Original Link: https://www.anandtech.com/show/13080/the-phison-e12-nvme-ssd-controller-preview
The Phison E12 Reference Design Preview: A Next-Gen NVMe SSD Controller
by Billy Tallis on July 18, 2018 10:30 AM ESTPhison is one of the most successful SSD controller designers that sells controllers on the open market. Phison's controllers range from tiny low-power chips for USB flash drives and memory cards up to high-end NVMe SSD controllers, ending up in drives such as the Kingston A1000, and the Patriot Hellfire. The Phison E12 is its latest flagship NVMe controller intended for high-end consumer SSDs and enterprise SSDs. We have an engineering sample of Phison's reference design for a consumer M.2 NVMe SSD with the E12 controller and 1TB of Toshiba 3D TLC NAND flash memory. Phison — and the many brands that base their SSDs on Phison designs — are hoping that this new combination will be able to compete against top SSDs from the vertically-integrated giants like Samsung and Western Digital.
Phison has a close relationship with Toshiba, and most Phison-based consumer SSDs from the past several years have used Toshiba NAND flash. This was a bit of a hindrance for brands using Phison controllers during the year where Samsung and IMFT were shipping 3D NAND but Toshiba was still primarily producing 15nm planar NAND. Phison's first NVMe controller (E7) was used almost exclusively with Toshiba's 15nm MLC NAND and consequently those drives tended to have more limited capacities and higher per-GB prices than their competition.
The situation this year is quite different. Toshiba's 64L 3D TLC is very competitive, and when paired with the right controller it can deliver great performance (as seen in the WD Black with SanDisk NAND) and great power efficiency (as seen in the Toshiba XG5).
Phison NVMe SSD Controller Comparison | ||||||
Controller | E12 | E12C | E8 | E8T | E7 | |
Model Number | PS5012-E12 | PS5012-E12C | PS5008-E8 | PS5008-E8T | PS5007-E7 | |
Host Interface | PCIe 3.0 x4 | PCIe 3.0 x4 | PCIe 3.0 x2 | PCIe 3.0 x2 | PCIe 3.0 x4 | |
Protocol | NVMe 1.3 | NVMe 1.2 | NVMe 1.1b | |||
NAND Channels | 8 | 4 | 4 | 4 | 8 | |
NAND Chip Enable lines | 32 | 16 | 32 | 32 | 64 | |
Typical NAND | 3D TLC, 3D QLC | 64L 3D TLC | 15nm MLC | |||
Max Capacity | 8 TB | 2 TB | 2 TB | 1 TB | 2 TB | |
DRAM Support | DDR4, DDR3L | DDR3(L) | None (HMB) | DDR3(L) | ||
Error Correction | LDPC, StrongECC | StrongECC | BCH | |||
Manufacturing Process | TSMC 28nm | UMC 40nm | TSMC 28nm | |||
Sequential Read | 3200 MB/s | 1700 MB/s | 1600 MB/s | 1600 MB/s | 2600 MB/s | |
Sequential Write | 3000 MB/s | 1700 MB/s | 1300 MB/s | 1300 MB/s | 1300 MB/s | |
4KB Random Read | 600k IOPS | 340k IOPS | 240k IOPS | 120k IOPS | 300k IOPS | |
4KB Random Write | 600k IOPS | 400k IOPS | 220k IOPS | 130k IOPS | 200k IOPS | |
Retail SSD Availability | Soon | ? | Q4 2017 | ? | Q1 2016 |
Late last year, Phison's E8 controller arrived in the retail market with the MyDigitalSSD SBX. This second-generation NVMe controller is aimed at the low-end NVMe SSD market: it features fewer PCIe lanes and NAND channels than a typical high-end SSD and is fabbed on a cheaper 40nm process rather than the 28nm process used by most cutting-edge SSD controllers. The Phison E12 is the high-end member of this second generation, and it brings all the features we expect from a controller for this market segment: support for features from the latest NVMe 1.3 standard, LDPC error correction robust enough to support 3D TLC and QLC NAND, enough channels and chip enable pins to support up to 8TB of flash, and support for DDR4 DRAM. Manufacturing on TSMC's 28nm process provides the transistor budget to enable all of those features while keeping power consumption reasonable.
