Western Digital is bringing 3D NAND to their high-end consumer SSD family with the launch today of two new NVMe SSDs featuring SanDisk's 64-layer 3D TLC NAND flash memory. As with the SATA SSDs that first brought 3D NAND to their consumer portfolio, Western Digital is releasing the same drive under both their WD and SanDisk brands. Under the stickers, the hardware is identical.

The names are recycled and familiar: the WD Black and SanDisk Extreme PRO. The first WD Black SSD was Western Digital's first consumer NVMe product. It used a Marvell controller and 15nm planar TLC NAND, and ended up near the bottom of the performance rankings for NVMe SSDs, with no appreciable performance advantage over SATA SSDs for heavier workloads. The SanDisk Extreme PRO name hasn't been used on an internal SSD for quite a while, but it carries a strong legacy: the original SanDisk Extreme PRO was a top-tier SATA SSD with MLC NAND and was competitive with the Samsung 850 PRO. The SATA SanDisk Extreme PRO hit the market right before the 850 PRO and was the first consumer SSD to carry a 10-year warranty, forcing Samsung to follow suit with the 850 PRO.

Re-using product names like this without any clear generational indicator or model year will cause confusion. Western Digital has at least ensured that the new drives are using different capacities from their predecessors: the first-generation WD Black was 256GB and 512GB, the original Extreme PRO was 240GB, 480GB and 960GB, and the new WD Black and SanDisk Extreme PRO are 250GB, 500GB and 1000GB. Still, last year's WD Black will be coexisting in the marketplace for several months with this year's model, and the two are very different products.

The new WD Black and SanDisk Extreme PRO SSDs are based on the same platform as the SN720 business/OEM SSD Western Digital announced earlier this year. In addition to the major advance of switching from 15nm planar TLC NAND to 64-layer BiCS 3D NAND, the new SSDs also feature Western Digital's own new SSD controller instead of using a controller from Marvell. This is a major shift toward vertical integration for Western Digital/SanDisk, and is the best strategy for Western Digital to differentiate their products in a market crowded with dozens of brands sourcing their controllers or the entire drive design from the same small handful of vendors.

Western Digital WD Black and SanDisk Extreme PRO Specifications
Capacity		250 GB	500 GB	1 TB
WD Black Model		WDS250G2X0C	WDS500G2X0C	WDS100T2X0C
SanDisk Extreme PRO Model		-	SDSSDXPM2-500G	SDSSDXPM2-1T00
Form Factor		M.2 2280 Single-Sided
Interface		NVMe PCIe 3 x4
Controller		Western Digital in-house
NAND		SanDisk 64-layer 3D TLC
DRAM		SK Hynix DDR4-2400
Sequential Read		3000 MB/s	3400 MB/s	3400 MB/s
Sequential Write		1600 MB/s	2500 MB/s	2800 MB/s
4KB Random Read		220k IOPS	410k IOPS	500k IOPS
4KB Random Write		170k IOPS	330k IOPS	400k IOPS
Power	Peak (10µs)	9.24 W	9.24 W	9.24 W
	PS3 Idle	70 mW	70 mW	100 mW
	PS4 Idle	2.5 mW	2.5 mW	2.5 mW
Write Endurance		200 TBW 0.4 DWPD	300 TBW 0.3 DWPD	600 TBW 0.3 DWPD
Warranty		5 years
MSRP		$119.99 (48¢/GB)	$229.99 (46¢/GB)	$449.99 (45¢/GB)
Amazon Price		$119.99 (48¢/GB)	$226.75 (45¢/GB)	$449.99 (45¢/GB)

The performance specifications of the new WD Black and SanDisk Extreme PRO promise a high-end drive, with sequential read speeds of 3+ GB/s even on the smallest 250GB model, and high random access specifications on the 500GB and larger models. Write endurance ratings are a reasonable 0.3-0.4 drive writes per day for five years. The MSRPs position the WD Black directly against the Samsung 960 EVO and above most other recent consumer NVMe SSDs—the fast-growing entry-level NVMe segment is what most brands are focusing on at the moment.

