Samsung’s SM951 PCIe SSD reviewed

PC enthusiasts give little credence to arbitrary product segmentation. We want the best parts, and we don’t care if they were never meant for our machines. From time to time, this desire has driven us to adopt enterprise-grade hardware with more brawn than consumer-grade gear. It’s also taken us into mobile territory, where lower-power components have appealing characteristics for smaller form factors and quieter cooling.

Now, imagine getting the best of both of those worlds—performance that trounces typical desktop parts in a tiny, mobile-friendly package. That’s the basic idea behind Samsung’s SM951 SSD, which promises speeds over 2GB/s from an M.2 gumstick small enough to pose with Lego minifigs.

The SM951 was never intended to be sold directly to end users. Instead, it was designed for big notebook makers to put into premium systems. But Newegg and other online vendors still stock the so-called OEM drive, making it an intriguing option for anyone seeking wicked-fast PCIe storage. We’ve tested the SM951 to see whether it’s worth appropriating for premium PCs, and the answer isn’t entirely straightforward. Let me explain.

Essential details

The SM951 is the follow-up to Samsung’s XP941, an older M.2 gumstick with a quad-lane PCIe Gen2 interface. This new generation cranks the same number of lanes up to Gen3 speeds, effectively doubling the aggregate bandwidth to 4GB/s. To put that figure into perspective, consider that the dual-lane Gen2 interface behind the M.2 slots on most motherboards is limited to a mere 1GB/s.

Only a handful of X99 and Z97 boards have “Ultra,” “Turbo,” or otherwise amped-up M.2 slots with enough bandwidth to fully exploit the SM951. Quad-lane slots should be more common on next-gen Skylake boards due later this summer. In the meantime, the SM951 can be mounted on an M.2 adapter card and plugged into any full-sized PCIe slot.

There are actually two versions of the drive. The initial release uses the familiar AHCI protocol, while a newer variant adheres to the NVM Express standard. Our focus today is on the AHCI model, which is compatible with a broader range of motherboards and available from a wider selection of vendors. We also have the NVMe edition in our labs, but it doesn’t work with conventional secure-erase tools, preventing us from wiping the drive between benchmarks. Secure-erasing helps ensure repeatable results, so we’re holding off on testing the NVMe version for now.

Like just about every other SSD, including its forebear, the SM951 sports an eight-channel flash controller. The chip’s internal cores run at 500MHz, up a healthy 120MHz from the core frequency in the old XP941.

The spec sheet claims end-to-end data protection, but it doesn’t promise to preserve in-flight data if the power cuts out unexpectedly. Hardware-accelerated encryption is also missing from the feature set. The drive supports PCIe’s L1.2 sleep state, which purportedly lowers power consumption to just two milliwatts, but the savings shouldn’t matter in desktop systems.

Samsung says the drive’s NAND is fabbed on a “10-nanometer class” process, which means the geometry could be as large as 19 nm. The chips use traditional planar tech rather than the company’s multi-layer 3D V-NAND. They pack two bits per cell and 16GB per die.

Capacity Die config Max sequential (MB/s) Max Random (IOps) Price $/GB
Read Write Read Write
128GB 8 x 16GB 2000 600 90k 70k $130 $1.02
256GB 16 x 16GB 2150 1200 90k 70k $240 $0.94
512GB 32 x 16GB 2150 1500 90k 70k $400 $0.78

The 512GB flagship is rated to hit a blistering 2150MB/s with sequential reads and 1500MB/s with writes. Those speeds fall short of the interface’s peak theoretical throughput, but they’re still faster than what you can expect from competing M.2 drives and plenty of full-sized PCIe SSDs. They’re also much better than the random I/O specs, which are average even when compared to SATA SSDs.

Samsung doesn’t publish an endurance rating for the SM951, but you probably don’t have to worry about burning through the flash. The results of our endurance experiment show that modern MLC drives can survive hundreds of terabytes of writes with barely a scratch. Some can even withstand petabytes of writes without issue.

For a cutting-edge PCIe SSD that’s not even supposed to be on the market, the SM951 is surprisingly affordable. The top capacity rings in at under $0.80/GB, which is less than the going rate for Intel’s 750 Series, the only other PCIe Gen3 drive on the market right now.

Online listings for the SM951 advertise three-year warranty coverage, but that’s between you and the retailer rather than you and Samsung. The warranty is a couple years shy of the five-year coverage typically attached to premium SSDs.

