Original Link: https://www.anandtech.com/show/11130/previewing-silicon-motion-sm2260-nvme-controller-with-3d-mlc-nand-512gb
Previewing Silicon Motion SM2260 NVMe Controller With 3D MLC NAND (512GB)
by Billy Tallis on February 17, 2017 9:00 AM EST
Silicon Motion's SM2260 is their first NVMe PCIe SSD controller. We've already reviewed one SSD using this controller: the Intel SSD 600p. That drive used Intel's 3D TLC NAND and set a new low for NVMe SSD prices while also offering performance that is mostly beyond the reach of any SATA SSD. Today we have an engineering sample from Silicon Motion that pairs the SM2260 controller with Micron's 3D MLC NAND—our first look at the performance of the Intel/Micron 3D MLC and a preview of what to expect from products like the ADATA XPG SX8000.
One of the first announced design wins for the SM2260 controller was Micron's Ballistix TX3 SSD, which was intended to be their first consumer PCIe SSD and a flagship product for their overhaul of the Ballistix brand. The drive was originally intended to hit the market in September 2016 but was canceled before launch. Since then, Micron has not introduced any consumer products using the MLC version of their 3D NAND, leaving the 3D TLC-based Crucial MX300 as their top consumer SSD. Intel shipped their 3D TLC-based 600p with a customized SM2260 controller, but the aging SSD 750 remains their high-end consumer NVMe offering.
With a controller that was spurned by Micron and 3D MLC NAND that its manufacturers have not tried to introduce to the consumer market, this engineering sample offers us some insight into Intel and Micron's consumer SSD strategies and the opportunity they passed up. It also offers us a preview of ADATA's XPG SX8000, their first NVMe SSD and one that is based on this controller and NAND flash combination. ADATA is the first brand to announce products with Micron's 3D MLC for the consumer market, with the XPG SX8000 NVMe SSD and the Ultimate SU900 and XPG 950 SATA SSDs, all using Silicon Motion controllers.
ADATA usually has a broader range of controller and NAND combinations than any other brand, but so far Micron is their only choice for 3D NAND for SSDs and Silicon Motion is their controller vendor. This is due in large part to Silicon Motion being the first controller vendor ready with a complete hardware and firmware reference design tuned for Micron's 3D NAND. At CES earlier this year, we also saw Mushkin announce SATA and NVMe SSDs based on the combination of Micron 3D NAND and Silicon Motion controllers.
The SM2260 controller supports a PCIe 3.0 x4 interface and implements the NVMe 1.2 protocol. The controller has a dual-core ARM Cortex-R processor and is manufactured on TSMC's 40nm LP process, an interesting juxtaposition when you consider that most of its competition is on 28nm. The controller is packaged with a thin copper heatspreader that we first saw relatively late in the chip's design cycle. LDPC error correction and TCG Opal encryption are supported.
The sample we are reviewing today is a M.2 2280 SSD with a total of 16 of Micron's 256Gb 3D MLC dies spread across four packages and the controller's 8 channels. This amount of flash would typically be used for a drive with an advertised capacity of 480GB, 500GB or 512GB, but this sample has a usable capacity closer to 515GB. This means its spare area is a little smaller than is normal—about 6.31% rather than the more common 6.85%. The firmware implements a dynamically-sized SLC cache whereas the Intel SSD 600p used a fixed-size cache.
Silicon Motion will surely face more competition this year for design wins for 3D NAND SSDs. Micron used a Marvell controller for their own Crucial MX300 SATA SSD, the only shipping SSD with Micron's 3D NAND that doesn't use a Silicon Motion controller. We expect Phison's controllers to be a popular choice once Toshiba has 3D NAND for the SSD market and Plextor is likely to continue using a mix of Marvell and Silicon Motion controllers when they begin transitioning to 3D NAND. This year we will probably also see Maxiotek score a design win for their MK8115 DRAM-less SATA SSD controller that is primarily intended for use with 3D TLC NAND. But in the meantime, Silicon Motion's SM2260 is the NVMe controller that's already shipping paired with 3D NAND.