The WD Black can at least momentarily hit the power limits of the M.2 form factor, but it doesn't feature any heatspreader. Instead, Western Digital is using an uncommon layout that places the controller in the middle of the stick with NAND flash memory on both sides of the controller. This was deemed adequate to prevent overheating, and has the side effect of making it easier to route the 8 channels from the controller to the NAND.

The new drives will initially be available in capacities from 250GB to 1TB, though the SanDisk-branded versions won't include the smallest 250GB model. These drives should all be shipping by the end of the month. Western Digital has not mentioned any plans for a 2TB models, but since they have already announced a 2TB SN720 they obviously have the option to quickly deploy a 2TB WD Black or SanDisk Extreme PRO model if the demand is sufficient.

AnandTech 2017/2018 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 3.1

• 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.

The Western Digital NVMe Architecture - NAND & Controller

The most interesting aspects of the WD Black 3D NAND SSD come from the new in-house SSD controller and, for the first time, the usage of the 64-layer BICS 3D NAND in one of their retail NVMe SSDs. The 64-layer 3D NAND (BiCS 3) is SanDisk/Toshiba's third-generation 3D NAND. The first two generations had 24 and 48 layers respectively, but saw extremely limited release and neither generation was manufactured in quantities sufficient to displace Toshiba/SanDisk 15nm planar NAND. The production of the 64-layer version is now fully ramped up, and we saw Western Digital use it in the SanDisk Ultra 3D and WD Blue 3D SATA SSDs late last year.

As a refresher, BiCS uses a charge-trap, stacked design that alleviates almost all shortcomings of planar NAND. Bit density can increase from one 3D generation to the next, thanks to predictable scaling in both vertical and lateral dimensions. Availability of a greater number of electrons per NAND cell in the charge trap design compared to planar NAND's floating gate design (at very small geometries) ensures that the reliability and endurance of a 3D NAND cell is almost always better than that of leading-edge planar NAND. The downside is the substantially higher capital investment required to upgrade the 2D NAND producing fab equipment to manufacture 3D NAND wafers.

BiCS 4 (96-layer) is also ramping up production. Most of the 3D NAND volume from Western Digital this year is expected to be the 64-layer BiCS 3 NAND, which is what we see in the WD Black 3D NAND SSD that is being reviewed today.

On the controller front, Western Digital has opted to move from a Marvell solution to an in-house design. The primary reason cited was that they could create an architecture that was optimized for BiCS flash - in particular, the controller wouldn't need to support NAND from other vendors, and, the controller could be architected with the future plans for the BiCS line in mind. This makes sense from a flagship perspective - NVMe SSDs push the performance limits, and it is essential to be able to extract every last bit of performance possible from the available NAND. It is no surprise that almost all flash vendors have their own controller for their flagship NVMe SSDs - Samsung uses its own controllers across all its SSDs and Intel uses its own controller for the SSD 900p (Optane). Even mainstream and low-end SSDs from top tier vendors (using merchant controller silicon) come with firmware customized in-house. Vertical integration (starting from flash fabrication and IC assembly to system integration in the form of NVMe or SATA SSDs) allows vendors to optimize performance for their customers.

The new controller has a tri-core architecture (probably using Arm Cortex-R cores) fabricated in a 28nm process. It is designed to be scalable - the current controller can interface with the host using a PCIe 3.0 x4 link, or an x2 link as in the Western Digital SN520. The architecture of the controller also allows future products using variants to come to market faster and with newer features. It also allows Western Digital to segment their NVMe product stack. The controller in the Western Digital Black 3D NAND SSD is optimized for client workloads including PC gaming and high-performance commercial applications. Western Digital expects this new controller architecture to last at least until NVMe SSDs move beyond PCIe 3 x4 interfaces.

