The M9Pe is the new flagship SSD from Plextor, and its first NVMe SSD to use 3D NAND flash memory.  The M9Pe is the successor to both the M8Pe, which used Toshiba 15nm MLC, and the M8Se, which used Toshiba 15nm planar TLC. As with most SSD product lines, as a general industry wide implementation, MLC NAND is being abandoned in the transition to 64-layer 3D NAND. The M9Pe was officially launched at the beginning of this year, but the M9Pe didn't hit the shelves until March, and supplies are still inconsistent.

Plextor in the SSD Market

Plextor is the retail SSD brand of Lite-On, one of the larger suppliers of SSDs for OEMs. Lite-On doesn't manufacture their own NAND flash memory or SSD controllers, so their primary means of product differentiation is through writing their own SSD controller firmware. This often gives them an edge against companies that simply re-brand reference designs from the controller vendors, but as the biggest players in the SSD market are increasingly vertically integrated it is getting harder for Plextor to compete. The Plextor-branded retail NVMe SSDs also offer the cosmetic distinction of optional heatspreaders and add-in card adapters with heatsinks and LEDs. Aside from those cosmetic touches, the M9Pe is essentially the same as the Lite-On CA3 client SSD we reported on previously.

Using Toshiba NAND and Marvell Controllers

Like most Plextor SSDs, the M9Pe uses Toshiba NAND flash memory. This time it is Toshiba's 64-layer third-generation 3D TLC NAND. This is Toshiba's first mainstream, mass-produced 3D NAND, branded by Toshiba as BiCS 3. The controller in the M9Pe is the same Marvell 88SS1093 "Eldora" used in the preceding M8Pe and M8Se SSDs, so the M9Pe derives almost all of its improvements from the upgraded NAND. Plextor has managed to implement NVMe 1.3 features for the M9Pe despite using an older controller that was originally designed to target the NVMe 1.1 specification.

As with other NVMe SSDs from this generation, the large 256Gbit or 512Gbit capacity of each NAND flash die hurts performance of the smallest drive capacities. The 128GB option is gone from this generation because that would only use half of the controller's eight NAND channels. The 256GB M9Pe still has distinctly lower performance specifications than the larger models, especially for write speeds. Overall, performance specs are a bit below other current high-end NVMe SSDs like the Samsung 970 EVO and Western Digital WD Black.

The five year warranty period and ~0.3 drive writes per day endurance rating are typical for this product segment.

Plextor M9Pe Series Specifications Comparison
		256 GB	512 GB	1 TB
Form Factor	M9PeY	Half-height half-length PCIe add-in card (HHHL)
	M9PeG	M.2 2280 with heatspreader
	M9PeGN	M.2 2280 without heatspreader
Controller		Marvell 88SS1093 "Eldora"
Interface		NVMe 1.3 PCIe 3.0 x4
DRAM		512MB LPDDR3		1024MB LPDDR3
NAND		Toshiba 64L BiCS3 3D TLC
Sequential Read		3000 MB/s	3200 MB/s	3200 MB/s
Sequential Write		1000 MB/s	2000 MB/s	2100 MB/s
4KB Random Read (QD32)		180k IOPS	340k IOPS	400k IOPS
4KB Random Write (QD32)		160k IOPS	280k IOPS	300k IOPS
Endurance		160 TBW 0.3 DWPD	320 TBW 0.3 DWPD	640 TBW 0.3 DWPD
Warranty		5 years
Current Price	M.2	$119.99 (47¢/GB)	$209.19 (41¢/GB)	$408.26 (40¢/GB)
	M.2 with heatspreader	$123.62 (48¢/GB)	$219.99 (43¢/GB)	$441.87 (43¢/GB)
	Add-in card	$160.72 (63¢/GB)	$245.02 (48¢/GB)	Unavailable

The M9Pe is SSD is packaged in three different variants. The cheapest M9PeGN is the bare M.2 SSD. The M9PeG adds a heatspreader to the drive that should help a bit with thermal throttling during sustained workloads but may be physically too large to fit in some notebook systems. The M9PeY mounts the M.2 module on a PCIe x4 half-height half-length adapter card with a large heatsink and a row of RGB LEDs. The previous generation M8Pe and M8Se featured red and blue LEDs respectively, so RGB is the obvious way to make the M9PeY even gaudier. The RGB LEDs cannot be controller by the user and instead are managed by the same ITE 8259FN chip we've seen on several motherboards that also feature RGB LEDs. Aside from the lighting, the heatsink of the M9PeY is almost identical to the M8SeY, with a blue plastic insert swapped for red.