Phison E7, E8 and E12 Controllers
Earlier Phison NVMe controllers have relied on BCH error correction and Phison's proprietary StrongECC error correction, but the Phison E12 finally brings full LDPC support. This comes at a relatively high transistor cost compared to simpler ECC schemes, but is allows for high throughput and low power encoding and decoding. The ECC engine can easily be one of the most performance-critical parts of a NVMe SSD controller, and the Phison E12's error correction should be much faster than any of their previous controllers. It should also provide a longer usable write endurance, which is why Phison also markets the E12 as their first controller suitable for use with QLC NAND.
Phison's performance estimates for their controllers are based on the contemporary NAND that they are expected to be paired with. In the case of the E12, this is Toshiba's current 64-layer 3D TLC NAND, though we may also see drives using their 96L NAND during the E12's lifetime. Using 3D TLC, the E12 promises to deliver much higher performance than the previous-generation E7 could even when using MLC NAND.
Phison E12 Engineering Sample Specifications | |
Capacity | 960 GB |
Controller | Phison PS5012-E12 |
Firmware Version | ECFM11.0 |
NAND Flash | Toshiba 256Gb 64-layer BiCS3 3D TLC |
Form-Factor | double-sided M.2-2280 |
Interface | PCIe 3.1 x4, NVMe 1.3 |
DRAM | 1 GB SK Hynix DDR4-2400 CL17 |
Our Phison E12 engineering sample features 1TB of Toshiba 64L 3D TLC NAND flash memory and 1GB of DDR4 DRAM, on a double-sided M.2 2280 card. The usable capacity is configured as 960 GB, so this drive has plenty of spare area to work with (by consumer SSD standards). Many brands that use the Phison E12 will also offer 2TB models, but performance shouldn't be much higher than this 1TB sample.
For this review, we are comparing the E12 sample against a variety of SSDs that have already hit the market. Retail drives based on the E12 should arrive very soon, probably starting with the MyDigitalSSD BPX Pro, successor to the BPX that was the most affordable E7-based SSD. MyDigitalSSD hasn't shared specific timing or pricing, but the BPX Pro should be on sale by the end of this summer. We expect pricing will be below the current top TLC-based SSDs, so the results from MLC-based drives and the Intel Optane SSD 900P represent a much higher price bracket than Phison E12 drives will occupy.
We are also expecting a wave of new SSD announcements over the next month leading in to Flash Memory Summit, August 7-9. We hope to hear plans for products using Silicon Motion's SM2262EN, an upgraded version of their SM2262 controller that is currently shipping in drives like the HP EX920 and ADATA SX8200. We will have a review of a SM2262EN engineering sample ready soon to see which new controller is likely to come out on top this fall. Unless Samsung, Western Digital or Toshiba brings out a more affordable high-end drive that delivers at least as much performance as their current flagships, the best value NVMe SSD going into the holiday season will probably be based on either the Phison E12 or the SM2262EN.
AnandTech 2018 Consumer SSD Testbed | |
CPU | Intel Xeon E3 1240 v5 |
Motherboard | ASRock Fatal1ty E3V5 Performance Gaming/OC |
Chipset | Intel C232 |
Memory | 4x 8GB G.SKILL Ripjaws DDR4-2400 CL15 |
Graphics | AMD Radeon HD 5450, 1920x1200@60Hz |
Software | Windows 10 x64, version 1709 |
Linux kernel version 4.14, fio version 3.6 | |
Spectre/Meltdown microcode and OS patches current as of May 2018 |
- Thanks to Intel for the Xeon E3 1240 v5 CPU
- Thanks to ASRock for the E3V5 Performance Gaming/OC
- Thanks to G.SKILL for the Ripjaws DDR4-2400 RAM
- Thanks to Corsair for the RM750 power supply, Carbide 200R case, and Hydro H60 CPU cooler
- Thanks to Quarch for the XLC Programmable Power Module and accessories
- Thanks to StarTech for providing a RK2236BKF 22U rack cabinet.
