Review Index:

PCIe SSD Roundup - Samsung SM951 NVMe vs. AHCI, XP941, SSD 750 and More!

Subject: Storage
Manufacturer: Samsung

Introduction, Specifications and Packaging


There's been a lot of recent talk about the Samsung SM951 M.2 PCIe SSD. It was supposed to launch as an NVMe product, but ended up coming out in AHCI form. We can only assume that Samsung chose to hold back on their NVMe-capable iteration because many devices are unable to boot fron an NVMe SSD. Sitting back for a few months was a wise choice in this case, as an NVMe-only version would limit the OEM products that could equip it. That new variant did finally end up launching, and we have rounded it and the other Samsung M.2 PCIe SSDs up for some much awaited testing:

View Full Size

I'll be comparing the three above units against some other PCIe SSDs, including the Intel SSD 750, Kingston HyperX Predator, G.Skill Phoenix Blade, Plextor M6e Black, and more!

Continue reading our review of these hot new M.2 products!


These are all OEM products, but here are some specs from various sources:

XP941 (source):

  • Form Factor: M.2 2280
  • Capacity (GB): 128, 256, 512
  • Host Interface: PCI-Express 2.0 x4
  • MTBF: 1.5 Million Hours
  • Uncorrectable Bit Error Rate (UBER): < 1 sector per 1015 bits read
  • Power Consumption (Active/Idle): 5.8W / 80mW
  • Peak Read Sequential Performance: Up to 1170 MB/s
  • Peak Write Sequential Performance: Up to 930 MB/s
  • Peak Random Performance Reads: Up to 122k IOPS
  • Peak Random Performance Writes: Up to 72k IOPS
  • Physical Dimesions: 22 x 80 x 4 mm
  • Weight: 8.5g

SM951 (AHCI) (source):

  • Form Factor: M.2 2280
  • Capacity (GB): 128, 256, 512
  • Host Interface: PCI-Express 3.0 x4
  • Controller: Samsung UBX 3-Core
  • Flash: Samsung 16nm MLC
  • Power Consumption (Active/Idle): 6.5W / 50mW
  • Peak Read Sequential Performance: 2050, 2150, 2150 MB/s
  • Peak Write Sequential Performance 600, 1200, 1500 MB/s
  • Peak Random Performance Reads: 90k IOPS
  • Peak Random Performance Writes: 70k IOPS
  • Physical Dimesions: 22 x 80 x 4 mm

SM951 (NVMe) (source):

  • Form Factor: M.2 2280
  • Capacity (GB): 256, 512
  • Host Interface: PCI-Express 3.0 x4
  • Flash: Samsung 16nm MLC (verified visually)
  • Peak Read Sequential Performance: 1.08, 1.17 GB/s *
  • Peak Write Sequential Performance 800, 931 MB/s *
  • Peak Random Performance Reads: 120k, 122k IOPS
  • Peak Random Performance Writes: 60k, 72k IOPS
  • Physical Dimesions: 22 x 80 x 4 mm

* We know the NVMe SM951 has higher specs than what was stated in our source, but we don't have more accurate figures from Samsung.


Our NVMe SM951 was 'packaged' with another product we got in for testing.

The AHCI SM951 came in a 2015 Lenovo X1 Carbon that Ryan has been testing. I liberated it while he wasn't looking (shhh).

The XP941 came with another product, but we have a brown box pic for that one:

View Full Size

Video News

April 17, 2015 | 08:14 AM - Posted by willmore (not verified)

On the second page "Internals, Testing Methodology and System Setup", the caption for the second photo is wrong. It says the NVMe SM951 is both the middle and the bottom card. If it like the original photo, the bottom one should be the AHCI SM951.

April 17, 2015 | 10:56 AM - Posted by Allyn Malventano

Thanks! Fixed.

April 17, 2015 | 08:23 AM - Posted by jerrytsao (not verified)

Nice review! Should wait for 512G NVMe instead of 750 series.

April 17, 2015 | 09:45 AM - Posted by Buyers

I know you said that you didn't see any kind of throttling with the temps you saw, but at ~156F (~69C) degrees do you feel as though future products will/should start coming with small heat spreaders just to mitigate the possibility of problems due to thermals?