Compared to other NVMe SSD controllers that come with a large number of CPU cores, Western Digital is relying less on firmware and more on hardware accelerators to perform the host-to-flash communication (NVMe command processing and data transfer from flash to the host). Power and thermal management tasks are also done without the aid of the CPU cores. The firmware is used only for selective command processing (such as fetching of S.M.A.R.T details, the flash translation layer (FTL) algorithm, and handling exceptions). Avoiding the CPU for all the performance-critical tasks is a means to ensure that the controller is not power hungry.

The WD NVMe SSD Architecture also implements tiered caching (nCache) to improve performance. The WD Black 3D NAND NVMe SSD uses nCache 3.0, which appears to include a host of updates over nCache 2.0.

Prior to going into the details, it is interesting to take a look at how nCache has evolved over the years. In its first generation, nCache was designed to cache the NAND mapping table and small writes (less than 4KB) in a SLC segment on the die. In the second version (first introduced in the SanDisk Ultra II), we saw the move to cache writes of all sizes in the SLC segment first. The size of the SLC cache was also increased to 5GB for every 120GB of user space. nCache 2.0 also implemented an on-chip-copy for the folding mechanism (migrating data from the SLC cache to the main TLC area) that freed up the controller from managing the data-heavy portions of the nCache operations. nCache 2.0 also placed the NAND mapping table in the SLC cache area. The end result being that nCache 2.0 had to route all writes through the SLC cache; this affected sustained write performance greatly, because the SLC cache had to be flushed out to the TLC segment before newer data could be written to it.

Meanwhile for current-generation products, the performance of 3D TLC is much better than planar TLC when it comes to sustained writes. The endurance is better as well. As a result, rather than sending everything through the SLC cache as with nCache 2.0, nCache 3.0 allows writes to go directly to the TLC space after exhaustion of the SLC segment.

The migration of data from the SLC cache to TLC blocks is performed mostly when the drive is idle, unless the host operating system uses the non-operational power state permissive mode feature of NVMe 1.3 to request that the drive defer background processing. That said, the evacuation policy is aggressive in order to maintain high availability of the SLC blocks. This causes the SSD to perform more writes than it would if data was kept in the cache longer, but Western Digital is confident that their flash has sufficient endurance.

Another major departure from nCache 2.0 seems to be the absence of the on-chip-copy feature available in the previously-used planar TLC dies. Given that 3D TLC allows for direct access during sustained writes, the on-chip-copy feature for folding purposes is not needed.

Finally, Western Digital's nCache 3.0 still uses a fixed size SLC cache, which saves the drive the trouble of converting blocks between SLC and TLC usage. Western Digital has not disclosed the SLC cache sizes for the WD Black, but it appears that our 1TB samples have about 20GB of SLC cache.

Like any good modern SSD controller, Western Digital's new architecture features multiple layers of error correction. The first three layers are different LDPC-style error correction codes for handling increasing bit error rates, which come at the cost of increasing power consumption and decreased performance. The base level of error correction is an LDPC code that is tuned to offer higher throughput and with lower power requirements than the BCH error correction that nearly all SSDs used before TLC NAND began to take over the market. This lowest layer of error correction is the only one needed during normal operation for most of the drive's lifespan, and this LDPC engine is responsible for less than 10% of the controller's power consumption. The second and third layers of error correction are intended to handle the increasing error rates of a drive that is nearing the end of its write endurance, and these codes are also entirely handled by dedicated hardware on the controller without taking the performance hit of involving the processor cores.

For handling severe data loss that cannot be recovered by the three layers of LDPC, the controller also performs traditional RAID5-like XOR parity. This can handle the failure of defects affecting multiple NAND pages, but consumer SSDs don't include enough excess flash to survive the complete failure of an entire NAND die. Data integrity is also protected by the use of ECC on all of the controller's SRAM and on the external DRAM.

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.

The average data rate from the new WD Black on The Destroyer is almost as fast as Samsung's TLC-based 960 EVO and their newer PM981 OEM drive. Where the original WD Black NVMe SSD was clearly a low-end NVMe drive and no faster than SATA SSDs on this test, the new WD Black is competitive at the high end.

The average latencies from the WD Black are competitive with Samsung's TLC drives, and the 99th percentile latencies are the fastest we've seen from any flash-based SSD for this capacity class.