AnandTech Storage Bench - The Destroyer
The Destroyer is an extremely long test replicating the access patterns of very IO-intensive desktop usage. A detailed breakdown can be found in this article. Like real-world usage, the drives do get the occasional break that allows for some background garbage collection and flushing caches, but those idle times are limited to 25ms so that it doesn't take all week to run the test. These AnandTech Storage Bench (ATSB) tests do not involve running the actual applications that generated the workloads, so the scores are relatively insensitive to changes in CPU performance and RAM from our new testbed, but the jump to a newer version of Windows and the newer storage drivers can have an impact.
We quantify performance on this test by reporting the drive's average data throughput, the average latency of the I/O operations, and the total energy used by the drive over the course of the test.
Starting off, the Phison E12 delivers a much higher average data rate on The Destroyer than their previous controllers, but it's not quite enough to match the top TLC-based drives on the market. Even the Plextor M9Pe with its aging Marvell 88SS1093 controller is a bit faster. On the other hand, the HP EX920 shows that Silicon Motion also has a lot of catching up to do with their next controller.
Average and 99th percentile latencies from the Phison E12 are among the best we've seen from a TLC-based SSD. Premium drives using 3D MLC or Intel's Optane SSDs can be a bit better, but with even 99th percentile latency approaching 1ms for TLC drives, there's not much room left for meaningful improvement on these scores.
The average read latency from the Phison E12 is similar to most other high-end SSDs, but the average write latency is better than any other TLC-based SSD.
The Phison E12 turns in great scores for both 99th percentile read and write latency. The write latency particularly stands out, and it is clear that the E12 behaves well under pressure.
The Phison E12 uses far less energy on The Destroyer than the previous-generation E7-based Patriot Hellfire, and is a bit more efficient than the slower but lower-power E8-based Kingston A1000. The WD Black still holds a significant lead in power efficiency over all the other NVMe drives, but the E12 is making some progress toward that goal.
AnandTech Storage Bench - Heavy
Our Heavy storage benchmark is proportionally more write-heavy than The Destroyer, but much shorter overall. The total writes in the Heavy test aren't enough to fill the drive, so performance never drops down to steady state. This test is far more representative of a power user's day to day usage, and is heavily influenced by the drive's peak performance. The Heavy workload test details can be found here. This test is run twice, once on a freshly erased drive and once after filling the drive with sequential writes.
The average data rates from the Phison E12 on the Heavy test are good, but the exceptional empty-drive performance of the HP EX920 is well out of reach for the Phison E12. However, the E12 doesn't sacrifice full-drive performance the way the EX920 does.
The average and 99th percentile latencies from the Phison E12 are great, leaving the E12 more or less tied for the lowest-latency flash-based SSD.
The average read latency of the Phison E12 on the Heavy test is good but not exceptional like the HP EX920 (in the empty-drive test run only). The average write latency of the E12 is great, but the Samsung 970 EVO is a bit faster when the test is run on an empty drive.
The 99th percentile read latency from the Phison E12 is excellent and essentially tied for first place among flash-based SSDs when the Heavy test is run on an empty drive. When the drives are full, the WD Black takes the lead over the Phison E12 and HP EX920. The 99th percentile write latency of the Phison E12 is nothing special, except that the full-drive score is just as good.
The Phison E12 is just behind the Phison E8-based Kingston A1000 for energy usage on the Heavy test. Both drives are quite efficient for a NVMe drive, but the WD Black is still the only one that matches the efficiency of a good SATA drive.
AnandTech Storage Bench - Light
Our Light storage test has relatively more sequential accesses and lower queue depths than The Destroyer or the Heavy test, and it's by far the shortest test overall. It's based largely on applications that aren't highly dependent on storage performance, so this is a test more of application launch times and file load times. This test can be seen as the sum of all the little delays in daily usage, but with the idle times trimmed to 25ms it takes less than half an hour to run. Details of the Light test can be found here. As with the ATSB Heavy test, this test is run with the drive both freshly erased and empty, and after filling the drive with sequential writes.
The average data rates of the Phison E12 on the Light test are solidly high-end, but not record setting and the full-drive performance loss is more significant than on most competing drives.
The Phison E12's average and 99th percentile latencies on the Light test are great, but it's hardly the only drive offering this level of performance. The Samsung 970 EVO has better latency scores all around, and a clear lead for the 99th percentile latency.