The main competition we'll be comparing this sample against includes:
- The Patriot Hellfire 480GB, using the Phison PS5007-E7 controller and Toshiba 15nm MLC
- The Intel SSD 600p 512GB, using the SM2260 controller and Intel 3D TLC
- The Plextor M8PeG(N) 512GB, using the Marvell 88SS1093 controller and Toshiba 15nm MLC
- The Samsung 960 EVO, using Samsung 3D TLC, tested in 250GB and 1TB capacities
| AnandTech 2015 SSD Test System | |
| CPU | Intel Core i7-4770K running at 3.5GHz (Turbo & EIST enabled, C-states disabled) |
| Motherboard | ASUS Z97 Pro (BIOS 2701) |
| Chipset | Intel Z97 |
| Memory | Corsair Vengeance DDR3-1866 2x8GB (9-10-9-27 2T) |
| Graphics | Intel HD Graphics 4600 |
| Desktop Resolution | 1920 x 1200 |
| OS | Windows 8.1 x64 |
- Thanks to Intel for the Core i7-4770K CPU
- Thanks to ASUS for the Z97 Deluxe motherboard
- Thanks to Corsair for the Vengeance 16GB DDR3-1866 DRAM kit, RM750 power supply, Carbide 200R case, and Hydro H60 CPU cooler
Performance Consistency
Our performance consistency test explores the extent to which a drive can reliably sustain performance during a long-duration random write test. Specifications for consumer drives typically list peak performance numbers only attainable in ideal conditions. The performance in a worst-case scenario can be drastically different as over the course of a long test drives can run out of spare area, have to start performing garbage collection, and sometimes even reach power or thermal limits.
In addition to an overall decline in performance, a long test can show patterns in how performance varies on shorter timescales. Some drives will exhibit very little variance in performance from second to second, while others will show massive drops in performance during each garbage collection cycle but otherwise maintain good performance, and others show constantly wide variance. If a drive periodically slows to hard drive levels of performance, it may feel slow to use even if its overall average performance is very high.
To maximally stress the drive's controller and force it to perform garbage collection and wear leveling, this test conducts 4kB random writes with a queue depth of 32. The drive is filled before the start of the test, and the test duration is one hour. Any spare area will be exhausted early in the test and by the end of the hour even the largest drives with the most overprovisioning will have reached a steady state. We use the last 400 seconds of the test to score the drive both on steady-state average writes per second and on its performance divided by the standard deviation.
The SM2260 sample has relatively poor steady-state random write performance given that it uses NVMe and MLC NAND. The use of SLC caching and the lower than normal spare area of this drive both contribute to poor steady-state performance but may not significantly impair short-term performance.
The SM2260 sample has an even lower consistency score than the Intel SSD 600p, which uses basically the same controller and TLC NAND, but has substantially more spare area.
 |
|||||||||
| Default | |||||||||
| 25% Over-Provisioning | |||||||||
After the very short initial burst of great performance around 140k IOPS, the SM2260 sample transitions abruptly to a steady state that it maintains throughout the rest of the test with no long-term shifts in behavior. With extra overprovisioning reserved, there's an intermediate phase consisting of mostly performance around 80k IOPS and second burst at 140k IOPS before a higher performance but no more consistent steady state is reached.
 |
|||||||||
| Default | |||||||||
| 25% Over-Provisioning | |||||||||
Looking more closely at the steady state, the SM2260 sample is mostly shifting between four performance levels, with the most common being around 5k IOPS. However, when it goes through periodic phases of lower performance, it is stuttering hard and will often go for an entire second without completing any I/O. This is clearly poorly-managed garbage collection, possibly exacerbated by thermal throttling and definitely suffering from insufficient spare area.
With more overprovisioning, the severe stuttering is all but eliminated and the normal performance range jumps to around 24k IOPS with periods where it drops to around 6k IOPS.
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 and unlike our Iometer tests, 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.
We quantify performance on this test by reporting the drive's average data throughput, a few data points about its latency, and the total energy used by the drive over the course of the test.
The SM2260 sample's average data rate on The Destroyer is just a hair slower than the Phison-based Patriot Hellfire. This makes the SM2260 sample the slowest NVMe SSD using MLC NAND, but it's still faster than any SATA SSD.
The SM2260 sample's average service time during The Destroyer is again in last place for NVMe/MLC SSDs, but the SATA SSDs and some of the TLC-based NVMe SSDs are trailing behind by a wide margin.
The SM2260 sample is not great at avoiding high-latency outliers above 100ms and ranks behind the Samsung 850 PRO. At the 10ms threshold, the SM2260 sample performs quite well with big advantage over the Patriot Hellfire, the Intel SSD 600p and the SATA SSDs.
