NVMe SSD very slow write speeds.

[Chuck Foltz]

I noticed the same thing with my 512 SSD. By chance do you have Bit Locker enabled? When you choose I own my surface and login with a MS account it will by default enable bit locker.

 

[John M Beauchemin]

I wonder if they chose this drive for low power consumption? They obviously did not choose it for performance, so either they went cheap for the price, or they chose it based on some other metric, like power consumption.

 

[Daniel Rubino]

I wonder if they chose this drive for low power consumption? They obviously did not choose it for performance, so either they went cheap for the price, or they chose it based on some other metric, like power consumption.

Not really sure. Look at my Surface Pro 4s I have two variations:

Core i5 at 256 GB with a Toshiba SSD (build 10586)
Core m3 at 128GB with a Samsung SSD (build 10240)

Toshiba wins for sequential and 512k, but loses in 4k and 4k QD32. Anyone make sense of that?

 

[ikjadoon]

Err, a few things: your CrystalDiskMark, I just realized, is a little old: 3.0.4. Everyone else is using 5.0.2 (actually, one guy's even using 5.0.3).

But, that withstanding, you actually have the slowest 4K speeds I've seen on a Toshiba drive (that 4K is like spindle disk level, man--spindle disks get around 1MB/s of 4K). Looks like it's on the new build: was Search Indexing completed/paused when you ran that test? Any other programs in the background?

Compare all these Toshiba results with yours:

https://i.imgur.com/oIzNu7R.jpg
http://images.anandtech.com/doci/9727/CrystalDiskMark.png
https://i.imgur.com/UcbSoPd.png

This Toshiba guy thinks Search Indexing was still going on:

https://i.imgur.com/tf9MLQI.png

These are all the same SSD: the 256GB Toshiba XG3. We shouldn't be getting such different results--the issues are Windows updates, CrystalDiskMark versions, firmware updates, search indexing, background programs, local vs Microsoft account (Bitlocker), etc.

So, I would be curious, if (when controlled for all these variables) your drive ever matched what other users were getting, but--I think overall--we may need to wait for a "final" verdict on this when the Anandtech review lands. I'm a little too tired to go back to all those people and ask them all these questions, in the end, haha. But, if you want those blazing results from CDM, I'd take a look at these extra variables.

 

[ikjadoon]

I noticed the same thing with my 512 SSD. By chance do you have Bit Locker enabled? When you choose I own my surface and login with a MS account it will by default enable bit locker.

That's a great point. I was surprised why so many people had turned it on (I thought it was more of an enterprise feature), but the Microsoft account bit seems to be the "culprit".

Bitlocker is an important variable, but from the benches I've seen, it mostly affects 4K speeds and not so much sequential speeds.

 

[ttsoldier]

That's a great point. I was surprised why so many people had turned it on (I thought it was more of an enterprise feature), but the Microsoft account bit seems to be the "culprit".

Bitlocker is an important variable, but from the benches I've seen, it mostly affects 4K speeds and not so much sequential speeds.

Are you saying that logging in with a Microsoft account enables bitlocker?

I'm using a local account right now and it's off.

When I sign in with my ms account, I don't want it on...

Bitlocker IS more geared towards enterprise users...

 

[ikjadoon]

Err, I don't know; that guy who posted earlier said it. I don't have a Surface Book yet and my current laptop doesn't have a TPM chip. And this guy (@ 7:29):

https://youtu.be/PuJK-01X_V0?t=451

Gives a tip that you should turn it off. So....presumably....it was turned on by default?

EDIT:

OK, yup, Windows 10 automatically enables Bitlocker (started with Windows 8.1) if your device has the necessary hardware. You have to opt-out.

Before Windows 8.1 automatically enables Device Encryption, the following must be true:

• The Windows device ?must support connected standby and meet the Windows Hardware Certification Kit (HCK) requirements for TPM and SecureBoot on ConnectedStandby systems.? Older Windows PCs won?t support this feature, while new Windows 8.1 devices you pick up will have this feature enabled by default.
• When Windows 8.1 installs cleanly and the computer is prepared, device encryption is ?initialized? on the system drive and other internal drives. Windows uses a clear key at this point, which is removed later when the recovery key is successfully backed up.
• The PC?s user must log in with a Microsoft account with administrator privileges or join the PC to a domain. If a Microsoft account is used, a recovery key will be backed up to Microsoft?s servers and encryption will be enabled. If a domain account is used, a recovery key will be backed up to Active Directory Domain Services and encryption will be enabled.

 

[lancorp]

Are you saying that logging in with a Microsoft account enables bitlocker?

I'm using a local account right now and it's off.

When I sign in with my ms account, I don't want it on...

Bitlocker IS more geared towards enterprise users...

Why do you say that? You think everyone doesn't have important personal information on their computers? Anyone with a super thin and light device like a Surface Pro is more susceptible to it being stolen. I'd sure sleep better at night knowing that if my SP3 or SP4 fell into the wrong hands, all my email, contacts and stored website passwords were safe from prying eyes and hands!

 

[ttsoldier]

Why do you say that? You think everyone doesn't have important personal information on their computers? Anyone with a super thin and light device like a Surface Pro is more susceptible to it being stolen. I'd sure sleep better at night knowing that if my SP3 or SP4 fell into the wrong hands, all my email, contacts and stored website passwords were safe from prying eyes and hands!

Well. To each their own

 

[lancorp]

Well. To each their own

I agree with that more than your first assertion.