April 17, 2015 | 11:05 AM - Posted by Allyn Malventano

Heat spreaders wouldn't fit in with the M.2 physical spec (might not fit into some notebooks, etc). If I was installing one into a desktop, I'd probably add one or two stick-on VRAM-style heat sinks though.

April 17, 2015 | 02:29 PM - Posted by willmore (not verified)

I wonder if we'll see thermal issues taken into concern with laptop m.2 bays. Since you, uhh, liberated one of these from a laptop, could you comment on any such considerations it may have had?

Secondly, any chance you could do some of this testing in the laptop to see if there is any performance delta?

April 17, 2015 | 04:46 PM - Posted by Buyers

Is temperature testing something you'll consider adding to the process for future drives? Presumably these will be fitting into notebooks and such with tight spacing, so perhaps you could 'insulate' one somehow to increase temps to see if/when they do start to show issues.

April 20, 2015 | 12:07 PM - Posted by Allyn Malventano

Benchmarking these puts way more load on them when compared to what they would see in a laptop. I have yet to come across one that seemed so 'sensitive' that you would hit a throttling limit in typical laptop use. You'd have to have some way to write to these at full speed for over a minute to get them to start building up considerable heat, and you're just not going to have anywhere to get that data *from* so quickly when on a laptop. Honestly even on a desktop I have to run them 100% to get them to warm up. They blow through the IO's so quickly that they would be idle most of the time in typical usage.

April 17, 2015 | 09:49 AM - Posted by citrix13

I am interested in Allyn adding more real world tests to the SSD reviews, perhaps ISO file creation some game load tests, examples here:

and here:,4096-6.html

April 17, 2015 | 09:54 AM - Posted by citrix13

Some more examples of Real World tests in SSD reviews Allyn here:

and again here:

April 17, 2015 | 11:01 AM - Posted by Allyn Malventano

You're pointing to a lot of file copy / creation types of tests. Not only do we perform similar tests, our file creation and copy tests are more thorough than those in your linked reviews. We are creating and copying four different sizes of files, repeating that process 3x, and averaging those results for each capacity run. Further, the way we perform our test reveals TRIM speed and other performance inconsistencies that many other review sites completely missed, like this glaring issue with the Vector 180.

April 18, 2015 | 02:38 PM - Posted by citrix13

Allyn, i was not pointing out how thorough, or meticulous the file creation and copy tests are at pcper, merely that there are no Real World Benchmark specifically how fast a given SSD will load a level in Battlefield 4 or application like Photoshop, Word, Powerpoint etc. In other words, at pcper the audience catered to is primarily made up of gamers and enthusiasts, however the ordinary guy/non enthusiast who goes to buy an SSD just wants to know how fast is my Call of Duty, Battlefield, Dying Light or GTA V gonna load. He may not necessarily fully understand, 10 files @ 1000 MB each, 100 files @ 100 MB each, 500 files @ 10 MB each and 1000 files at 1 MB each, as it's laid out here at pcper's file copy / creation test.
I'm not knocking how meticulous your test are of course. Just helping you to think like an ordinary consumer, and i'm asking that you include Real World examples in the benchmarks, for the ordinary guys, non enthusiasts.

April 20, 2015 | 12:10 PM - Posted by Allyn Malventano

Understood. I have actually still been running and recording various PCMark test runs of all SSDs tested. We stopped publishing those results because it got to the point where there was little to no difference across various SATA SSDs. PCIe speeds did bump these results up somewhat, so I may reintroduce those results in some future pieces.

April 20, 2015 | 03:06 PM - Posted by citrix13

Yes i support a return of PCMark 7 and PCMark 8 to the benchmarks.

April 17, 2015 | 11:42 AM - Posted by Brian Hoyt (not verified)

Can you mention what the NVMe version was "packaged" with? Macboook by chance? Since the AHCI one has Lenovo FRU it should be able to be ordered by any place that can order repair parts, won't be cheap though.