The SM2260 sample with 3D MLC improves significantly over the poor power efficiency of the TLC-based Intel SSD 600p, but it still ranks poorly overall. It is tied with the Phison E7-based Patriot Hellfire.
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.
The SM2260 sample's average data rate on the Heavy test is a little bit slower than the Patriot Hellfire when the test is conducted starting with an empty drive, but when starting on a full drive the SM2260 has a very slight lead. The empty drive performance of the SM2260 is still significantly better than any SATA SSD, but the full drive performance drops slightly below the Samsung 850 PRO.
The average service time of the SM2260 sample was slightly better than the Patriot Hellfire for both runs of the test. At best, the SM2260 is roughly on par with the Plextor M8Pe, OCZ RD400 and Intel SSD 750, and at its worst it still holds on to a lead over the best SATA SSDs.
Full or fresh, the SM2260 keeps latency well under control during the Heavy test, where the Patriot Hellfire began to struggle with a full drive.
The SM2260 sample uses less energy over the course of the test than the Patriot Hellfire or Plextor M8Pe, and especially the Intel SSD 600p. But other than that, the power efficiency is still poor and nowhere close to what Samsung delivers.
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.
The SM2260 sample's average data rate on the Light test once again puts it as the slowest NVMe SSD using MLC NAND, but also still faster than SATA SSDs and the Intel SSD 600p. Even Samsung's smallest 250GB 960 EVO substantially outperforms the SM2260 sample.
The average service times of the SM2260 sample put it near the bottom of its class of NVMe SSDs, but this metric also shows a wide lead over SATA SSDs.
The SM2260 sample is not in the top tier of NVMe SSDs for its ability to keep latency low, but the number of outliers above 10ms is slightly better than the OCZ RD400 and far better than SATA SSDs and the Intel 600p.
The SM2260 is essentially tied for last place in power efficiency, using only slightly less energy than the Patriot Hellfire and the Plextor M8Pe.
Random Read Performance
The random read test requests 4kB blocks and tests queue depths ranging from 1 to 32. The queue depth is doubled every three minutes, for a total test duration of 18 minutes. The test spans the entire drive, which is filled before the test starts. The primary score we report is an average of performances at queue depths 1, 2 and 4, as client usage typically consists mostly of low queue depth operations.
The SM2260 sample delivers random read performance that is on par with most of its competitors, both vetter than any SATA SSD and notably better than the Patriot Hellfire, OCZ RD400 and Intel SSD 600p NVMe SSDs.
The SM2260 sample delivers random read performance that is on par with most of its competitors, both vetter than any SATA SSD and notably better than the Patriot Hellfire, OCZ RD400 and Intel SSD 600p NVMe SSDs.
 |
|||||||||
The SM2260 scales to much higher performance at higher queue depths than the Patriot Hellfire, but doesn't hit quit as high a peak as the rest of the MLC NVMe SSDs, some of which are also near full speed at QD16.
Random Write Performance
The random write test writes 4kB blocks and tests queue depths ranging from 1 to 32. The queue depth is doubled every three minutes, for a total test duration of 18 minutes. The test is limited to a 16GB portion of the drive, and the drive is empty save for the 16GB test file. The primary score we report is an average of performances at queue depths 1, 2 and 4, as client usage typically consists mostly of low queue depth operations.
The SM2260 sample delivers better random write performance than Samsung's 950 PRO, but it can't match the more recent MLC-based NVMe SSDs.
The SM2260 also beats the Samsung 950 PRO on power consumption and efficiency, but against the more recent competition it fares poorly.
 |
|||||||||
Performance scaling for the SM2260 is rough as it runs out of SLC cache and potentially begins to thermally throttle as the test progresses to higher queue depths. Its average of around 450 MB/s during the second half is substantially less than what most of its closest competitors manage even when they are thermally limited.
Sequential Read Performance
The sequential read test requests 128kB blocks and tests queue depths ranging from 1 to 32. The queue depth is doubled every three minutes, for a total test duration of 18 minutes. The test spans the entire drive, and the drive is filled before the test begins. The primary score we report is an average of performances at queue depths 1, 2 and 4, as client usage typically consists mostly of low queue depth operations.