 

[Daniel Rubino]

Although I prefer Bitlocker no doubt turning it off can have significant improvements, at least in benchmarks (not sure about 'real world' feel just yet)

 

[Yonic Boom]

I'm not sure I'd be comfortable leaving my data unencrypted, but that is a pretty big jump.

 

[ttsoldier]

Here's mine

i5/8gb/512ssd

View attachment 115758

Mine was done with bit locker off. This was before I logged into my sp4 with my microsoft account.

 

[ikjadoon]

Mine was done with bit locker off. This was before I logged into my sp4 with my microsoft account.

Exactly! That's what irks me really bad. A 512GB model getting half the sequential write speed of a 256GB model. That's just absurd.

Those Bitlocker-off Toshiba results are what I expected out of a PCIe NVMe drive released in 2015. I had no idea getting such results would require such hoops: being a lucky enough to get a Toshiba drive + disabling Bitlocker.

Regarding security impact: good point and an important consideration to weigh.

 

[Daniel Rubino]

Mine was done with bit locker off. This was before I logged into my sp4 with my microsoft account.

Legit question: Build 10240 or Insider build 10586? I'm on the latter and as today demonstrated, there are some enhancements with TH2.

 

[ttsoldier]

Legit question: Build 10240 or Insider build 10586? I'm on the latter and as today demonstrated, there are some enhancements with TH2.

10240.

 

[ikjadoon]

The pitifully performing PM951 128GB mystery is solved. Spoiler: it's the same issue that was on the 128GB Surface Pro 3's, except that TLC 19nm NAND was released in 2012 on SATA and now we're on TLC 40nm V-NAND released in 2015 on NVMe. What stayed the same? Oh, yeah...T-L-frigging-C.

You guys seen this yet? Crucial's first TLC SSD? It's not the Year of the Sheep, guys, but the the Year of going back 5 years in SSD performance.

At least we should be thankful Microsoft didn't pick this drive (keep scrolling):



Also, the M.2 850 EVO results results seem correlate strongly: the PM951 is the "OEM" NVMe-enabled 850 EVO. Note how weak the M.2 120GB is, the moderate improvement but still weak the 240GB is, and the normal performance of the 500GB.

So, looking at the 120GB Samsung 850 EVO, let's see: it uses a single 128GiB chip (greatly decreasing parallelization):



Samsung, in analysis, decided that it was OK to butcher the 128GB model because these big TLC NAND chips would finally allow them to release affordable 1TB and 2TB SSDs. Looking at the above performance, we can see how badly the 120GB drive was affected. Fewer chips seemingly affect TLC drives way more than MLC drives because the 128GB 850 Pro has no issues like this.

So, why wasn't there an uproar with the Samsung 850 EVO 120GB model? It's as ****ty as the PM951.

Ah, it supports TurboWrite (the SLC cache in the 850 EVO that allows it to put up decent numbers, at least up to 3GB transfers):



Samsung masks this terrible write performance with TurboWrite (turning ~3GB of the TLC into a super-fast SLC cache). Problem is, on the PM951, it doesn't have TurboWrite, so we're stuck with the original TLC performance.

Just for arguments sake, this is why people still have issues recommending TLC NAND: most consumer workloads fit just fine into that ~3GB SLC cache on the 850 EVO. But, even on those, if you push them, the 3GB cache runs out and you get the original poopy performance:



Again, all these issues only affect Samsung TLC drives. But, these affected the original Surface line, too, because they use the PM851! I wondered...Turbo Write will definitely help these pitiful PM951 drives, too, (and let Samsung advertise a 500MB/s write speed instead of the more honest 150MB/s and 330MB/s). Did the PM851 (the 840 EVO drive) ever get Turbo Write? Maybe in the beginning, it wasn't ready, but they enabled it later? From the PM851 tech. sheet:



Noooopee. hahaha, got a little optimistic, didn't I?

Maybe we are crazy lucky, guys. Now, with the Toshiba XG3, Microsoft is finally waking up and giving us a normal SSD drive. I hope you guys with Toshiba drives have a death-grip on them because I'm getting real jealous, haha.

 

[ikjadoon]

Quick question, Daniel. On your Toshiba XG3, would you remind running a quick AS-SSD test? The download link is at the very bottom.

I'm curious on the write speeds.

 

[Daniel Rubino]

Quick question, Daniel. On your Toshiba XG3, would you remind running a quick AS-SSD test? The download link is at the very bottom.

I'm curious on the write speeds.

I'll take a look today.

 

[Devhux]

Looks like Anandtech has revealed what we already know:

http://www.anandtech.com/show/9767/microsoft-surface-book-2015-review/3

"You can see that read speeds are still quite good, even surpassing the Toshiba some of the time, but write speeds are quite a bit lower. In particular, the sequential writes can be less than half the speed of the XG3 drive. This can be compensated for by using some SLC (Single-Level Cell) as a cache, but it does not appear that the PM951 has any caching available to it."

"Summing the storage up, we have a situation that is not ideal. It?s fine to source from different vendors, but the performance differences between the two devices are quite substantial, let alone the long term durability of the TLC NAND"

In other words, this is a lottery we simply shouldn't have with the Surface Book. As the article also points out, Microsoft really should have gone with the Samsung SM951 instead of the PM951.

FWIW, I exchanged my Surface Book for a 256Gb Surface Pro 4, and lucked out in getting a Toshiba drive. Kind of ridiculous that I spent less on the Pro 4 to get double the storage with a much faster drive!

Posted via the Windows Central App for Android