The average write latencies from the Phison E12 are top-notch and similar to the best drives from Samsung and WD, but the average read latencies are slightly slower than the top drives, especially when the test is run on a full drive.
The 99th percentile read and write latencies are all excellent. The HP EX920 and the Intel Optane SSD 900P manage to have better full-drive read latency, but otherwise the Phison E12 is as good as anything at keeping latency reliably low.
The Phison E12's energy usage ranks similarly to how it ranked on the other ATSB tests: The WD Black is still the most efficient NVMe SSD and the only one that really rivals good SATA drives, and the latest Phison controllers are the next most efficient, with the low-end/low-power E8 slightly ahead of the E12.
Random Read Performance
Our first test of random read performance uses very short bursts of operations issued one at a time with no queuing. The drives are given enough idle time between bursts to yield an overall duty cycle of 20%, so thermal throttling is impossible. Each burst consists of a total of 32MB of 4kB random reads, from a 16GB span of the disk. The total data read is 1GB.
The burst random read speed of the Phison E12 is very good, but the HP EX920 still stands out as the leading flash-based SSD on this test.
Our sustained random read performance is similar to the random read test from our 2015 test suite: queue depths from 1 to 32 are tested, and the average performance and power efficiency across QD1, QD2 and QD4 are reported as the primary scores. Each queue depth is tested for one minute or 32GB of data transferred, whichever is shorter. After each queue depth is tested, the drive is given up to one minute to cool off so that the higher queue depths are unlikely to be affected by accumulated heat build-up. The individual read operations are again 4kB, and cover a 64GB span of the drive.
On the longer random read test where some higher queue depths come into play, the Phison E12's performance is a bit below par as several other TLC-based SSDs are a bit faster, even those using the same BiCS3 3D TLC.
Power Efficiency in MB/s/W | Average Power in W |
The power efficiency of the Phison E12 during the sustained random read test is decent, and the overall power consumption is only a bit higher than the Phison E8-based Kingston A1000. The only TLC SSD that's both faster and more efficient than the Phison E12 is the WD Black, which just barely faster.
The random read speed of the Phison E12 scales up nicely as queue depths increase, and the power consumption doesn't get out of hand. By QD32 a few TLC drives are a bit ahead of the Phison E12, but the only big performance gaps are relative to the MLC SSDs and the Intel Optane SSD.
Random Write Performance
Our test of random write burst performance is structured similarly to the random read burst test, but each burst is only 4MB and the total test length is 128MB. The 4kB random write operations are distributed over a 16GB span of the drive, and the operations are issued one at a time with no queuing.
The Phison E12 sets a new record for burst random write performance, improving slightly over the WD Black that formerly held a wide lead.
As with the sustained random read test, our sustained 4kB random write test runs for up to one minute or 32GB per queue depth, covering a 64GB span of the drive and giving the drive up to 1 minute of idle time between queue depths to allow for write caches to be flushed and for the drive to cool down.
On the longer random write test, the Phison E12 is in a multi-way tie for second place, slightly behind the WD Black.
Power Efficiency in MB/s/W | Average Power in W |
The Phison E12 matches the Samsung 970 EVO for power efficiency on the sustained random write test. The WD Black and Toshiba XG5 both have significantly better power efficiency with the same NAND. The WD Black is faster while drawing a bit less power, while the XG5 is significantly slower but also much less power-hungry.
Even though quite a few drives turned in very similar performance scores for the average of QD1-QD4, the actual performance scaling curves vary quite a bit. The Phison E12's random write performance is very good at QD1 and QD2, but at higher queue depths many drives overtake it and saturate at a much higher speed.
Sequential Read Performance
Our first test of sequential read performance uses short bursts of 128MB, issued as 128kB operations with no queuing. The test averages performance across eight bursts for a total of 1GB of data transferred from a drive containing 16GB of data. Between each burst the drive is given enough idle time to keep the overall duty cycle at 20%.
The burst sequential read performance of the Phison E12 is just shy of 2GB/s—much better than most TLC SSDs, but the HP EX920 has a pretty big lead at 2.4GB/s.