The SM2260 sample's sequential read performance is identical to the TLC-based Intel 600p that uses the same controller. Both fall right in the middle of the large gap between SATA SSDs and the next slowest NVMe SSD.
The story is the same for power consumption: the SM2260 sample with MLC is tied with the Intel SSD 600p with TLC. They are not as power-hungry as most NVMe SSDs, but the low performance means they are still less efficient.
 |
|||||||||
The SM2260 sample's power consumption tops out at QD4 but performance increases a little further at QD8 to just under 1500 MB/s. This performance at high queue depths is reasonable, but most of the competition can reach these speeds at much lower queue depths.
Sequential Write Performance
The sequential write test writes 128kB blocks and tests queue depths ranging from 1 to 32. The queue depth is doubled every three minutes, for a total test duration of 18 minutes. The test spans the entire drive, and the drive is filled before the test begins. The primary score we report is an average of performances at queue depths 1, 2 and 4, as client usage typically consists mostly of low queue depth operations.
The sequential write speed of the SM2260 sample is slightly slower than the top tier of SATA SSDs, which the rest of the MLC-based NVMe SSDs have no trouble beating.
The SM2260 uses about the same amount of power as its NVMe competition, but delivers much less performance for it.
 |
|||||||||
The SM2260 sample's write performance starts out at around 900 MB/s but starts dropping due to garbage collection and an exhausted SLC cache less than a third of the way through the QD1 phase of the test. The average performance continues to drop throughout the rest of the test as the drive spends an increasing portion of its time on garbage collection, but it also continues to recover periodically to the 900 MB/s level.
Mixed Random Read/Write Performance
The mixed random I/O benchmark starts with a pure read test and gradually increases the proportion of writes, finishing with pure writes. The queue depth is 3 for the entire test and each subtest lasts for 3 minutes, for a total test duration of 18 minutes. As with the pure random write test, this test is restricted to a 16GB span of the drive, which is empty save for the 16GB test file.
The SM2260 sample scores in the lower tier of NVMe SSDs for mixed random I/O, but this is still much faster than SATA SSDs. The Patriot Hellfire and OCZ RD400 have a bit of an edge, and the Plextor M8Pe is very slightly faster.
The SM2260 sample has the highest power consumption among M.2 or SATA SSDs, so it's less efficient than all of its competition.
 |
|||||||||
The SM2260 sample's performance gradually increases as the portion of writes grows. Power consumption jumps substantially in the final phase of pure writes, but the performance increase is disappointing.
Mixed Sequential Read/Write Performance
The mixed sequential access test covers the entire span of the drive and uses a queue depth of one. It starts with a pure read test and gradually increases the proportion of writes, finishing with pure writes. Each subtest lasts for 3 minutes, for a total test duration of 18 minutes. The drive is filled before the test starts.
The mixed sequential I/O performance of the SM2260 sample is a little bit slower than the next slowest MLC-based NVMe SSD and a little bit faster than the best SATA SSD.
The SM2260 sample's power consumption is a little high by NVMe standards but it isn't setting a record. The efficiency is sub-par due to the low performance.
 |
|||||||||
Performance from the SM2260 sample wobbles a little over the course of the mixed sequential I/O test but is mostly steady. Most drives perform substantially better at either end of the test when the workload is predominantly reads or writes, but the SM2260 does manage to maintain a slightly better minimum than the Patriot Hellfire's worst.
ATTO
ATTO's Disk Benchmark is a quick and easy freeware tool to measure drive performance across various transfer sizes.
 |
|||||||||
The SM2260 is close to its full write speed when the transfer size is 16kB, but full read speed takes much longer to attain and occurs at 512kB, after which thermal throttling slows things back down.
AS-SSD
AS-SSD is another quick and free benchmark tool. It uses incompressible data for all of its tests, making it an easy way to keep an eye on which drives are relying on transparent data compression. The short duration of the test makes it a decent indicator of peak drive performance.
The peak sequential read speed of the SM2260 sample as measured by AS-SSD is slightly higher than the Intel SSD 600p instead of being tied as shown by our longer IOmeter test, but they both are still behind the rest of the NVMe crowd. For write speed, the SLC cache enables the SM2260 to beat the Intel SSD 750 in addition to the 600p, but it is also slightly slower than its more recent NVMe competition.