The SM951 is sold as a bare drive, sans accessories and software. It’s not supported by the SSD Magician utility Samsung provides for its consumer-oriented SSD, but third-party tools can still read the SMART data. HD Sentinel shows the following attributes:

Between the media wearout indicator and the counters for bad blocks, uncorrectable errors, and flash reserves, there’s loads of useful information for monitoring drive health over the long haul. Too bad some of the attributes are obfuscated behind vendor-specific labels.

With the basics covered, we can move on to our performance tests. You won’t believe the shocking results on the next page—no, seriously.


The competition

Before delivering on that BuzzFeed bait, I should take a moment to introduce the competition. We’ve tested the SM951 against a bunch of PCIe SSDs, including its XP941 predecessor. The results also include Plextor’s M6e, an early M.2 drive with a much slower dual-lane Gen2 interface.

Intel’s 750 Series matches the SM951’s Gen3 interface, but it comes on a much larger expansion card with a beefy heatsink. This product of datacenter trickle-down is derived from the server-grade DC P3700, which we’ve included strictly for reference. The P3700 sells for roughly two bucks per gig and devotes a much larger percentage of flash to overprovisioned area. It’s a very different class of PCIe SSD.

Speaking of a different class, the results also cover a stack of the latest SATA SSDs, plus the old Intel X25-M. This collection should give us broader context for the SM951’s performance.

IOMeter — Sequential and random performance

IOMeter fuels much of our new storage suite, including our sequential and random I/O tests. These tests are run across the full extent of the drive at two queue depths. The QD1 tests simulate a single thread, while the QD4 results emulate a more demanding desktop workload. (87% of the requests in our old DriveBench 2.0 trace of real-world desktop activity have a queue depth of four or less.) Clicking the buttons below the graphs switches between the different queue depths.

Our sequential tests use a relatively large 128KB block size.

The SM951 beats the 750 Series with reads—and quite comfortably at QD1. It’s substantially slower in the write speed tests, though. Samsung’s new hotness even slips behind the old XP941 when slinging sequential writes at QD4.

“New hotness” is more than just a reference to Men in Black; the data suggest the SM951 is throttling to prevent overheating. The graphs above show the median result from a trio of three-minute averages, but IOMeter also logs transfer rates for every second of each run. A sampling of that data is presented below, along with results from a separate test that cooled the drive with direct airflow from a 120-mm fan.

Yikes. Although the uncooled config reaches 2000MB/s in the read speed test, it only holds on for about 25 seconds before plunging below 100MB/s. The slowdown lasts for less than ten seconds, after which the drive ramps up for another spurt at top speed.

Sequential writes are evidently more problematic, forcing the SM951 to spend more time in the valleys than it does at the peaks. Similar oscillations are also evident with sequential writes at QD1—but not with reads at that queue depth. Meanwhile, the air-cooled config maintains top speed over the duration of each test.

The uncooled XP941 experiences similar slowdowns, just to a lesser degree. It’s the only other SSD we’ve seen exhibit this behavior.

For what it’s worth, our IOMeter tests run consecutively, with no breaks between them. The drive is also filled with a test file immediately before the benchmark begins, effectively conditioning it with a sequential write workload. The overall pattern of peaks and valleys is consistent from one run to the next, but the duration of the initial burst is different in each run, likely due to whatever residual heat is left over from the previous test.

Next, we’ll turn our attention to performance with 4KB random I/O. We’ve reported average response times rather than raw throughput, which we think makes sense in the context of system responsiveness.

Impressively, the SM951 has the quickest read response times of the bunch, beating even Intel’s best datacenter SSD. The Samsung drive’s read response times are consistent across each test run, as well.

Oscillations appear again in the random write tests, where the SM951 is less competitive with the Intel PCIe drives. To be fair, though, the differences between the SSDs amount to just fractions of a millisecond. Also, all the drives suffer a decline in random write performance over time. We explore that slowdown more closely on the following page.


IOMeter — Sustained and scaling I/O rates

Our sustained IOMeter test hammers drives with 4KB random writes for 30 minutes straight. It uses a queue depth of 32, which should result in higher speeds that saturate each drive’s overprovisioned area more quickly. This lengthy—and heavy—workload isn’t indicative of typical PC use, but it provides a sense of how the drives react when they’re pushed to the brink.

We’re reporting IOps rather than response times for these tests. Click the buttons below the graph to switch between SSDs.