Since it first hit the shelves, prices on the M9Pe have come down to roughly match other top SSDs when comparing against the M9PeGN variant that lacks a heatsink. The versions with extra cooling carry a substantial premium. Newegg is the only major online retailer selling the M9Pe, and they're not currently listing the 1TB M9PeY add-in card model.

Our review samples are the 512GB M9PeGN (M.2) and the 1TB M9PeY (AIC), but we've also swapped the drives in the adapter card to present results for both capacities running as standalone M.2 drives and in the adapter with the heatsink and LEDs.

Competition for the M9Pe

Most of the competition for the M9Pe uses substantially different hardware. The second-generation Western Digital WD Black uses the same BiCS3 3D TLC NAND as the M9Pe but features Western Digital's new in-house controller. The M9Pe shows us what the new WD Black would have been like if WD's new controller hadn't been ready and if they had stuck with the Marvell controller for another generation. The WD Black and the Samsung 970 EVO showcase the strengths of a vertically integrated strategy with the NAND, SSD controller, firmware and drive all designed by the same company.

A little lower on the price scale are several competitors based around Silicon Motion's second generation SM2262 controller, a huge improvement over their earlier SM2260 controller. Intel uses the SM2262 in their own 760p SSD, and cheaper drives based on the SM2262 reference design are being sold by ADATA and HP. These drives offer performance close enough to the top SSDs that it's hard to pass up the savings.

For comparison against other market segments, this review also includes results from the ultra-premium Intel Optane SSD 900P and Samsung 960 PRO, the entry-level NVMe MyDigitalSSD SBX, and the mainstream SATA Crucial MX500.

This is our last SSD review using our 2017 SSD test suite. Future reviews will feature benchmark results from a testbed with Meltdown and Spectre patches and microcode updates current as of this month, and a few other tweaks to the test procedure. The Windows 10 version will still be 1709, because Microsoft has not yet fixed all the new bugs introduced in the NVMe driver in Windows 10 version 1803.

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.

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 of the Plextor M9Pe on The Destroyer is a clear step forward from Plextor's previous TLC-base NVMe SSD (the M8Se), but there are plenty of other recent TLC-based drives that perform much better, and the M8Pe with planar MLC is still faster than the M9Pe.

The average latency of the 512GB M9Pe matches that of the Intel SSD 760p, but without a 1TB 760p to compare against the 1TB M9Pe's average latency looks quite poor. The 99th percentile latencies of both M9Pe capacities are also worse than they should be, but don't stand out quite as much from the relevant competition.

The average write latency scores from the M9Pe rank a bit better than the average read latency scores, but in both cases the M9Pe is clearly slower than the top tier drives from Samsung and Western Digital.

The 99th percentile read and write latency scores from the M9Pe are an improvement over the planar TLC-based M8Se, but the scores still aren't great. The 512GB model has trouble staying ahead of the Crucial MX500 and MyDigitalSSD SBX.

Energy consumption from the M9Pe during The Destroyer is on the high side, but not as bad as the M8Se was. The Samsung 970 EVO's energy consumption is also a bit higher than the M9Pe. The WD Black impresses the most, with higher performance than the Samsung drives while using half the energy.

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 Plextor M9Pe offers fairly low average data rates on the Heavy test, by the standards of contemporary high-end SSDs. However, it doesn't lose too much performance when full, which allows the 512GB model to score a win over the 500GB Samsung 970 EVO.

The average latency scores from the M9Pe place it at the bottom of the NVMe segment, but still ahead of the SATA drives. The 99th percentile scores of the 1TB model are worse than the other 1TB NVMe drives but not too bad overall, while the 512GB M9Pe is nearly tied with the Crucial MX500 SATA drive.

The average read latency scores from the Plextor M9Pe show an unusual inversion, with better performance when the drive is full. Anomalies like this are often an artifact of the drive lying about when it's done with a secure erase operation, leading to the empty-drive test run starting too soon. Even discounting those results, the M9Pe doesn't look competitive. The average write latency scores look more sensible, but are also bad news for the M9Pe's aspirations to high-end status.

None of the 99th percentile read or write latency scores from the Plextor M9Pe are actually good by the standards of a current-generation high-end SSD, but the M9Pe does avoid the problems that befall some of the drives when they're full.