Our test of sustained sequential reads uses queue depths from 1 to 32, with the performance and power scores computed as the average of QD1, QD2 and QD4. Each queue depth is tested for up to one minute or 32GB transferred, from a drive containing 64GB of data. This test is run twice: once with the drive prepared by sequentially writing the test data, and again after the random write test has mixed things up, causing fragmentation inside the SSD that isn't visible to the OS. These two scores represent the two extremes of how the drive would perform under real-world usage, where wear leveling and modifications to some existing data will create some internal fragmentation that degrades performance, but usually not to the extent shown here.
On the longer sequential read test, the Phison E12 doesn't keep pace, and it isn't even the fastest controller for this kind of NAND: the Toshiba XG5 is a bit faster. The E12's performance doesn't suffer acutely from internal fragmentation like the SM2262-based HP EX920 does.
Power Efficiency in MB/s/W | Average Power in W |
The Phison E12 is reasonably power efficient on the sequential read test, but it doesn't stand out from the crowd. It is one of the few drives that draws more power when reading from a data with internal fragmentation, because that fragmentation slows most drives down to the point that they also start drawing less power overall.
The sequential read performance of the Phison E12 is mostly flat at low queue depths, and then jumps up toward saturation between QD4 and QD16. There are several other drives that require similarly high queue depths to reach full performance, but the HP EX920 does more than 3 GB/s at QD2.
Sequential Write Performance
Our test of sequential write burst performance is structured identically to the sequential read burst performance test save for the direction of the data transfer. Each burst writes 128MB as 128kB operations issued at QD1, for a total of 1GB of data written to a drive containing 16GB of data.
After seeing the Phison E12 set a record for burst random writes, it is not too surprising to see it also set the record for burst sequential write performance. Second place goes to the Samsung 970 EVO, and then it's a long way down to the next fastest TLC drive.
Our test of sustained sequential writes is structured identically to our sustained sequential read test, save for the direction of the data transfers. Queue depths range from 1 to 32 and each queue depth is tested for up to one minute or 32GB, followed by up to one minute of idle time for the drive to cool off and perform garbage collection. The test is confined to a 64GB span of the drive.
On the longer sequential write test, the Phison E12 doesn't hold on to its performance as well as the Samsung 970 EVO does, but the E12 still manages to stay ahead of the rest of the TLC drives.
Power Efficiency in MB/s/W | Average Power in W |
Several SSDs with BiCS3 3D TLC take the top spots for power efficiency during sequential writes. The Toshiba XG5 is still on top, but the Phison E12 ties the WD Black for second place. The E12 uses a bit more power than the WD Black but delivers speed to match.
The Phison E12 hits an excellent sequential write speed at QD2, but then bounces around as the drive struggles to complete garbage collection to empty the SLC cache during the idle times between test phases. The drive still stays above 1GB/s, so it isn't slowing down to anywhere near SATA speeds. Power consumption varies little; the drive is 100% busy writing to the flash throughout the test phases, even when a significant portion of those writes aren't new data coming in from the host system.
Mixed Random Performance
Our test of mixed random reads and writes covers mixes varying from pure reads to pure writes at 10% increments. Each mix is tested for up to 1 minute or 32GB of data transferred. The test is conducted with a queue depth of 4, and is limited to a 64GB span of the drive. In between each mix, the drive is given idle time of up to one minute so that the overall duty cycle is 50%.
The mixed random I/O performance of the Phison E12 is unimpressive for its intended market segment: The HP EX920 and WD Black are 20% faster overall and the Samsung 970 EVO is 40% faster. The E12 delivers only a modest improvement over the previous-generation E7 controller.
Power Efficiency in MB/s/W | Average Power in W |
The power efficiency of the Phison E12 during the mixed random I/O test is better than most TLC SSDs, but it is slightly beat out by the low-power Toshiba XG5 and top-performing Samsung 970 EVO, and they're all overshadowed by the extremely efficient WD Black.
The Phison E12 gains performance slowly as the test progresses from pure reads toward write-heavy workloads. At the very end of the test the high random write speed of the Phison E12 helps bring the average up, but not enough to overcome the drives that show much more pronounced performance growth throughout the test.