April 21, 2015 | 04:49 PM - Posted by vailr

May have been sourced from the Intel NUC5i7RYH

August 28, 2015 | 12:03 PM - Posted by NUCi (not verified)

the link indicates that NUC5i7RYH was tested with

Samsung SM951 MZVPV256 256 GB M.2 NVMe SSD version,
but that is not available for me. Instead I have tried
Samsung SM951 MZHPV256HDGL AHCI version.

I have in NUC5i7RYH the latest bios .350

USB installation with Ubuntu 14.04 64bit goes fine and 951 AHCI version MZHPV is found by installer. Before the installation BIOS finds the HD ok with correct identifier.

Ubuntu installation goes lightning fast and ok until end, BUT after the restart comes visible on screen:
and then It stops with blinking "-" at top left corner
and after a long while appears text:
"Reboot and Select proper Boot device
or Insert Boot Media in selected Boot device and press a key"

BIOS is in AHCI mode without UEFI option. Even with the option (x) it stops in shell mode.

Any ideas, what to try next.

September 28, 2015 | 09:49 PM - Posted by Matt (not verified)

I thought NVMe support was in the linux kernel since 3.3? ( I don't have one at the moment to test, but may have an SM951-nvme in a couple of weeks. If I do I'll update here.

April 17, 2015 | 12:14 PM - Posted by homerdog (not verified)

In the event that my Z77 mobo gets boot support for NVME drives and I put a 750 PCIe drive in it, how will the PCIe lines get divided between my GPU and the SSD? Will the SSD connect to the chipset rather than the PCIe lanes on the CPU?

April 18, 2015 | 05:22 AM - Posted by Anonymous (not verified)

That depends on the board. Does it have an x4 available from the chipset? If you have an SLI board and you plug it into an x16 slot connected directly to the cpu, you will lose some connectivity to the gpu on most boards. Both slots would go down to x8, but the Intel part is only x4, so 4 lanes would be wasted. I was looking at mother boards for a friend, and I was recommending a Z97 board with 2 x16 slots (usually x8/x8 when multiple cards are used) to support dual video cards and an x4 (in an x8 or x16 slot) off the chipset to support a pci-e ssd if they wanted one later. Some of them end up sharing bandwidth between the x4 slot , the m.2 slot, and/or other x1 pci-e slots. You can't use all of them at the same time.

November 17, 2015 | 11:25 AM - Posted by Anonymous (not verified)

I don't know if you have bought any PCIe SSD but I also have a mobo with Z77 chipset. The Z77 chipset support 1x16 lane PCIe 3.0 and 2x8 lanes PCIe 3.0. I've checked my mobo manual (Asus P8Z77-V LX) and have only two PCIe x16 slots. One slot is for PCIe 3.0/2.0 operating at x16 and the other one is for PCIe 2.0 operating at x4. This means that in my case I can have the graphic board working at x16 and the PCIe SSD at x4

April 17, 2015 | 01:13 PM - Posted by Snoopvelo (not verified)

512GB model available at MacMall for $403.99

April 17, 2015 | 01:15 PM - Posted by Snoopvelo (not verified)


April 17, 2015 | 01:53 PM - Posted by VintageDon (not verified)

Good find on that one; Amazon does have them, but they're $500 for the 512GB version. I've got an ASRock X99 Extreme4 with the 512GB ACHI version and love it, but I'm going to go ahead and update to the NVME version.
Also, thanks for the review on this. Great stuff.

April 17, 2015 | 02:14 PM - Posted by Allyn Malventano

The part number listed in that link is *not* the NVMe version. Buyer beware.

April 18, 2015 | 03:58 PM - Posted by Anonymous (not verified)


April 17, 2015 | 06:21 PM - Posted by RamCity

This is an interesting review. Kudos for taking the initiative on buying the hardware to get hold of an NVMe SM951. There are a few things to note here:

1. The Lenovo specific SM951 has a thermal limiter of some sort that prevents it from achieving top sequential reads/writes. The 'standard' OEM version of the SM951 has no such limiter and is considerably quicker. There will be another review online about that next week.

2. The comment regarding 'All X99 and Z97 motherboards should have no problem booting an NVMe SSD' is more or less true, but whether maximum speeds can be achieved is another story entirely.