Idle Power Consumption
Since the ATSB tests based on real-world usage cut idle times short to 25ms, their power consumption scores paint an inaccurate picture of the relative suitability of drives for mobile use. During real-world client use, a solid state drive will spend far more time idle than actively processing commands. We report two measures of idle power consumption: active idle where the SSD is not in use but has not been put in to any low-power sleep state, and idle power consumption in the deepest sleep state supported by our testbed. For NVMe SSDs, the lowest drive power state is measured but PCIe Active State Power Management (ASPM) is not used due to limitations of this motherboard. For SATA SSDs, aggressive link power management is used to put the SATA link into slumber state. Many SSDs support a deeper DevSleep state, but this cannot be engaged using ordinary desktop platforms.
The SM2260 sample has slightly worse idle power consumption than the Intel SSD 600p, but it's ahead of the Plextor M8Pe and OCZ RD400, while the Phison E7-based Patriot Hellfire is in a distant last place among M.2 PCIe SSDs. Samsung continues to deliver far better idle power that is close to what we typically see for SATA SSDs.
For active idle power without making use of any explicit power saving modes, the SM2260 does better than any PCIe SSD and better than at least one SATA SSD, but the SM2260 sample is still using a bit more power than the Intel SSD 600p.
Final Words
This SM2260 engineering sample with Micron 3D MLC is not exactly representative of any retail product, but it does hint at what we can expect when we get the ADATA XPG SX8000 in for testing, and it shows what Micron was dealing with last summer when they canceled the Ballistix TX3 SSD. If we make the very reasonable assumption that the firmware on this sample is more mature than what Micron had last summer, then it is clear that Micron made the right choice in canceling the TX3. It was not going to be able to compete at the high end of the NVMe SSD market. At the time of the planned September launch Micron was still ramping up their 3D NAND production capacity and the NAND shortage was just beginning to hit, so Micron would not have been able to offer great pricing.
Today's market is quite different from the end of last summer. Samsung has put the performance crown well out of reach, but they're leaving plenty of room for competition among more affordable NVMe SSDs. The SM2260 has already made a good showing in that category with the Intel SSD 600p with 3D TLC NAND that is no more expensive than the best SATA SSDs, but the 600p suffers under particularly heavy workloads. This SM2260 sample with MLC NAND is much better equipped to handle our more intense tests, especially our relatively long-running synthetic benchmarks. It still has some performance issues that make for low benchmark scores, but for the most part they are not important for real-world use.
The SM2260 is not the only controller competing for entry-level NVMe SSDs. We recently tested Phison's PS5007-E7 controller in the Patriot Hellfire, which is a moderately faster drive overall than this SM2260 sample. The Plextor M8Pe is even faster and has been priced reasonably when it's been in stock. To compete in today's market, SM2260 solutions will need to match or beat Phison E7 drives on price. There is room for some performance improvement from retail SM2260 SSDs over this sample, especially for sustained write speeds where this sample suffered due to its unusually small spare area. ADATA's XPG SX8000 should have the normal usable capacity for a 512GB SSD and will probably be a reasonable purchase when priced to match the 480GB Phison solutions. The SX8000 also promises to have a wider range of capacities available than most of the Phison solutions and has a decent 5-year warranty, but we'll save the final verdict for when we have performance numbers from the real thing.
The SM2260 is a nice reminder of how much the SSD market has progressed, in spite of the current NAND flash shortage that is pushing some prices up. This entry-level NVMe controller that is similar to a drive currently selling for $220 was able to beat the Intel SSD 750 and the Samsung 950 PRO on several benchmarks. Those drives debuted at $389 and $350 respectively and are thoroughly outdated in the face of the wide range of NVMe options now on the market. Even though this SM2260 sample showed relatively poor power efficiency compared to its contemporary competitors (in part due to being manufactured on an outdated but cheap 40nm process), there's no question that it is far more efficient than the enterprise-grade controller in the Intel SSD 750 that idles at 4 W and precluded the use of the M.2 form factor.
Silicon Motion hasn't had any big hits recently like they had with their SM2246EN controller back when mainstream SATA SSDs were still all using planar MLC NAND and the Crucial BX100 offered a combination of good performance and great power efficiency and price. But Silicon Motion is still in competition and has reasonable offerings for value-oriented product segments. Going forward they'll have to improve significantly and move to 28nm fabrication in order to stay competitive, but for the time being SMI-based SSDs are still definitely worth paying attention to.
Facebook
Twitter
RSS