Mixed Sequential Performance
Our test of mixed sequential reads and writes differs from the mixed random I/O test by performing 128kB sequential accesses rather than 4kB accesses at random locations, and the sequential test is conducted at queue depth 1. The range of mixes tested is the same, and the timing and limits on data transfers are also the same as above.
The mixed sequential I/O performance of the Phison E12 is very good, beating all TLC drives except the Samsung 970 EVO, which is only about 7% faster overall.
Power Efficiency in MB/s/W | Average Power in W |
The power efficiency of the Phison E12 on the mixed sequential test is again very good, but still can't match the WD Black. Toshiba/SanDisk BiCS3 3D TLC seems to be very efficient for sequential I/O, but Phison's controller is a bit more power-hungry than Western Digital's in-house controller.
The Phison E12 starts the test with a middling sequential read speed, but it barely loses any performance when writes are added to the mix and instead turns around to deliver modest performance growth, peaking in the second half of the test when most drives are at their worst. Near the end, the write volume is high enough that the E12's slow garbage collection begins to hurt performance.
Power Management Features
Real-world client storage workloads leave SSDs idle most of the time, so the active power measurements presented earlier in this review only account for a small part of what determines a drive's suitability for battery-powered use. Especially under light use, the power efficiency of a SSD is determined mostly be how well it can save power when idle.
For many NVMe SSDs, the closely related matter of thermal management can also be important. M.2 SSDs can concentrate a lot of power in a very small space. They may also be used in locations with high ambient temperatures and poor cooling, such as tucked under a GPU on a desktop motherboard, or in a poorly-ventilated notebook.
Phison E12 960GB Engineering Sample NVMe Power and Thermal Management Features |
|||
Controller | Phison PS5012-E12 | ||
Firmware | ECFM11.0 | ||
NVMe Version |
Feature | Status | |
1.0 | Number of operational (active) power states | 3 | |
1.1 | Number of non-operational (idle) power states | 2 | |
Autonomous Power State Transition (APST) | Supported | ||
1.2 | Warning Temperature | 70 °C | |
Critical Temperature | 90 °C | ||
1.3 | Host Controlled Thermal Management | Supported | |
Non-Operational Power State Permissive Mode | Not Supported |
The Phison E12 implements most of the optional power and thermal management features defined by the current NVMe 1.3 standard. The two idle states advertise great power savings and reasonably quick transition times.
Phison E12 960GB Engineering Sample NVMe Power States |
|||||
Controller | Phison PS5012-E12 | ||||
Firmware | ECFM11.0 | ||||
Power State |
Maximum Power |
Active/Idle | Entry Latency |
Exit Latency |
|
PS 0 | 8.9 W | Active | - | - | |
PS 1 | 2.3 W | Active | - | - | |
PS 2 | 1.8 W | Active | - | - | |
PS 3 | 49 mW | Idle | 2 ms | 2 ms | |
PS 4 | 1.8 mW | Idle | 25 ms | 25 ms |
Note that the above tables reflect only the information provided by the drive to the OS. The power and latency numbers are often very conservative estimates, but they are what the OS uses to determine which idle states to use and how long to wait before dropping to a deeper idle state.
Idle Power Measurement
SATA SSDs are tested with SATA link power management disabled to measure their active idle power draw, and with it enabled for the deeper idle power consumption score and the idle wake-up latency test. Our testbed, like any ordinary desktop system, cannot trigger the deepest DevSleep idle state.
Idle power management for NVMe SSDs is far more complicated than for SATA SSDs. NVMe SSDs can support several different idle power states, and through the Autonomous Power State Transition (APST) feature the operating system can set a drive's policy for when to drop down to a lower power state. There is typically a tradeoff in that lower-power states take longer to enter and wake up from, so the choice about what power states to use may differ for desktop and notebooks.
We report two idle power measurements. Active idle is representative of a typical desktop, where none of the advanced PCIe link or NVMe power saving features are enabled and the drive is immediately ready to process new commands. The idle power consumption metric is measured with PCIe Active State Power Management L1.2 state enabled and NVMe APST enabled if supported.
The active idle power draw of the Phison E12 is pretty good: slightly higher than the smaller Phison E8 controller and significantly more power-hungry than many SATA SSDs, but half the draw of most other high-end NVMe SSDs. The HP EX920 draws about 20% more in its active idle state.