Many Z97 systems, especially those from Gigabyte and MSI, have M.2 sockets that are limited to just 2x PCIe 2.0 lanes, so that maximum sequential speeds top out at around 750MB/s. In addition, there are some X99 motherboards (even rather high end ones like the MSI X99S SLI PLUS) that require a 40-lane CPU to allow 4x PCIe lanes to be delivered to the M.2 socket. If you bought a 5820K CPU with 28 lanes, then you'll once again be restricted to just 2 lanes on the M.2 socket, and severely limiting the performance of any PCIe 4x M.2 SSD. You could work around it by using an adapter to connect via a standard PCIe socket, if you have one spare, but to date we've seen very few Z97 motherboards that can boot windows from an adapter.

The other thing to note is that installing windows on a PCIe M.2 SSD is not always a straight forward affair, due to their inherent lack of a built-in option ROM (with the exception of the old Plextor M6e and the new Kingston HyperX Predator, which do have option ROM's).

Most systems will at least require a BIOS update (especially if installing to the SM951, since it's is so new). Installing Windows 8.1 is relatively simple so long as you install via a USB drive, or a USB connected DVD-ROM. Windows 7 typically requires a bit of extra work, but it's doable with a bit of help from forums or online videos.

cheers, Rod

April 17, 2015 | 06:47 PM - Posted by Allyn Malventano

1. The Lenovo part (the AHCI SM951) had no issues reaching top rated speeds in our testing. The NVMe part (with lower write speeds) was not sourced from Lenovo, bit it's possible it may have had software-limited speeds in the interest of reduced heat production. With zero airflow, we did see thermal throttling kick in on some of these, but that only occurs intermittently and is not really a hard limit on speeds.

2. Good point on the number of available lanes. Buyers will have to do their research prior to purchase. I will say that if the motherboard is using a PCIe 3.0 capable CPU, and if those lanes are routed to the M.2 slot, it should run at 3.0 speeds, meaning that with only two lanes you would cap out at 1.5 GB/sec with PCIe 3.0 x2.

More data is available in the SSD 750 compatibility testing that Ken worked on last week.

April 18, 2015 | 03:49 AM - Posted by D00d (not verified)

@ Allyn -- Jon L. Jacobi states here, "The real issue with M.2 is that on Intel systems it's "generally implemented behind the PCH (Platform Controller Hub), which features only Gen 2 PCIe. That's because the PCH lies behind the DMI (Direct Media Interface) which is capped at 2GBps."

Does that seem in line with your info?

April 20, 2015 | 12:10 PM - Posted by Allyn Malventano

Yup, that's very true, but not all motherboards drive their M.2 port from the PCH.

April 18, 2015 | 04:36 AM - Posted by D00d (not verified)

Say, Rod -- What has been your experience regarding successful booting of Windows from a Samsung M.2 SSD on a PCIe adaptor, on an A88X motherboard with a Kaveri APU (specifically, A10-7850K on a G1.Sniper A88X)? I have yet to see anyone test M.2 PCIe SSDs on AMD APU-based platforms, and this crucial information is sorely lacking, since I don't want to spend my dosh on a drive which will be extremely problematic in said regard. Thanks!

April 18, 2015 | 06:40 AM - Posted by RamCity

It occurred to me after I read the pcper Intel 750 review that we very seldom encounter customers wanting to install PCIe M.2 SSD's AMD platforms, or at least we've never fielded a question from a customer wanting to. For some reason, most of them are using Intel.

What I do know is that this time last year there were virtually no motherboard manufacturers with the necessary boot support built-in to the BIOS so that the XP941 could boot (the SM951's predecessor). ASRock were the very early adopters followed soon after by ASUS, so I would think that any ASUS board, released in the last 12 months, be it Intel or AMD based, will at worst require a BIOS update to support the XP941, and possibly just a bit of user pestering to get them to add support for the SM951.