The SM951 exceeds its official specifications at the beginning of the test, but it suffers a brief slowdown during the initial burst. After that, it doesn’t take long for the drive to run out of overprovisioned area, causing the random write rate to fall by more than an order of magnitude. Performance oscillates within a narrow range for the remainder of the test.

All the other SSDs fire a frenetic early barrage before falling back to a much slower steady-state rate. To show the data in a slightly different light, we’ve graphed the peak random write rate and the average, steady-state speed over the last minute of the test.

Despite beating most of the competition at the beginning, the SM951 falls to the middle of the pack over the last minute. At least it trumps the XP941 handily in both cases. Samsung’s SATA-based 850 Pro has a higher steady-state rate, though, and the Intel 750 Series is way ahead across the board.

Our final IOMeter test examines performance scaling across a broad range of queue depths. We ramp all the way up to a queue depth of 128. Don’t expect AHCI-based drives to scale past 32, though; that’s the maximum depth of their native command queues.

We use a database access pattern comprising 66% reads and 33% writes, all of which are random. The test runs after 30 minutes of continuous random writes that put the drives in a simulated used state. Click the buttons below the graph to switch between the different drives. And note that the P3700 plot uses a much larger scale.

To Samsung’s credit, the SM951 fares much better than the XP941. But its performance barely scales up as the load intensifies, while some of the other SSDs rise to the occasion. Intel’s 750 Series runs away from the SM951 at higher queue depths, as do a bunch of SATA drives.

IOMeter’s per-second results for this test show signs of throttling on the SM951 and XP941. Overheating doesn’t appear to be a problem on for the other SSDs, including those that achieve much higher I/O rates.


TR RoboBench — Real-world transfers

RoboBench trades synthetic tests with random data for real-world transfers with a range of file types. Developed by our in-house coder, Bruno “morphine” Ferreira, this benchmark relies on the multi-threaded robocopy command build into Windows. We copy files to and from a wicked-fast RAM disk to measure read and write performance. We also cut the RAM disk out of the loop for a copy test that transfers the files to a different location on the SSD.

Robocopy uses eight threads by default, and we’ve also run it with a single thread. Our results are split between two file sets, whose vital statistics are detailed below. The compressibility percentage is based on the size of the file set after it’s been crunched by 7-Zip.

  Number of files Average file size Total size Compressibility
Media 459 21.4MB 9.58GB 0.8%
Work 84,652 48.0KB 3.87GB 59%

The media set is made up of large movie files, high-bitrate MP3s, and 18-megapixel RAW and JPG images. There are only a few hundred files in total, and the data set isn’t amenable to compression. The work set comprises loads of TR files, including documents, spreadsheets, and web-optimized images. It also includes a stack of programming-related files associated with our old Mozilla compiling test and the Visual Studio test on the next page. The average file size is measured in kilobytes rather than megabytes, and the files are mostly compressible.

RoboBench’s write and copy tests run after the drives have been put into a simulated used state with 30 minutes of 4KB random writes. The pre-conditioning process is scripted, as is the rest of the test, ensuring that drives have the same amount of time to recover.

Read speeds are up first. Click the buttons below the graphs to switch between one and eight threads.

The SM951 battles for podium spots with the Intel PCIe drives. It’s well ahead of the XP941 and every other SSD we’ve tested.

That trend continues with writes. The SM951 beats the 750 series with eight threads in flight, but it’s a little slower in the single-threaded tests. Either way, the drive performs better than not only its predecessor, but also the entire SATA field.

There are no surprises in RoboBench’s copy tests. Again, the SM951 hangs out with the Intel PCIe drives at the front of the pack.

RoboBench only reports the overall transfer rate for each run, so we can’t check for slowdowns during the test. The run-to-run scores show no evidence of throttling, though. The scripted test sequence also reboots before each run and then waits a couple minutes for the OS to finish loading. That pre-programmed delay, coupled with the fact that the SM951 completes some of the RoboBench tests in as little as seven seconds, may keep thermals within acceptable bounds.


Boot times

Thus far, all of our tests have been conducted with the SSDs connected as secondary storage. This next batch uses them as system drives.

We’ll start with boot times measured two ways. The bare test depicts the time between hitting the power button and reaching the Windows desktop, while the loaded test adds the time needed to load four applications—Avidemux, LibreOffice, GIMP, and Visual Studio Express—automatically from the startup folder. Our old boot tests only focused on the time required to load the OS, but these new ones cover the entire process, including drive initialization.