The average read latencies from the WD Black on The Destroyer are as good as any flash-based SSD we've tested. Average write latencies are great but Samsung's top drives are still clearly faster.

The WD Black has the best 99th percentile read latency scores aside from Intel's Optane SSD 900P, but the 99th percentile write latency scores are only in the second tier of drives.

The load power consumption of the new WD Black is a huge improvement over the previous SSD to bear this name. The new model uses less than half as much energy over the course of The Destroyer, putting it in first place slightly ahead of the Toshiba XG5.

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 new WD Black SSD on the Heavy test are essentially tied with the Samsung 960 EVO. Premium drives like the Samsung 960 PRO and Intel Optane SSD 900P are faster, but the WD Black and SanDisk Extreme PRO NVMe SSDs still clearly belong in the high-end market segment.

The average and 99th percentile latency scores from the WD Black on the Heavy test are among the best from any flash-based SSD. The 99th percentile write latency of the WD Black shows much less performance loss from a full drive than the Toshiba XG5 or Samsung 960 EVO.

The WD Black is one of the top drives for average read latency, and the average write latency is only slightly higher than that of the Samsung 960 EVO. The performance hit when the test is run on a full drive is no worse than what most MLC-based drives suffer.

Western Digital's new controller architecture provides great QoS for read operations, with 99th percentile latencies lower than any of the competing flash-based SSDs. The 99th percentile write latencies are top notch but don't stand out from the crowd.

The WD Black and SanDisk Extreme PRO join the Toshiba XG5 as some of the few NVMe SSDs that offer load power efficiency comparable to good SATA SSDs. The total energy used during the heavy test is only slightly higher than the Crucial MX500 and Western Digital's own SATA drives with the same 64L 3D TLC NAND.

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 WD Black's average data rates on the Light test are slightly slower than the Samsung 960 EVO when the test is run on an empty drive, and a bit faster when the drive is full. The Samsung PM981 is the only drive that has a clear lead in both cases, and even then it isn't a very big margin. The worst-case performance here from the new WD Black is substantially faster than the best-case from last year's WD Black.

The average latencies from the WD Black during the Light test are as low as any SSD offers. The 99th percentile latencies are not quite as fast as Samsung's best drives offer, except that the full-drive performance is better than the 960 EVO.

There are quite a few SSDs with average read latency scores that are close to or slightly better than the WD Black, and even the low-end NVMe SSDs keep the average read latency down to a fraction of a millisecond on the Light test. The average write latencies from the WD Black are essentially tied for first place with Samsung's drives.

The WD Black offers great 99th percentile write latency on the Light test as its SLC cache never fills. The 99th percentile read latency doesn't rank quite as high, but the full-drive score is very good.

As with the Heavy test, the only NVMe SSD we've tested that can match the WD Black's power efficiency is the Toshiba XG5. These drives get the job done much faster than a SATA drive without using any more energy.

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 performance of the WD Black isn't exceptional, but it is an improvement over the original WD Black SSD and is only slightly behind the Samsung 960 EVO.

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.

The sustained random read performance of the WD Black is a small improvement over last year's model, but not quite enough to catch up to Samsung. In addition, the recent Intel 760p also comes out slightly ahead of the WD Black.

Power Efficiency in MB/s/W	Average Power in W

The power efficiency of the WD Black during random reads is better than any other TLC drive as it barely draws any more power than a SATA drive during this test.

Western Digital WD Black 1TB (3D NAND)Intel SSD 760p 512GBSamsung 960 EVO 1TBSamsung 960 PRO 1TBPatriot Hellfire 480GBSamsung PM981 1TBOCZ RD400 1TBADATA XPG SX8000 512GBWestern Digital WD Black 512GB (2D NAND)SanDisk Extreme Pro 1TBToshiba XG5 1TBSamsung 860 EVO 2TB M.2Crucial MX500 1TBSanDisk Ultra 3D 1TBIntel Optane SSD 900P 280GBWestern Digital WD Black 7200RPM 1TB

At higher queue depths, the Samsung drives build a small performance lead over the WD Black, but most other drives fall far behind as the queue depth increases.