Energy usage from the M9Pe again falls between the results from its MLC and TLC based predecessors, and aren't great overall. The 500GB Samsung 970 EVO also requires a lot of energy, while its larger counterpart does quite well when its SLC cache isn't filled.

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 from the Plextor M9Pe on the Light test again show it taking less of a performance hit from being full than what most TLC-based drives suffer. Unfortunately, most other recent TLC-based drives have enough of an advantage in the empty-drive performance that their full-drive scores are still ahead of the M9Pe.

All of these SSDs have average latencies well below 1ms on the Light test, and even the SATA drive is more than fast enough. Some of the 99th percentile scores are high enough to potentially matter, but the 1.8ms from the M9Pe 512GB when the drive is full isn't quite that bad.

Average read latencies from the M9Pe are close enough to the fastest drives that the differences don't matter at all. Average write latencies are clearly worse for the M9Pe than for the top tier of drives, but even 100µs is insignificant for this kind of workload.

The 99th percentile read latency scores from the M9Pe stand out a bit for the full-drive test runs. The 99th percentile write latency scores show that the M9Pe avoids the problems that the Intel 760p and MyDigitalSSD SBX have with occasionally dropping down to SATA performance levels.

Energy usage is once more worse than average for the M9Pe, but several of the fastest drives also use more energy on the Light test than the M9Pe does.

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 Plextor M9Pe is good at either of the two tested capacities. Only a handful of flash-based SSDs outperform the 1TB M9Pe, and the 512GB model is just over 5% slower.

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 speeds of the M9Pe are better than most TLC drives, though the Samsung 970 EVO is a bit faster still. The MLC-based M8Pe is also slightly faster than the M9Pe on this test.

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

The power efficiency of the Plextor M9Pe during random reads is quite poor, but it's only slightly worse than the Samsung 970 EVO and the previous generation of Plextor drives.

Plextor M9PeY 1TB (M.2)MyDigitalSSD SBX 512GBIntel SSD 760p 512GBSamsung 960 PRO 1TBSamsung 960 PRO 512GBSamsung 970 EVO 500GBSamsung 970 EVO 1TBPlextor M8PeGN 512GBPlextor M8SeY 1TB (M.2)Plextor M9PeGN 512GB (M.2)Western Digital WD Black 1TB (3D NAND)Intel Optane SSD 900P 280GBPlextor M9PeY 1TB (AIC)Plextor M9PeY 512GB (AIC)Crucial MX500 1TBCrucial MX500 500GB

At low queue depths, the random read performance of the two tested capacities of the Plextor M9Pe is nearly identical, and the larger model doesn't gain a significant lead until queue depths are well above 8.

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 burst random write speeds from the Plextor M9Pe are about average for a high-end NVMe SSD. Several Samsung drives score about the same as the M9Pe. The WD Black and Intel 760p lead with the fastest write caches.

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 Plextor drives are all clustered together with very similar performance, though it appears the heatsink helps a little bit here. The Samsung NVMe drives are all performing much better than the Plextor drives, and the WD Black holds the spot as the top flash-based SSD on this test.

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

Even with 3D NAND, the Plextor drives still deliver relatively poor power efficiency during random writes. The controller and DRAM both need to be updated to newer, lower-power designs.

Plextor M9PeY 1TB (M.2)MyDigitalSSD SBX 512GBIntel SSD 760p 512GBSamsung 960 PRO 1TBSamsung 960 PRO 512GBSamsung 970 EVO 500GBSamsung 970 EVO 1TBPlextor M8PeGN 512GBPlextor M8SeY 1TB (M.2)Plextor M9PeGN 512GB (M.2)Western Digital WD Black 1TB (3D NAND)Intel Optane SSD 900P 280GBPlextor M9PeY 1TB (AIC)Plextor M9PeY 512GB (AIC)Crucial MX500 1TBCrucial MX500 500GB

The 512GB M9Pe hits its full random write speed at QD4, while the 1TB model is able to continue increasing performance up to QD8, leaving it with a much higher overall limit. Both capacities are well-behaved once reaching saturation, and don't appear to be having major garbage collection issues.

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 speeds from the Plextor M9Pe are worse than most NVMe SSDs including the low-end MyDigitalSSD SBX, but at least it's still well above SATA speeds (unlike the Plextor M8Se).

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 longer sequential read test, the M9Pe fares a little better as the MyDigitalSSD SBX and Intel SSD 760p drop down to about the same level. As with the burst test, the Samsung drives dominate.