The SM951 is the fastest-booting SSD we’ve tested to date. It gets to the Windows desktop just a smidgen ahead of Plextor’s M6e—and several seconds faster than the XP941. The Intel 750 Series boots much slower than even the SATA SSDs.

Load times

Next, we’ll tackle load times with two sets of tests. The first group focuses on the time required to load larger files in a collection of desktop applications. We open a 790MB 4K video in Avidemux, a 30MB spreadsheet in LibreOffice, and a 523MB image file in GIMP. In the Visual Studio Express test, we open a 159MB project containing source code for the LLVM toolchain. Thanks to Rui Figueira for providing the project code.

There’s little reason to call out individual results. The SM951 bounces around the standings in our load-time tests, but all the SSDs are evenly matched, with typically less than a second separating the fastest examples from the slowest ones. At least with traditional desktop applications, PCIe SSDs don’t have appreciably quicker load times than even ancient SATA drives like the X25-M.

The following page is loaded with detailed system specifications and other testing details. It’s there mostly for reference, so we won’t be offended if you skip to the conclusion.


Test notes and methods

Here are the essential details for all the drives we tested:

  Interface Flash controller NAND
Crucial BX100 500GB SATA 6Gbps Silicon Motion SM2246EN 16-nm Micron MLC
Crucial MX200 500GB SATA 6Gbps Marvell 88SS9189 16-nm Micron MLC
Intel X25-M G2 160GB SATA 3Gbps Intel PC29AS21BA0 34-nm Intel MLC
Intel 335 Series 240GB SATA 6Gbps SandForce SF-2281 20-nm Intel MLC
Intel 730 Series 480GB SATA 6Gbps Intel PC29AS21CA0 20-nm Intel MLC
Intel 750 Series 1.2TB PCIe Gen3 x4 Intel CH29AE41AB0 20-nm Intel MLC
Intel DC P3700 800GB PCIe Gen3 x4 Intel CH29AE41AB0 20-nm Intel MLC
Plextor M6e 256GB PCIe Gen2 x2 Marvell 88SS9183 19-nm Toshiba MLC
Samsung 850 EV0 250GB SATA 6Gbps Samsung MGX 32-layer Samsung TLC
Samsung 850 EV0 1TB SATA 6Gbps Samsung MEX 32-layer Samsung TLC
Samsung 850 Pro 500GB SATA 6Gbps Samsung MEX 32-layer Samsung MLC
Samsung XP941 256GB PCIe Gen2 x4 Samsung S4LN053X01 19-nm Samsung MLC
Samsung SM951 512GB PCIe Gen3 x4 Samsung S4LN058A01X01 1x-nm Samsung MLC
Samsung 850 Pro 500GB SATA 6Gbps Samsung MEX 32-layer Samsung MLC
OCZ Vector 180 240GB SATA 6Gbps Indilinx Barefoot 3 M10 A19-nm Toshiba MLC
OCZ Vector 180 960GB SATA 6Gbps Indilinx Barefoot 3 M10 A19-nm Toshiba MLC

All the SATA SSDs were connected to the motherboard’s Z97 chipset. The M6e was connected to the Z97 via the motherboard’s M.2 slot, which is how we’d expect most folks to run that drive. Since the SM951 and XP941 require more lanes, they were connected to the CPU via a PCIe adapter card. The 750 Series and DC P3700 were hooked up to the CPU via the same full-sized PCIe slot.

As a reward for not skipping to the conclusion, here’s a bonus shot of the SM951:

We used the following system for testing:

Processor Intel Core i5-4690K 3.5GHz
Motherboard Asus Z97-Pro
Firmware 1304
Platform hub Intel Z97
Platform drivers Chipset:


Memory size 16GB (2 DIMMs)
Memory type Adata XPG V3 DDR3 at 1600 MT/s
Memory timings 11-11-11-28-1T
Audio Realtek ALC1150 with drivers
System drive Corsair Force LS 240GB with S8FM07.9 firmware
Storage Crucial BX100 500GB with MU01 firmware