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.

Our WD Black sample oddly returned a substantially better burst random write score than the SanDisk Extreme PRO that should be identical. Since both scores are at the top of the chart, unusually high variance doesn't actually present a problem.

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.

The new WD Black offers top-tier performance on the sustained random write test, well ahead of Samsung's current retail offerings and just barely behind the PM981 OEM drive that Samsung's next generation retail drives will be based upon. Last year's WD Black was just barely faster than SATA drives.

Power Efficiency in MB/s/W	Average Power in W

The overhaul of the NAND and the controller has taken the WD Black from the bottom of the efficiency chart with last year's model to the very top, where it has a small lead over the Toshiba XG5 and Samsung 960 PRO.

Western Digital WD Black 1TB (3D NAND)Intel SSD 760p 512GBSamsung 960 EVO 1TBSamsung 960 PRO 1TBPatriot Hellfire 480GBSamsung PM981 1TBOCZ RD400 1TBADATA XPG SX8000 512GBWestern Digital WD Black 512GB (2D NAND)SanDisk Extreme Pro 1TBToshiba XG5 1TBSamsung 860 EVO 2TB M.2Crucial MX500 1TBSanDisk Ultra 3D 1TBIntel Optane SSD 900P 280GBWestern Digital WD Black 7200RPM 1TB

The WD Black's random write performance saturates at QD4 while the Samsung drives and several other models continue improving and can hit much higher performance levels at high queue depths. However, the WD Black has all the random write performance it needs at the more important low queue depths.

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 WD Black is several times higher than last year's model, but doesn't come close to setting any records.

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.

On the sustained sequential read test, the Samsung NVMe drives have a clear lead over the WD Black, which is tied with Toshiba's drives.

Power Efficiency in MB/s/W	Average Power in W

In terms of power efficiency for sequential reads, the WD Black is much closer to the top drives, with the exception of the Samsung 960 PRO.

Western Digital WD Black 1TB (3D NAND)Intel SSD 760p 512GBSamsung 960 EVO 1TBSamsung 960 PRO 1TBPatriot Hellfire 480GBSamsung PM981 1TBOCZ RD400 1TBADATA XPG SX8000 512GBWestern Digital WD Black 512GB (2D NAND)SanDisk Extreme Pro 1TBToshiba XG5 1TBSamsung 860 EVO 2TB M.2Crucial MX500 1TBSanDisk Ultra 3D 1TBIntel Optane SSD 900P 280GBWestern Digital WD Black 7200RPM 1TB

The sequential read performance of the WD Black starts out rather poor at QD1 but grows steadily all the way up to QD16, by which point it is outperforming everything except the Optane SSD. The Toshiba XG5 shows similar scaling behavior but can't quite keep pace with the WD Black.

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.

As with the burst random write test, our two samples show surprising differences in burst sequential write speeds. The difference amounts to the WD Black/SanDisk Extreme PRO either being tied for second place with the Samsung 960 EVO, or almost tied with the PM981 that the 960 EVO's replacement will be based on.

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.

The sustained sequential write performance of the WD Black is not quite the best, but it is well ahead of everything except the best drives from Samsung and Intel. The WD Black is almost twice as fast as the Toshiba XG5 that uses essentially the same flash.

Power Efficiency in MB/s/W	Average Power in W

Despite not having the best performance on the sequential write test, the WD Black is the clear winner on the efficiency metric. With power draw of just over 4W it isn't close to being the least power-hungry drive, but it get so much done on that budget that the efficiency score beats everything else.

Western Digital WD Black 1TB (3D NAND)Intel SSD 760p 512GBSamsung 960 EVO 1TBSamsung 960 PRO 1TBPatriot Hellfire 480GBSamsung PM981 1TBOCZ RD400 1TBADATA XPG SX8000 512GBWestern Digital WD Black 512GB (2D NAND)SanDisk Extreme Pro 1TBToshiba XG5 1TBSamsung 860 EVO 2TB M.2Crucial MX500 1TBSanDisk Ultra 3D 1TBIntel Optane SSD 900P 280GBWestern Digital WD Black 7200RPM 1TB

The sequential write speed of the WD Black is quite steady across the range of queue depths, with just a small increase from QD1 to QD2 and no signs of degraded performance from excessive garbage collection after the SLC cache is full.