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

The Plextor M9Pe's power efficiency on the sequential read speed is below average, but it's not actually the worst-scoring NVMe SSD for this test—that distinction belongs to the Intel SSD 760p, which is barely more efficient than the Crucial MX500 SATA SSD.

Plextor M9PeY 1TB (M.2)MyDigitalSSD SBX 512GBIntel SSD 760p 512GBSamsung 960 PRO 1TBSamsung 960 PRO 512GBSamsung 970 EVO 500GBSamsung 970 EVO 1TBPlextor M8PeGN 512GBPlextor M8SeY 1TB (M.2)Plextor M9PeGN 512GB (M.2)Western Digital WD Black 1TB (3D NAND)Intel Optane SSD 900P 280GBPlextor M9PeY 1TB (AIC)Plextor M9PeY 512GB (AIC)Crucial MX500 1TBCrucial MX500 500GB

Both M9Pe models hit full sequential read speed at QD4, which is relatively late, but not as slow a ramp up as the WD Black shows.

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.

The burst sequential write speeds from the Plextor M9Pe are good, with only the WD Black and a few Samsung drives scoring better. The older generation of Plextor drives performed much worse on this test.

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 M9Pe manages to stay ahead of the low-end NVMe drives but otherwise doesn't perform well. Capacity is a major factor in performance here: the 1TB model is 80% faster than the 512GB model.

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

With almost average performance, the 1TB M9Pe has reasonable but not top-tier power efficiency on the sustained sequential write test. The 512GB model has the second-worst efficiency score out of this batch of drives.

Plextor M9PeY 1TB (M.2)MyDigitalSSD SBX 512GBIntel SSD 760p 512GBSamsung 960 PRO 1TBSamsung 960 PRO 512GBSamsung 970 EVO 500GBSamsung 970 EVO 1TBPlextor M8PeGN 512GBPlextor M8SeY 1TB (M.2)Plextor M9PeGN 512GB (M.2)Western Digital WD Black 1TB (3D NAND)Intel Optane SSD 900P 280GBPlextor M9PeY 1TB (AIC)Plextor M9PeY 512GB (AIC)Crucial MX500 1TBCrucial MX500 500GB

The sequential performance of the Plextor M9Pe increases slightly from QD1 to QD2 and is steady thereafter, with no obvious signs of SLC write cache exhaustion or intermittent garbage collection getting in the way.

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 performance of the Plextor M9Pe on the mixed random I/O test is better than the SATA drive and the low-end NVMe drive, but is otherwise disappointing and not competitive with the other high-end SSDs or even the preceding M8Pe.

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

Power efficiency from the M9Pe on the mixed random I/O test is poor, but similar to the earlier Plextor drives. The drives with newer controllers not only perform better, but most also require much less power.

Plextor M9PeY 1TB (M.2)MyDigitalSSD SBX 512GBIntel SSD 760p 512GBSamsung 960 PRO 1TBSamsung 960 PRO 512GBSamsung 970 EVO 500GBSamsung 970 EVO 1TBPlextor M8PeGN 512GBPlextor M8SeY 1TB (M.2)Plextor M9PeGN 512GB (M.2)Western Digital WD Black 1TB (3D NAND)Intel Optane SSD 900P 280GBPlextor M9PeY 1TB (AIC)Plextor M9PeY 512GB (AIC)Crucial MX500 1TBCrucial MX500 500GB

The two capacities of the Plextor M9Pe show very similar performance scaling across the random I/O test, with very slow performance growth as the proportion of writes increases, until near the end of the test when there are enough writes to perform effective caching and coalescing. Most drives show much more substantial performance growth during the first 80% of 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 test again puts the M9Pe in the situation of being the slowest drive that's trying to be high-end.  It does outperform the SATA drive and the MyDigitalSSD SBX, but drives like the WD Black and the Samsung NVMe SSDs are far faster.

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

Power efficiency from the M9Pe on the mixed sequential I/O test is actually worse than either of its predecessors, and worse than the Crucial MX500 SATA drive. The Intel SSD 760p and MyDigitalSSD SBX are slightly worse than the M9Pe, and the WD Black has more than 2.5 times the performance per Watt of the M9Pe.

