i.e. GPU) were single chips in the 70's. These chips were the actual hardware responsible for drawing graphics, sprites, and colors on a screen.

    https://en.wikipedia.org/wiki/MOS_Technology_VIC

    Nvidia, AFTER Voodoo helped popularize the *3D* graphics consumer segment. Voodoo created the 3D PC graphics market that Nvidia latched on to. Nvidia is responsible for a lot of great things in the GPU space, but they haven't created any markets. They've always innovated into markets already created by others.
  • Notmyusualid - Monday, May 2, 2016 - link

    I have a feeling that Samsung has a better team of lawyers...
  • invasmani - Tuesday, May 3, 2016 - link

    I'd imagine so given that I'm pretty certain it's the top business in South Korea after all they make just about everything too it's a rather diverse company more so than people might first expect.
  • BurntMyBacon - Tuesday, May 3, 2016 - link

    @Notmyusualid: "I have a feeling that Samsung has a better team of lawyers..."

    Well, they have been practicing with Apple, so ...
  • Samus - Tuesday, May 3, 2016 - link

    Well, obviously they would, they are a much larger company than nVidia, and Samsung lawyers have had a lot more experience with these sorts of cases over the last decade.
  • aryonoco - Monday, May 2, 2016 - link

    Good. This was s stupid move on Nvidia's part anyway.

    Nvidia is a stable, profitable engineering company with growth potential and admirable products. They are not a Rambus or a Tivo, and don't need to resort to those tactics.
  • Yojimbo - Tuesday, May 3, 2016 - link

    It wasn't a stupid move. It was an attempt to "open ports" like Commodore Perry bringing a warship to Japan, I think. Qualcomm has crucial patents regarding CDMA and they use them to leverage sales of their modems and Snapdragon chips. It forced TI out, NVIDIA out, and Icera already had filed a suit against them before NVIDIA bought them. I think NVIDIA tried to use their GPU patents to gain leverage with Samsung to compete on a more equal footing with Qualcomm: get them to license their IP or buy their SOCs, to fight fire with fire, so to speak. It didn't work. Not sure why, I would have thought that AMDs and NVIDIAs patents were crucial to the modern GPU features that mobile GPUs have started to use now, just as Quallcomms patents are crucial to CDMA-capable modems. Maybe NVIDIA doesn't have a strong patent position, but who knows with this patent system. It seems to be broken. As I recall, Apple won a case that it had patents on a parallelpiped shaped phone and icon movements which follow real-world physics. Ridiculous.

    Anyway, now that China has punished Qualcomm, the US and EU now seem to be taking the antitrust allegations against Qualcomm seriously, whereas they weren't previously. NVIDIA wants $350M in damages from Qualcomm saying that Qualcomm's illegal actions caused the failure of their Icera unit. If Qualcomm is forced to change their tactics NVIDIA may well get at least part of what they wanted by their patent suit, anyway.
  • Yojimbo - Tuesday, May 3, 2016 - link

    I meant to say "It forced TI out, NVIDIA out, and Icera already had filed a suit against Qualcomm before NVIDIA bought Icera."
  • BurntMyBacon - Tuesday, May 3, 2016 - link

    @Yojimbo: "... , I would have thought that AMDs and NVIDIAs patents were crucial to the modern GPU features that mobile GPUs have started to use now, ..."

    Qualcomm's current GPU division originated at ATi, so it's possible that they have the patents they need. Purchase of ATi handheld graphics division and Imageon graphics technology for $65M completed early 2009.
  • Yojimbo - Tuesday, May 3, 2016 - link

    They actually purchased it from AMD. AMD purchased ATI in 2006. But the Imageon line did not have modern GPU architectures, I think. I'm not really sure, though.
  • tuxRoller - Tuesday, May 3, 2016 - link

    Actually, qualcomm's gpu division ORIGINATED with bitboys before it was purchased by ati.

    http://www.tomshardware.com/reviews/medfield-krait...
  • Kvaern2 - Tuesday, May 3, 2016 - link

    I believe the major mobile GPU designs all go a way back.
    Imagination and Falanx (Mali) are failed desktop graphics businesses which transitioned into mobile.
  • Communism - Monday, May 2, 2016 - link

    I called it from the beginning.