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 WD Black offers great mixed random I/O performance, but it is still slightly slower overall than the best drives from Samsung, and the Optane SSD is in an entirely different league.

Power Efficiency in MB/s/W	Average Power in W

The WD Black's power efficiency on the mixed random I/O test is about the same as that of the Samsung 960 PRO, and close to the Optane SSD in spite of the vast difference in absolute performance level.

Western Digital WD Black 1TB (3D NAND)Intel SSD 760p 512GBSamsung 960 EVO 1TBSamsung 960 PRO 1TBPatriot Hellfire 480GBSamsung PM981 1TBOCZ RD400 1TBADATA XPG SX8000 512GBWestern Digital WD Black 512GB (2D NAND)SanDisk Extreme Pro 1TBToshiba XG5 1TBSamsung 860 EVO 2TB M.2Crucial MX500 1TBSanDisk Ultra 3D 1TBIntel Optane SSD 900P 280GBWestern Digital WD Black 7200RPM 1TB

The performance of the WD Black grows very slowly as the workload shifts from reads toward writes, but near the end of the test the good SLC write caching implementation gives the WD Black steep gains. Power consumption is under 2W for most of the test and doesn't quite reach 4W at the very end.

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 workload performance of the WD Black is surprisingly good, just barely behind the Optane SSD and far ahead of almost all flash-based SSDs.

Power Efficiency in MB/s/W	Average Power in W

The WD Black draws about the same power as other SSDs during the mixed sequential test, and combined with the great performance that translates to a huge lead in power efficiency.

Western Digital WD Black 1TB (3D NAND)Intel SSD 760p 512GBSamsung 960 EVO 1TBSamsung 960 PRO 1TBPatriot Hellfire 480GBSamsung PM981 1TBOCZ RD400 1TBADATA XPG SX8000 512GBWestern Digital WD Black 512GB (2D NAND)SanDisk Extreme Pro 1TBToshiba XG5 1TBSamsung 860 EVO 2TB M.2Crucial MX500 1TBSanDisk Ultra 3D 1TBIntel Optane SSD 900P 280GBWestern Digital WD Black 7200RPM 1TB

The performance scaling pattern of the WD Black on the mixed sequential test is quite unusual. Many drives show a bathtub curve with peak performance at either end of the test when the workload is pure reads or pure writes, and the worst performance in the middle of the test. By contrast, the WD Black starts out rather slowly but rapidly speeds up during the first half of the test, and stays near full performance for the entire second half.

Power Management

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.

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.

Like most NVMe SSDs, the WD Black has a fairly high active idle power draw—the cost of keeping a PCIe 3 x4 link active. The active idle power is a bit higher than the previous WD Black SSD but is in line with drives from Samsung, Toshiba and Phison.

Enabling all the advanced PCIe and NVMe power management features doesn't have the desired effect on the WD Black SSD. The drops by almost half, but it should have dropped by at least an order of magnitude. The original WD Black SSD used aggressive power management whether or not the operating system requested it. The new WD Black seems to be unable to save much power when used on our desktop testbed, no matter what NVMe power states are requested. We will work with Western Digital to try to isolate the cause of this poor behavior. In the meantime, the WD Black is hardly the only NVMe drive where power management has problems out of the box, but Intel and Samsung have managed to produce drives that achieve very low idle power on our testbed with little or no tuning required.

Since the WD Black is clearly unable to engage its full array of power management capabilities on our testbed, it is unsurprising to see that its wake-up latency is quite short. It is not the minimal ~15µs we usually observe from drives that aren't enabling any power savings at all, but ~230µs is still a very quick wake-up from sleep.