So, if I were you I'd contact ASUS to get their official stance on XP941/SM951 support and see what they have to say.

cheers, Rod

April 18, 2015 | 05:42 AM - Posted by Anonymous (not verified)

Thank you for supplying additional info. From what I have read about the m.2 specification, with all of the possible different types of keying, this does not seem to really be a consumer friendly specification. I had assumed that the m.2 slots would mostly stay in oem produced machines with little direct interaction with consumers, but since we have them on motherboards that is not how it is going to work. I would say a full pci-e card is better if you have the space and an available slot. M.2 obviously makes sense in SFF systems, although performance may be limited by heat. The sata-express connector seems like it is going to be DOA. It is a large connector that looks like a kludge, and I believe it only delivers pci-e x2.

For most consumers, unless they want to be on the bleeding edge, they should wait and let the market settle a bit. For most consumer applications, I don't think these new drives are work the extra cost. They are obviously faster, but this is of limited usefulness given the current ecosystem. I would probably still recommend buying a Samsung 850 evo, unless they have some specific use case for new NVMe parts.

July 1, 2015 | 03:56 AM - Posted by Anonymous (not verified)

This is how it is, I have been working in IT as administrator for years, so I consider my abilities in knowing my way when building a new system to be as good
I Got the SM951, 256GB and the 512GB
one as NVMe and the other as AHCI
I got Asus deluxe X99 (comes with M.2 4x by the memory sides, and an PCIe card 4x)
Latest BIOS
no matter how I try, the SM951 is not bootable
I tried Windows 7 ultimate 64bit
Windows 7 Pro 64bit
Windows 8 64bit
Windows 8.1 64bit
Ubuntu 15 64bit
Same Same
PS: I managed to install with a few tweaks windows 7, ended up with extremely lagging speed
for additional 30 USD I bought an adapter M.2 to 2.5" S-ATA
and it worked with no issues only at lower speed (550 read/550 write)
hoping that soon there will be a newer version of BIOS that at least shows the SM951 in BIOS (it never showed it so far)
So my advice to every one is
If you are planning to buy the SM951 and expecting it to work out of the box just like that, Don't even think of it
If you are willing to invest long hours in trying it, Good luck

I checked on almost all available videos on youtube, there was one advising to shut down the CSM under first boot to install, the system BIOS insisted (yup, insisted) on reactivating it

this SM951 is new, installing windows on another SSD and using SM951 as D: was ok, the idea was to use it as booting device not as slave device,

I hope this would be helpful for those planing to buy it

Kind Regard

January 15, 2016 | 12:00 AM - Posted by dovedescent7

thanks for the fantastic post, I am one of the unfortunates who apparently didn't do my research and purchased an extreme4 and 5820k with the hopes of saving my pennies for the 750 or 950 pro to run at Full speeds. I just opened the box today from Newegg. I'm happy so far.i signed up here specifically to respond to your post, and I'm wondering now if full nvme speeds still aren't able to be achieved via m.2 because of the lanes? I really want the full benefits of nvme, I don't mind going pcie because this is all workstation,very cpu heavy stuff,no graphics.

I am still within the return window at Newegg, think I should return the extreme 4 and cpu and go with a 6700k setup?

January 15, 2016 | 12:01 AM - Posted by dovedescent7

Rod, thanks for the fantastic post, I am one of the unfortunates who apparently didn't do my research and purchased an extreme4 and 5820k with the hopes of saving my pennies for the 750 or 950 pro to run at Full speeds. I just opened the box today from Newegg. I'm happy so far.i signed up here specifically to respond to your post, and I'm wondering now if full nvme speeds still aren't able to be achieved via m.2 because of the lanes? I really want the full benefits of nvme, I don't mind going pcie because this is all workstation,very cpu heavy stuff,no graphics.

I am still within the return window at Newegg, think I should return the extreme 4 and cpu and go with a 6700k setup?

April 17, 2015 | 08:08 PM - Posted by Cyclops

So how much did you say it was? Oh wait a minute...

April 18, 2015 | 03:33 AM - Posted by Anonymous (not verified)

How fat is the boot up speed?

April 18, 2015 | 03:33 AM - Posted by Anonymous (not verified)

How fast is the boot up, sorry for the typo.