    There was no way Nvidia was possibly going to be able to outbribe, out-backroom deal a company that essentially owns a nation-state. South Korea is a wholely owned subsidiary of Samsung.
  • beginner99 - Tuesday, May 3, 2016 - link

    David vs. Goliath doesn't always end that well. NV made the classic mistake to never wake a sleeping dragon. I hope they have to dearly pay for it. I hate patent trolling.
  • Yojimbo - Tuesday, May 3, 2016 - link

    I doubt Samsung holds much ill-will towards NVIDIA. They were able to defend their position and that's that. Now I'm imagining both Samsung and NVIDIA want the same thing. Restrictions on Qualcomm's licensing practices. If NVIDIA had won Samsung would have been stuck between a rock and a hard place, needing NVIDIA's patents on one hand and Qualcomm's on the other. I can imagine both Apple and Samsung, as well as Qualcomm and ARM didn't want NVIDIA to succeed in their patent claims. That's a massive headwind if one has some doubts about the patent process. Probably only AMD wanted NVIDIA to succeed besides NVIDIA. But I doubt all those companies are out for any retribution against NVIDIA's attempt, including Samsung. NVIDIA is buying Samsung HBM2, in fact, and Samsung is trying to court NVIDIA's foundry demand.
  • invasmani - Tuesday, May 3, 2016 - link

    I highly doubt AMD wanted Nvidia to succeed and I'm so certain ARM cared one way or another they will make their licensing money either way without or without one or the being the bigger fish in the swimming pool.
  • BurntMyBacon - Tuesday, May 3, 2016 - link

    @invasmani: "I highly doubt AMD wanted Nvidia to succeed and I'm so certain ARM cared one way or another they will make their licensing money either way without or without one or the being the bigger fish in the swimming pool."

    It would effect their ability to license Mali and the value of that license.
  • Yojimbo - Tuesday, May 3, 2016 - link

    If NVIDIA could create a precedent that their patents are crucial, then AMD could more easily get licensing fees, too.
    From http://wccftech.com/amd-engage-nvidia-samsung-pate...
    "During the Raymond James’ Systems, Semiconductors, Software & Supply Chain Conference on December 8th. AMD’s Chief Financial Officer Devinder Kumar revealed that AMD has in fact considered legal action.

    Hans Mosesmann – Raymond James
    I think investors are interested in seeing how this plays out with Nvidia as part of their lawsuit against the Samsung and Qualcomm regarding fundamental Graphics IP, now AMD obviously has as well a very deep portfolio of Graphics IP. Has management considered a strategy to monetize that like Nvidia is doing, and what would you need to do this and start that effort?

    Devinder Kumar – SVP and CFO
    Yes. We have considered that and we will explore possibilities in that area as we move forward".

    ARM cares not for their CPUs or their ARM architecture, but because they also design GPUs, and they, too, might have had to license IP if NVIDIA had won.
  • testbug00 - Tuesday, May 3, 2016 - link

    AMD is monetizing it's IP. Did you miss their deal JV with a chinese company where AMD provides IP and is getting payment? AMD prefers to get things done and maintain good relationships with companies.
  • Mondozai - Wednesday, May 4, 2016 - link

    20% of all replies in this thread is from you, all of them furiously defending NV like a fanboy.

    As Testbug00 wrote below, AMD doesn't resort to these patent-troll tactics that NV often does.
  • Michael Bay - Wednesday, May 4, 2016 - link

    Spoken like a rabid AMD fanboy, congratulations.
  • TheJian - Wednesday, February 8, 2017 - link

    https://pipedot.org/article/2BF9W
    Just thought a year later interesting to update your "amd doesn't resort to these patent-troll tactics that NV often does". Umm..They just did, and BTW that was NV's 1ST suit filed by them. Any previous NV attempts have only been in response to someone else (like Intel breaking their agreement over chipsets).
  • JoeMonco - Tuesday, May 3, 2016 - link

    Couldn't bother reading the whole article? The only outcome was a narrow-focused, cross-licensig deal and nothing more.
  • Yojimbo - Tuesday, May 3, 2016 - link

    Yes, that was the outcome. The outcome is not known beforehand. Think a little.
  • webdoctors - Tuesday, May 3, 2016 - link

    It doesn't sound like this will stop future lawsuits. This was only a settlement of <10 patents in question. Both companies have 1000+ patents. The real winners will be known if we see NV licensing happening or Mali or QCOM GPUs growing in complexity without any IP payments.
  • Vlad_Da_Great - Tuesday, May 3, 2016 - link

    SAMS gave'em an easy pass. I am sure NVDA said, lets settle. NVDA are suing QCom again in EU.

Log in

Don't have an account? Sign up now