Crucial MX200 500GB with MU01 firmware

Intel 335 Series 240GB with 335u firmware

Intel 730 Series 480GB with L2010400 firmware

Intel DC P3700 800GB with 8DV10043 firmware

Intel X25-M G2 160GB with 8820 firmware

Plextor M6e 256GB with 1.04 firmware

OCZ Vector 180 240GB with 1.0 firmware

OCZ Vector 180 960GB with 1.0 firmware

Samsung 850 EVO 250GB with EMT01B6Q firmware

Samsung 850 EVO 1TB with EMT01B6Q firmware

Samsung 850 Pro 500GB with EMXM01B6Q firmware

Samsung XP941 256GB with UXM6501Q firmware

Samsung SM951 512GB with BXW2500Q firmware

Power supply Corsair Professional Series AX650 650W
Operating system Windows 8.1 Pro x64

Thanks to Asus for providing the systems’ motherboards, Intel for the CPUs, Adata for the memory, and Corsair for the system drives and PSUs. And thanks to the drive makers for supplying the rest of the SSDs.

We used the following versions of our test applications:

Some further notes on our test methods:

  • To ensure consistent and repeatable results, the SSDs were secure-erased before every component of our test suite. For the IOMeter database, RoboBench write, and RoboBench copy tests, the drives were put in a simulated used state that better exposes long-term performance characteristics. Those tests are all scripted, ensuring an even playing field that gives the drives the same amount of time to recover from the initial used state.

  • We run virtually all our tests three times and report the median of the results. Our sustained IOMeter test is run a second time to verify the results of the first test and additional times only if necessary. The sustained test runs for 30 minutes continuously, so it already samples performance over a long period.

  • Steps have been taken to ensure the CPU’s power-saving features don’t taint any of our results. All of the CPU’s low-power states have been disabled, effectively pegging the frequency at 3.5GHz. Transitioning between power states can affect the performance of storage benchmarks, especially when dealing with short burst transfers.

The test systems’ Windows desktop was set at 1920×1200 at 60Hz. Most of the tests and methods we employed are publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.



Getting the most out of Samsung’s SM951 SSD is a little complicated. First, you’ll need a motherboard with enough PCIe Gen3 lanes running to its M.2 slot—or a full-sized adapter and equivalent expansion slot. You’ll also need to think about cooling. Our testing shows that sustained loads can induce throttling on an open test bench running at normal room temperatures. Adding a fan solved the problem on that rig, but I worry about enclosed systems packed with toasty GPUs, especially since a lot of motherboards put the M.2 slot right below the primary graphics card.

Now, there’s nothing wrong with throttling performance when thermals exceed acceptable thresholds. Plenty of other SSDs and PC hardware employ similar mechanisms to prevent catastrophic meltdown. It’s just a shame the SM951 seems to require dedicated cooling to maintain peak performance over prolonged periods. Thankfully, common desktop workloads are dominated by shorter bursts of storage activity rather than sustained slogging.

Under the right conditions, the SM951 is capable of delivering spectacular performance. Its sequential throughput is especially impressive, with real-world transfer rates rivaling Intel’s much larger 750 Series SSD, and its boot times are the fastest we’ve measured. The drive’s random I/O rates aren’t as competitive at higher queue depths, but they’re still a step up from the old XP941. Heavy random workloads are pretty rare outside of servers and workstations, anyway.

The kicker is that everything is delivered in a tiny M.2 package. If there was a pound-for-pound ranking for solid-state drives, the SM951 would arguably be at the top of the list. The 512GB version’s $400 sticker is pretty reasonable, too, considering the 750 Series provides only 400GB of storage at that price. SATA drives still retain the cost-per-gigabyte crown, but the SM951 is remarkably affordable compared to its PCIe competition.

As much as I’d like to avoid closing with a clichéd car analogy, I can’t help but feel like the SM951 stuffs the engine from a supercar into the body of a hot hatchback. Extra cooling is required to avoid the rev limiter, but this unique combination has undeniable appeal for hardcore enthusiasts, in part because it beats so many rivals off the line. The fact that you can’t find one on the lot of every dealer only adds to the cachet.

Comments closed
    • dark-knight59
    • 7 years ago

    I’m in. skylake NUC is the future?

    • dark-knight59
    • 7 years ago

    super great article….the bigger surprise for me…the crucial BX100 staying right in the middle of the game…. that will be my next SSD

    • technospares
    • 7 years ago

    Future is SSD! Samsung’s SM951 PCIe SSD sounds great. It has replaced hard disk in several storage medium as it supports the systems with higher speeds. Heard about the most recent innovation of [url=<]solid state drives[/url<], the 3D NAND, provides a promising future for SSD.