April 18, 2015 | 03:52 AM - Posted by Jordi (not verified)

Please, I don,t like Windows 8.1
I have Windows 7, I will buy a Asrock Z97 Extreme 6 with M.2 Ultimate ( Gen 3 compatible)
I need SM951 for OS boot.
Can I boot the NVMe version with windows 7 or i need to buy the "old" AHCI ?
Thanks from Barcelona, Spain

April 19, 2015 | 06:47 PM - Posted by Anonymous (not verified)

It´s complicated, but it can be done.

April 22, 2015 | 09:17 PM - Posted by Allyn Malventano

I'm betting that reviewer didn't format his Windows 7 USB installer for UEFI boot, resulting in his 'complicated' method. Ken had no issue installed Windows 7 to the SSD 750 with the Intel driver. You just have to format the installer drive properly. I recommend Rufus.

April 18, 2015 | 04:53 AM - Posted by Anonymous (not verified)

UBER = "< 1 sector per 1015 bits read"

Looks like the superscript didn't copy from the pdf, or isn't displaying for me. It is supposed to be 10 to the 15 power.

April 19, 2015 | 03:00 AM - Posted by KrypteX (not verified)

@Allyn Have you noticed the Thermal Throttling on these Samsung M.2 PCIe SSDs, as seen in this review ?

April 19, 2015 | 10:19 AM - Posted by Patrick Reynolds (not verified)

Would like to see CPU usage comparison between ahci and nvme versions. Right now for is compatibility the ahci version looks to be the better solution.

April 20, 2015 | 12:26 PM - Posted by Allyn Malventano

CPU usage is not apples to apples since the NVMe parts are reaching far higher IOPS figures than their AHCI counterparts.

April 19, 2015 | 11:18 AM - Posted by Anonymous (not verified)

Wich storage driver did you use Allyn? Was it the Intel RST one or the inbox Nvme driver provided by Mirosoft?

April 20, 2015 | 12:24 PM - Posted by Allyn Malventano

Microsoft Windows 8.1 'inbox' driver (native).

April 19, 2015 | 06:55 PM - Posted by Anonymous (not verified)

Both SM951's are PCIE 3.0 x 4 not PCIE 2.0 x 4 as you have on the first page.

April 22, 2015 | 02:36 PM - Posted by Allyn Malventano

Thanks for the catch - fixed.

April 19, 2015 | 11:23 PM - Posted by Matman (not verified)

To me it looks like AHCI is the better performance pick for desktop workloads. I'm wonder if NVMe is inherently less efficient at low queue depths or we're just seeing the result of immature OS/driver/firmware support.

April 20, 2015 | 12:19 PM - Posted by Allyn Malventano

That particular delta you saw (in Iometer I assume) was between the AHCI and NVMe 951's. The AHCI version had higher write speeds, which accounted for it getting a slight early jump in results at lower queue depths. This is not as much an issue with NVMe as it was a relatively large delta in performance specs between the two samples. It wasn't as apples to apples as I would have liked, but it was all we had on hand to test.

April 20, 2015 | 09:03 AM - Posted by vailr

Can you provide any details about backwards compatibility? If someone were to purchase the NVMe version SM951, will it still function in an older system configuration? Maybe use it now with an existing AHCI-only system, and then plan on re-using 6 months from now in a Skylake Z107 system, as an NVMe drive?
Also: adapter cards. Ram City offers the "Lycom M.2 PCIe SSD to PCIe 3.0 x4 adapter"
Are both the AHCI & the NVMe versions of SM951 going to work with that?

Edit: does Windows 8.1 include an NVMe driver? What does Windows Device Manager show, as far as driver version and date?

April 20, 2015 | 12:22 PM - Posted by Allyn Malventano

8.1 does include an NVMe driver (it's what we used for testing - Samsung does not have a driver). The trick on older system is not recognition under Windows 8.1. The problem is the BIOS not enumerating the device for boot (the part *before* entering Windows protected mode). With that part missing from older system BIOS, you would only be able to use an NVMe device as a secondary drive.

April 26, 2015 | 04:13 AM - Posted by Dark_wizzie (not verified)

Hey Allyn,

I approach most of this from a gamer's perspective. Of course, I wouldn't expect night and day difference with a very high end SSD vs a normal SSD. But what about for something like running around Skyrim with a butt-ton of mods? Open world games, tons of things to load on the fly.