Conclusion

We knew from tests last year of Western Digital's SATA drives and the Toshiba XG5 that the SanDisk/Toshiba 64-layer 3D TLC was a huge improvement over their planar NAND, and possibly was even the fastest and most power efficient TLC NAND yet. It is now clear that those drives weren't even making the best possible use of that flash. With Western Digital's new in-house controller, the BiCS 3 TLC really shines. The new WD Black and SanDisk Extreme PRO are unquestionably high-end NVMe SSDs that match the Samsung 960 EVO and sometimes even beat the 960 PRO.

There are very few disappointing results from the WD Black. Even when it isn't tied for first or second place, it performs well above the low-end NVMe drives. The two biggest problems appear to be a poor start to the sequential read test, and another round of NVMe idle power management bugs to puzzle through. Almost all NVMe drives have at least some quirks when it comes to idle power management, in stark contrast to the nearly universal and flawless support among SATA drives for at least the slumber state and usually also DevSleep (which cannot be used on desktops). The power efficiency of the WD Black under load is excellent, so it is clear that the Western Digital NVMe controller isn't inherently a power hog. Whatever incompatibility the WD Black's power management currently has with our testbed won't matter to other desktop users, and hopefully isn't representative of today's notebooks. The bigger surprises from the WD Black are when it performs much better than expected, especially during the mixed sequential I/O test where nothing comes close.

Samsung established an early lead in the NVMe SSD race and has held on to their top spot as many brands have tried and failed to introduce high-end NVMe SSDs with either planar NAND or the lackluster first-generation Intel/Micron 3D NAND. None of those SSDs was a more obvious underachiever than the original WD Black NVMe SSD from last year, which used 15nm planar TLC and could barely outperform a decent SATA drive. The first WD Black SSD didn't deserve Western Digital's high-performance branding. This new WD Black is everything last year's model should have been, and it should be able to stay relevant throughout this year even when Samsung gets around to releasing the successors to the 960 PRO and 960 EVO—which they really need to do soon.

NVMe SSD Price Comparison
	120-128GB	240-256GB	400-512GB	960-1200GB
WD Black (3D NAND) SanDisk Extreme PRO		$119.99 (48¢/GB)	$226.75 (45¢/GB)	$449.99 (45¢/GB)
Intel SSD 760p	$88.32 (69¢/GB)	$122.25 (48¢/GB)	$223.26 (44¢/GB)	$471.52 (46¢/GB)
Samsung 960 PRO			$327.99 (64¢/GB)	$608.70 (59¢/GB)
Samsung 960 EVO		$119.99 (48¢/GB)	$199.99 (40¢/GB)	$449.99 (45¢/GB)
WD Black (2D NAND)		$104.28 (41¢/GB)	$182.00 (36¢/GB)
Plextor M9Pe		$119.99 (47¢/GB)	$213.43 (42¢/GB)	$408.26 (40¢/GB)
MyDigitalSSD SBX	$59.99 (47¢/GB)	$99.99 (39¢/GB)	$159.99 (31¢/GB)	$339.99 (33¢/GB)
Toshiba OCZ RD400	$109.99 (86¢/GB)	$114.99 (45¢/GB)	$309.99 (61¢/GB)	$466.45 (46¢/GB)

The MSRPs for the WD Black roughly match current street prices for the Samsung 960 EVO, which is exactly what the WD Black should be competing against. Neither drive has a clear overall performance advantage in the 1TB capacity we've analyzed with this review, though the WD Black has a modest power efficiency advantage (our idle power problems notwithstanding). Since release, the Intel 760p has also climbed up to this price range, and it doesn't belong there.

The Plextor M9Pe is finally available for purchase after a paper launch early this year. It uses Toshiba's 64L TLC and a Marvell controller, so it closely represents what this year's WD Black would have been without Western Digital's new in-house controller. We will have preformance results for the M9Pe soon.

Western Digital's long years working to develop 3D NAND and their new NVMe controller have paid off. They're once again a credible contender in the high end space, and their latest SATA SSDs are doing pretty well, too. This year's SSD market now has serious competition in almost every price bracket.