    • MacGyverSG1
    • 7 years ago

    Is it possible to test the Samsung SM951 AHCI on a Z77 motherboard (obviously using PCI-E adaptor card)? I have the ASUS MAXIMUS V EXTREME Z77 and I’m not sure if the SM951 will boot Windows.

    The Kingston HyperX Predator will work (it has boot OROM built in), but the SM951 is a faster performer.

    • Rageypoo
    • 7 years ago

    I’m so glad to see the use of onboard storage directly on the motherboard. I’ve never been a fan of sata connection and IDE was just…wow so happy we’re past that. Direct board connections make sense, they’re faster, smaller, and perform better. This will help make computers finally come out of the dark ages with unecessary front storage slots and wasted unused space and more room friendly builds. I love efficiency, and power performance from a smaller form factor. Less heat, less frill, less BS really. I’m loving the new stuff, I hope it keeps getting better and better!

    • 7 years ago

    Maybe we can synchronize it with Intel’s-Race to sleep Core M

    • UberGerbil
    • 7 years ago

    But we’ve seen a lot of motherboards that put the M.2 slot behind the x16 slot, beneath the area where the video card sits; there’s no headroom for any kind of functional heat spreader (and with a hot video card directly above, it might serve more as a thermal collectors anyway). As mczak notes, the form factor and connector arrangement is kind of pessimal for dissipating heat: if the edge connector was vertical and used the long edge of the drive (so it looked more like a DIMM) you might have more of a chance to get heat sinks on both sides and it would be more likely to be in any airflow within the case. But of course space for that kind of thing is at a premium, especially in the smaller mobos. We have seen some boards that put the M.2. connector on the underside of the motherboard, which at least gets it away from other hot components but does nothing to get it into an airflow (and, depending on your case, may leave even less room for heat sinks).

    Of course you could put it on an adapter and stick it in a PCIe slot, which essentially negates everything about M.2. As I said when the first PCIe SSDs started appearing: there are a lot of things to like about getting away from SATA, but there are some advantages to sticking your secondary storage on the end of a long flexible cable.

    • Mr Bill
    • 7 years ago

    You just know these are going to start selling with DDR-like heatsinks. I bet someone is clipping some on right now.

    • UberGerbil
    • 7 years ago

    [quote<]Would it have killed them to include it in Disk Management?[/quote<][url=<]It is[/url<]. But like a lot of things in the management snap-ins, it's kind of hidden. [url=<]This[/url<] is from the article Freon linked.

    • derFunkenstein
    • 7 years ago

    works for all of AMD’s top end stuff, right?

    • Visigoth
    • 7 years ago

    Well, when I think enterprise, the only name that comes into my mind is Intel, nothing else. Yes, it’s expensive as hell, but you do get what you pay for.

    • derFunkenstein
    • 7 years ago

    Happy to be wrong. Would it have killed them to include it in Disk Management?

    Storage Spaces is on my PC but I’d never noticed it nor opened it before. That’s much better.

    • derFunkenstein
    • 7 years ago

    edit: nevermind, I was wrong, happily so. Thank you. 🙂

    • chuckula
    • 7 years ago

    I bet that MS spent a crapton of money coming up with wacky names to describe RAID-1, 5, and 0.

    • Deanjo
    • 7 years ago

    Windows has been able to do software raid for quite a while. Freon has the link for Windows seven above but it is also possible in Win8+ with Storage Spaces.

    [url<][/url<] [quote<] Resilient storage. Storage Spaces provides three storage layouts (also known as resiliency types): Mirror. Writes data in a stripe across multiple disks while also writing one or two extra copies of the data. Use the mirror layout for most workloads – it helps protect your data from disk failures and provides great performance, especially when you add some SSDs to your storage pool and use storage tiers. Parity. Writes data in a stripe across physical disks while also writing one or two copies of parity information. Use the parity layout for archival and streaming media workloads, or other workloads where you want to maximize capacity and you’re OK with lower write performance. Simple (no resiliency). Writes data in a stripe across physical disks without any extra copies or parity information. Because the simple layout doesn’t provide any protection from disk failures, use it only when you require the highest performance and capacity and you’re OK with losing or recreating the data if a disk fails. You can also use the simple layout when your application provides its own data protection.[/quote<]

    • Waco
    • 7 years ago

    Especially not the buggy TRIM implementation on their enterprise drives…

    • koaschten
    • 7 years ago

    Anyone else thinking … “Easily solved by watercooling”? 😉

    • Freon
    • 7 years ago


    • derFunkenstein
    • 7 years ago

    On OS X and Linux, but as far as I can find (and I’m open to the idea that I’m wrong) the only software RAID built into windows is JBOD spanning (which is hardly RAID).