It could still be CPU bottlenecked though. Tom's Hardware did an article a long time ago about SSD load in gaming. The guy used some sort of trace-based analysis tool from Intel to check if the reads from the SSD during game startup, level loading, and playtime are sequential or random, what size, and what queue depth. It's very interesting and I think many gamers would like to see such an article.

I'm looking at all the graphs and frankly it doesn't mean much to me. I don't run file servers, I load a ton of maps.


April 30, 2015 | 04:42 PM - Posted by jimv (not verified)

Great review. I was hoping you could help out by comparing my workflow to which above benchmark best applies to me.
My apps use up to 29GBs of RAM where 1000s of 64k buffers are used as targets for various streams of audio stored on SSDs.

When I press a key on an 88 note keyboard/synth it goes 1st to the 64k buffer in RAM then a stream of audio follows.
Obviuosly random applies to the 64k RAM buffers and read to the streaming audio files.
Maybe the Workstation benchmark....?

Thanks again for a great source of comparisons on SSDs.

May 3, 2015 | 12:30 PM - Posted by Anonymous (not verified)

The X99 Sabertooth allow one to conceal their M.2 SSD completely under the Thermal Armor. Is that recommended given the heat output for SM951 NVMe?

May 8, 2015 | 03:57 AM - Posted by Anonymous (not verified)


May 18, 2015 | 09:07 AM - Posted by Anonymous (not verified)

Have you heard about a 1TB version of the SM951 being release soon?

June 20, 2015 | 11:45 PM - Posted by Brighttail (not verified)

I'm using the SM951 ACHI in a m.2 to PCI x4 card that has a heatsink on it and is plugged into a x4 PCI-E slot on my x99 motherboard. I love it but I'm wondering if I managed to get my hands on one of the new NVME, would it fit into the same heatsink slot (the pins looks the same) or would I have to use the M.2 slot on the motherboard? I'm assuming either would work and just change to NVME in the BIOS.

Second, I have two Samsung 850 pros running in RAID 0 as my applications drive. Would I still be able to keep this in RAID while using NVME on the Asus x99 motherboard?

July 22, 2015 | 10:55 AM - Posted by Christiopher Caruk (not verified)

Hello, I've just been comparing a 512GB 951 NVMe variant that I purchased yesterday with an existing 512GB 951 AHCI. Apparently it's a sample rather than a production unit but I'm seeing fantastic read speeds but horrific write speeds. In my case I'm using with an Asus Z97i-plus with the latest BIOS. The board identifies the 951 and allows me to install windows (8.1 all latest updates)... so far so good. Unfortunately when I run speed tests against the NVMe variant I get 10 times slower write speeds compared to the AHCI 951.

CrystalDiskMark: AHCI variant (connected to PCIe 3.0 bus)

Seq Q32T1 - 1172MB/s read | 1043MB/s write
4k Q32T1 - 398MB/s read | 289MB/s write
Seq - 1052MB/s read | 900MB/s write
4k - 35MB/s read | 128MB/s write

CrystalDiskMark: NVMe variant (connected to PCIe 3.0 bus)

Seq Q32T1 - 2264MB/s read | 501MB/s write
4k Q32T1 - 563 MB/s read | 21 MB/s write
Seq - 1299 MB/s read | 170 MB/s write
4k - 54 MB/s read | 0.98 MB/s write

Blindingly fast read but horrifically slow write speeds.

I've also tested using the Z97i-plus's M.2 slot. I see reduced read speeds due to the limited, 10Gbps, speed of the M.2 on this board but the same horrific write speeds.

Is there something that I might be doing wrong? Could this be a BIOS problem? A Windows NVMe driver problem?

July 22, 2015 | 12:56 PM - Posted by Allyn Malventano

That's odd, but I believe Kristian from Anandtech had a similar issue with one of his samples. It was an actual defect I believe and they had to swap out his sample, IIRC.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

This question is for testing whether you are a human visitor and to prevent automated spam submissions.