Plextor M9PeY 1TB (M.2)MyDigitalSSD SBX 512GBIntel SSD 760p 512GBSamsung 960 PRO 1TBSamsung 960 PRO 512GBSamsung 970 EVO 500GBSamsung 970 EVO 1TBPlextor M8PeGN 512GBPlextor M8SeY 1TB (M.2)Plextor M9PeGN 512GB (M.2)Western Digital WD Black 1TB (3D NAND)Intel Optane SSD 900P 280GBPlextor M9PeY 1TB (AIC)Plextor M9PeY 512GB (AIC)Crucial MX500 1TBCrucial MX500 500GB

Neither capacity of the M9Pe does well during the more read-heavy half of the test, and only the 1TB model gets a substantial boost at the end of the test where the write caches are most effective.

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.

Plextor M9Pe NVMe Power and Thermal Management Features
Controller	Marvell 88SS1093 "Eldora"
Firmware	1.04
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	83 °C
	Critical Temperature	85 °C
1.3	Host Controlled Thermal Management	Supported
	Non-Operational Power State Permissive Mode	Supported

The Plextor M9Pe supports the full range of power and thermal management features from the latest NVMe standard, with Plextor's firmware implementing features that didn't exist when the Marvell 88SS1093 controller first hit the market. The power state table looks quite reasonable with two idle states that shouldn't take too long to enter or leave. The maximum active power rating of 8W is a bit high for an M.2 drive, but the M8Pe was also rather power-hungry at its worst.

Plextor M9Pe NVMe Power States
Controller	Marvell 88SS1093 "Eldora"
Firmware	1.04
Power State	Maximum Power	Active/Idle	Entry Latency	Exit Latency
PS 0	8 W	Active	-	-
PS 1	4.5 W	Active	-	-
PS 2	3 W	Active	-	-
PS 3	70 mW	Idle	1 ms	5 ms
PS 4	10 mW	Idle	5 ms	50 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 LEDs on the M9PeY's add-in card adapter ruin its idle power scores, but the M.2 drive itself behaves about the same as the older M8Se. The M9Pe doesn't manage to drop down to the deepest sleep states on our testbed, but enabling idle power management does cut its draw in half.

The M9Pe wakes up from its ~1W intermediate idle in well under a millisecond, which guarantees that this level of power management won't harm performance.

Conclusion

The past year has seen almost all remaining SSD product lines with MLC NAND switch to 64-layer 3D TLC NAND. This is the only significant hardware change the Plextor M9Pe brings over its predecessor, along with updates to the NVMe 1.3 standard in its controller.

The second-generation Western Digital WD Black SSD showed that Toshiba/SanDisk BiCS 3 64-layer 3D TLC can be very fast, but that performance is not on display with the Plextor M9Pe. The WD Black owes much of its success to Western Digital's new in-house NVMe SSD controller that has replaced the Marvell controller used by the first-generation WD Black. It is that older Marvell controller which powers Plextor's NVMe SSDs, including the M9Pe.

It appears that the Plextor M9Pe is held back by the outdated SSD controller. The Marvell 88SS1093 "Eldora" was one of the first NVMe SSD controllers to hit the market. Until recently, it was also clearly the best M.2-sized controller on the open market, but Silicon Motion's SM2262 controller has now taken that spot. (Samsung and Western Digital have clearly superior controllers for their own SSDs, but they aren't sharing them with other vendors.) Marvell has recently announced the 88SS1098 "Zao" as a successor to the 88SS1093, but it is too late for the first wave of 64L 3D NAND SSDs. With a fourth CPU core, an upgraded LDPC engine, a 50% faster flash interface and support for LPDDR4 DRAM, the 88SS1098 should bring Marvell back into competition, but by the time it is ready, the Phison E12 and Silicon Motion SM2262EN will also be shipping. In the meantime, the 88SS1093 "Eldora" controller needs to be retired.

At Computex next month we hope to get a preview of Plextor's plans for a successor to the M9Pe. It's possible they may have something in the works based on the Phison E12 controller and Toshiba NAND or Silicon Motion SM2262(EN) and Micron NAND, but if they stick with Marvell for their high-end NVMe SSDs, we won't see anything new ship until 2019.

That leaves Plextor with pretty much the slowest flagship SSD of any brand. The M9Pe is consistently behind the Intel 760p, WD Black and Samsung 970 EVO. The Plextor M9Pe can usually match the performance of the MLC-base M8Pe, but on heavier workloads it falls behind. The performance situation isn't all bad: the M9Pe is definitely a step up from entry-level NVMe SSDs like the MyDigitalSSD SBX, and also from the planar TLC-based M8Se. Unlike low-end NVMe SSDs, the M9Pe doesn't have any acute weaknesses where performance drops below that of a decent SATA SSD. Generally speaking, the M9Pe is fast enough that it is hard to notice a difference between it and the top SSDs without breaking out special-purpose storage benchmark software. The M9Pe still manages to qualify as a high-end NVMe SSD, just not the very top tier.