    • Deanjo
    • 7 years ago

    See what you did!!!! Way to go !!!! Now Geoff is leaving!!!

    • Deanjo
    • 7 years ago

    [quote<]Striping would require some sort of RAID controller, I think.[/quote<] Nope, you can use a pure software RAID as well (which has its advantages).

    • derFunkenstein
    • 7 years ago

    Striping would require some sort of RAID controller, I think. Addonics made a neat 4-drive RAID card supporting RAID 0, 1, and 10 for mSATA SSDs. A similar card for M.2 form factor would be required, and it’d need a full Gen3 x16 slot to really give you all that bandwidth.

    It’d be crazy fun, that’s for sure.

    • derFunkenstein
    • 7 years ago

    Yeah, I guess that might be the case. Pretty soon they’ll have “heat spreaders” like DIMMs do.

    • Krogoth
    • 7 years ago

    Not going to be happening for a long time.

    NVM Express cards are going to be marketed towards prosumers and enterprise users.

    • Deanjo
    • 7 years ago

    Yes there is but there is also the advantage of knowing the very specific type environment that ssd is placed in and being able to adjust the behaviour of the thermal throttling accordingly with specific firmware.

    • mczak
    • 7 years ago

    Well, blame the form factor. So far I haven’t really seen anything which would indicate with today’s nand technology it’s even possible to build it significantly more energy efficient. Look at the intel 750 series and compare it to this one and guess why one throttles and one does not – it’s definitely not because the 750 would have smaller power consumption (fwiw the 750 400GB is rated at 12W for writes, 9W for reads, the SM951 should be 6.5W max).
    For cooling m.2 actually seems to be a terrible form factor. There’s no clearance for heatsinks whatsoever, even if you’d try to do something custom you probably can only do that on one side. Too bad there’s tons of memory dimms out there with useless heatsinks, but m.2 can’t have any… Granted for things going into notebooks this makes sense, but maybe m.2 for the desktop wasn’t such a good idea after all…
    Maybe it would also be possible to make it more “heat-resistant” (so throttling would only happen at higher temperatures) but I’m sceptical – some other sites have shown throttling to happen at about 80 degrees temperature for the sm951 (measured at controller), which is sort of high already though probably the nand chips shouldn’t get too hot neither.

    • Chrispy_
    • 7 years ago

    Screwing up would be fine if they weren’t trying to sweep the issues under the rug. People screw up all the time, but what separates the good guys from the bad guys is how they deal with the aftermath.

    Samsung definitely aren’t conforming to “good guy” behaviour.

    • rpjkw11
    • 7 years ago

    I have three 840 Pro and three 850 Pro SSDs, so far no problems, but with each new Samsung “revelation” I’m starting to reconsider other SSDs, especially Intel.

    • UnfriendlyFire
    • 7 years ago

    Anyone that wants the M.2 drive should wait for a few months for prices to drop, and more importantly, for any firmware bugs to come out into the light.

    (On a side note, there’s a laptop that uses a M.2 SSD that is significantly slower in terms of writing speed than the last year laptop model’s mSATA SSD.)

    • Deanjo
    • 7 years ago

    Had a bit of rain over the last few weeks. 6/10’s of an inch which wasn’t enough to raise water levels but just enough to get the clay berms and road slick again.

    • InfiniXtive
    • 7 years ago

    Fascinating, as always.

    I have a request for future analyses. Given that many of the end users who would purchase this are doing so for a gaming machine, it would be instructive to see how games perform with the hardware.

    In particular, level load times, installation times, and especially copy times. I keep backups of all my games on a large 7k tertiary drive, but swap out frequently used games to my SSDs, so I do a lot of copying and moving. I wouldn’t be surprised if that is common for gaming machines these days.

    One other thing: can these drives be striped or would the earth melt if we tried that?

    Thank you,


    • derFunkenstein
    • 7 years ago

    Yeah, throttling on write makes a little sense, since it’s actually doing stuff to the flash. Doing it on read is what annoys me the most. By the time Windows is loaded, is it already starting to throttle?

    • Waco
    • 7 years ago

    Throttling on write I could almost accept in a consumer drive. Throttling on READ though? Ugh.