With all the power management features turned on, the Phison E12 can no longer match the SM2262 controller in the HP EX920 but it does provide an incremental improvement over most other high-end NVMe controllers. Given how problematic NVMe power management was for Phison's earlier controllers even after several firmware revisions, this is a very welcome result.
The idle wake-up latency is slightly faster than the drive advertises for PS3, and the the power draw was also pretty close to spec for that power state. Aside from not being able to enter PS4 on our desktop testbed (a very common limitation), the power management situation on the Phison E12 seems to be trouble-free.
Looking Forward
For months now, it has been clear that the best TLC-based SSDs are fast enough that even more expensive drives using MLC NAND or 3D XPoint memory are almost never justified. The Phison E12 controller does very little to further raise the bar for TLC SSDs, but it brings another player back in to competition at the high end. Within a few months, almost every SSD brand should be offering a respectably competitive high-end NVMe SSD. Most importantly for consumers, this will include brands such as MyDigitalSSD that are known for aggressively low pricing but have so far been limited to SATA and low-end NVMe drives when using Phison reference designs.
The Phison E12's best test results came from the write tests, where it set new records for burst sequential and random writes. Other drives with Toshiba/SanDisk 3D TLC tend to do well on write performance or power efficiency, so the Phison E12 isn't the top drive in all write-heavy tests. The E12's performance drops significantly when the SLC cache gets filled up and if the drive doesn't have enough time to finish flushing the cache before the next test phase, but even those performance drops only bring it down to about twice as fast as any SATA SSD. Phison probably could tune their SLC caching and garbage collection to offer better or more consistent sustained write performance on our synthetic tests, but it doesn't appear that such optimization is necessary for great real-world performance.
On the read side of things, the E12 offers good high-end performance but the HP EX920 with the Silicon Motion SM2262 controller and Micron 64L 3D TLC currently holds on to some impressive performance records. Silicon Motion should be able to expand that lead with their upcoming SM2262EN controller, so Phison either needs to figure out how to get better read performance out of Toshiba's 64L TLC, or hope that their 96L TLC will take care of the problem.
The Phison E12 consistently delivered very low read and write latencies on all of the ATSB tests, even when the Heavy and Light tests were run on a full drive. This applies not just to average latency but also 99th percentile latencies; the Phison E12's performance on these real-world storage workloads is consistently good.
The power efficiency of the Phison E12 is generally pretty good, even when the performance is not competitive with the top drives. The Phison E12 isn't quite as low-power as the year-old Toshiba XG5 that was the first shipping drive with this BiCS3 3D TLC NAND, but most of the tests where the E12 is more power-hungry are tests where the E12 is delivering enough extra performance to justify the power consumption. However, on almost all tests, the most efficient NVMe SSD is the impressive WD Black, which combines the same 3D TLC with Western Digital's in-house controller to deliver top performance with efficiency comparable to SATA SSDs.
Overall, the performance of the Phison E12 is competitive for today's SSD market given the price range I expect retail versions to show up in. Phison hasn't managed to leapfrog the competition in any big way, but they've caught up with the high-end for the first time since NVMe came to the consumer SSD market.
Phison may be able to continue incrementally improving performance through firmware updates, but it doesn't appear that the combination of the E12 controller and 64L TLC is at all future-proof when it comes to competing against the other drives that will be hitting the market during the second half of this year. Drives based on the Phison E12 will probably end up with more of a mid-range status within the NVMe SSD market, unable to deliver top performance but coming fairly close for a much smaller price. That should make E12 drives a pretty successful family of products with more mainstream appeal and affordability than Samsung's premium NVMe drives.
Looking back at how the Phison E7 fared as their first NVMe controller, the E12 is a huge improvement. The E7 was touted by many brands as a component for high-end SSDs, but it never really delivered. The most successful E7 products were the cheapest ones that didn't try to compete head-on with Samsung. The E7 controller also initially hit the market with relatively immature firmware, so many later products had substantially better (or at least different) performance from the first wave of releases. Phison has no obvious firmware flaws with the E12 platform, and it's probably ready to hit the stores.