Power efficiency is also a challenge for the M9Pe. It seldom demonstrates much improvement over its predecessors that used planar NAND. From other vendors, we've seen drives in this generation show impressive power efficiency improvements as they move to 64L 3D NAND, and we were hoping to see the same from the M9Pe. Users concerned about power consumption will obviously not be interested in the PCIe add-in card version of the M9Pe with more than 1W of LEDs, but even the bare M.2 module is a poor choice for battery-powered systems.

The big selling points of the M9PeY variant are the adapter card's large heatsink and the RGB LED lighting. The LEDs are purely a matter of taste (or lack thereof), but M.2 cooling solutions are often marketed as having some functional benefit. In our testing, the heatsink didn't make any difference to performance. Even our synthetic benchmarks are designed to focus on conditions representative of real-world client usage, and at low queue depths and for sustained transfers that are limited to just tens of GB at a time, the M9Pe doesn't overheat. Long-running tests that simply read or write continuously at very high queue depths may be able to show the heatsink making a difference after a few minutes, but such a test doesn't say anything about real-world use cases. Even when transferring data to or from another high-end NVMe SSD, it's hard to produce the conditions necessary to overheat the drive: either filesystem overhead slows things down, or you quickly run out of data to transfer.

NVMe SSD Price Comparison (2018-05-24)
	120-128GB	240-256GB	400-512GB	960-1200GB
Plextor M9PeGN (M.2)		$119.99 (47¢/GB)	$209.19 (41¢/GB)	$408.26 (40¢/GB)
Plextor M9PeG (M.2 with heatspreader)		$123.62 (48¢/GB)	$219.99 (43¢/GB)	$441.87 (43¢/GB)
Plextor M9PeY (add-in card)		$160.72 (63¢/GB)	$245.02 (48¢/GB)
ADATA XPG SX8200		$99.99 (42¢/GB)	$189.99 (40¢/GB)
HP EX920		$109.99 (43¢/GB)	$179.99 (35¢/GB)	$299.99 (29¢/GB)
MyDigitalSSD SBX	$52.99 (41¢/GB)	$84.99 (33¢/GB)	$157.99 (31¢/GB)	$309.99 (30¢/GB)
Intel SSD 760p	$82.95 (65¢/GB)	$116.99 (46¢/GB)	$219.45 (43¢/GB)	$414.20 (40¢/GB)
Samsung 970 EVO		$109.99 (44¢/GB)	$199.99 (40¢/GB)	$399.99 (40¢/GB)
Western Digital WD Black (3D NAND)		$109.99 (44¢/GB)	$199.99 (40¢/GB)	$449.99 (45¢/GB)

While even the Samsung 970 EVO would save money and deliver better performance than the M9Pe, the most interesting deals at the moment are on the drives using the Silicon Motion SM2262 controller and Intel/Micron 64L 3D TLC. The Intel SSD 760p's prices are still elevated compared to the launch prices, but the HP EX920 and ADATA SX8200 offer similar performance for much lower prices. In particular, the current sale price of $299.99 for the 1TB EX920 is astounding, and I'm surprised Newegg hasn't sold out of the EX920 in the time it takes me to write this. 

The current pricing of the Plextor M9Pe is unfortunate. The bare M.2 drive is priced slightly higher than the Samsung 970 EVO and WD Black. The heatspreader or add-in card options only drive the price further beyond reasonable. Based on performance and power consumption, the M9Pe should be the second-cheapest drive on this chart, not the most expensive. To some extent, it isn't a surprise that Plextor would have trouble beating the vertically-integrated NAND manufacturers on price, but the HP EX920 and ADATA SX8200 show that it's possible while still budgeting for a high-end controller. Lite-On has enough purchasing power to get NAND at similar prices to ADATA, so I suspect they're simply not trying hard to make the M9Pe price competitve, while they direct most of their NAND supplies to their enterprise and OEM drives. This would explain the spotty availability of the M9Pe, with the 1TB AIC model being unavailable at the moment and the rest only available from Newegg. Either that or they have a customer contract and had no major plans to bring this unit to retail.