    And here I thought only ioDrive Octals suffered from thermal write throttling…

    • gigafinger
    • 7 years ago

    [quote<]As much as I'd like to avoid closing with a clichéd car analogy, I can't help but feel like the SM951 stuffs the engine from a supercar into the body of a hot hatchback. Extra cooling is required to avoid the rev limiter, but this unique combination has undeniable appeal for hardcore enthusiasts, in part because it beats so many rivals off the line. The fact that you can't find one on the lot of every dealer only adds to the cachet.[/quote<] Replace hatchback, with rear engine, and you've described a Porsche.

    • Dissonance
    • 7 years ago

    Doh. He’ll always be the fresh prince to me.

    • anotherengineer
    • 7 years ago

    Don’t be a hater Deanjo 😉

    In other news, hows the weather there, still dry or floods return?

    • UberGerbil
    • 7 years ago

    I can’t be the only reader to notice Bane-Leia, but maybe I’m the only one disturbed by it? “Aren’t you a little short for the darkness?”

    • Deanjo
    • 7 years ago


    • anotherengineer
    • 7 years ago

    Agreed, but there is a difference between some throttling and epic throttling though.

    • leor
    • 7 years ago

    “new hotness” is in fact from Men In Black, not Fresh Prince. New hotness as opposed to Tommy Lee Jones’ “old and busted.” Right actor, wrong role 😉

    • anotherengineer
    • 7 years ago

    Geoff try to get one of these to review


    • Deanjo
    • 7 years ago

    I wouldn’t be so sure of that. As mentioned in the article, some throttling was observed with the previous gen as well in their labs. But I can tell you from first hand experience using my 2014 MBP with the 1 TB Apple version of that older gen, no throttling has occurred on it (and I’ve hammered that thing hard with long sequential writes with 100+ Gig video file writes over extended periods).

    • albundy
    • 7 years ago

    Sounds like M.2 drives will need heatsinks in the very near future to keep their speeds consistent. Great review!

    • Jeff Kampman
    • 7 years ago

    Whoops, fixed.

    • anotherengineer
    • 7 years ago

    If it throttles like that on an open bench motherboard, I would imagine the throttling would be even worse in a macbook, and notebook!!

    • Convert
    • 7 years ago

    Minor typo on page 6: “Here’s are the essential details for all the drives we tested:”

    Thanks for the review! Not impressed with the throttling though.

    • willmore
    • 7 years ago

    Sounds like those installing it might want to have a bunch of silpad and some aluminum sheet to bend into a nice heatsink–expecially in a laptop. Maybe a heatpipe and a heatsink from an old laptop or something.

    • derFunkenstein
    • 7 years ago

    That might actually help, but I wonder what the retailers that have to process the warranty would think.

    • DPete27
    • 7 years ago

    Just add [url=<]some bling[/url<] and you're set!

    • anotherengineer
    • 7 years ago

    *Samsung Disclaimer
    * To get full rated speeds active cooling is required
    *fan sold separately

    • Spyder22446688
    • 7 years ago

    Thank you for the review.

    I can’t wait for the day when NVM Express models are actually marketed towards the PC enthusiast market. I’d love to be able to purchase one for installation in a Skylake motherboard. Not sure if there is space on a motherboard, but be great to also see a model with low-profile copper heatsinks attached. I am sure that would go a great length towards cooling.

    • derFunkenstein
    • 7 years ago

    I don’t know about the latest 2015 models, but previous Apple drives had weird Apple form-factors and [url=<]required specific adapter boards[/url<] you bought on eBay and had shipped in from China in order to use in anything other than Apple hardware (or even older hardware, like the Nehalem Mac Pros). If that has continued, it might explain the Apple exclusivity.

    • derFunkenstein
    • 7 years ago

    Wow, that throttling is just unacceptable. Yes it’s faster than SATA3, but if I’m paying such a huge $/GB premium, I expect it to run full tilt. I’m sure that since it’s so fast, it’s fine in bursts that are typical of desktop or notebook SSDs. But if it’s working that hard, it’s working that hard for a reason, and it needs to work that hard without throttling.

    • ronch
    • 7 years ago

    After Samsung’s string of SSD buggies? No thanks.

    • Deanjo
    • 7 years ago

    No 1tb version for the PC? Guess it’s an Apple exclusive.

    • Firestarter
    • 7 years ago

    TBH the AHCI version doesn’t do it for me, I want to see the numbers on that NVMe beast!

Pin It on Pinterest

Share This

Share this post with your friends!