Samsung vs Toshiba SSD (now with poll)

[flar]

The recent Anandtech review dinged the SB for shipping recent models with a Samsung SSD based on the PM951 - both for the fact that it is a TLC device, but also for poor performance compared to similar SB models that have a Toshiba SSD. They had one of each to compare in their tests and the Toshiba beat the Samsung 2:1 on sequential writes, 3:2 on random writes and was also generally faster on reads, though less dramatically so (and even lost on the 4K non-threaded benchmark by a small margin).

Another source pointed out that most of the review models had come with the Toshiba drive and that the Samsung drives were only showing up more recently.

My SB has one of the Samsung drives and I get similar benchmark numbers as shown in the Anandtech review. Does anyone have a retail SB with the Toshiba drive? Does it achieve the higher numbers in the Anandtech review?

The Microsoft Surface Book Review

 

[Telstar1948]

My Surface Book that I bought Tuesday has the Samsung drive

 

[BearFlag]

Mine has the Samsung drive too. Maybe you should do a poll. Kind of curious how many got the Samsung vs Toshiba drive.

 

[GreyFox7]

Didn't I hear about drive failure causing some issues... maybe the Toshiba drives were unreliable so they switched.

 

[flar]

Mine has the Samsung drive too. Maybe you should do a poll. Kind of curious how many got the Samsung vs Toshiba drive.

Great idea - I added a poll.

I included all sizes as well as Samsung/Toshiba in case there is a breakdown based on size vs manufacturer. I should probably add something about purchase date, but that would make the poll too unwieldy. Perhaps people could add their purchase date in a reply if they have one of the rare Toshiba drives?

 

[GreyFox7]

Revised testing reveals better results although also vastly better with the same model Samsung SSD in the Surface Book.

This highlights one should not put too much stock in any single benchmark particularly early on and going off half cocked.

Revised Notebookcheck Update (11/13/15): A solution to this problem was discovered after this article was written and the data was compiled ... corrects the 4K speeds in AS SSD

Also note that the same Samsung MZFLV256 model SSD in the Surface Book got vastly better results.

'Interestingly, these same constraints do not apply to the NVMe-equipped Surface Book, which posts a total AS SSD score of 2712, demolishing both the Surface Pro 3 (965) and Surface Pro 4 (896)."​

There's no reason the SSD in the SP4 should not get the same performance as in the Surface Book given they use the same model Samsung MZFLVxxx SSDs.

There's more here than is obvious to the casual observer.

 

[mcsenerd]

"Revised Notebookcheck Update (11/13/15): A solution to this problem was discovered after this article was written and the data was compiled ... corrects the 4K speeds in AS SSD ".

Ok.... and the solution was????

 

[flar]

I saw the notification of that post via email and came on here to find a link to the article, but sadly none was posted. I rechecked the NBR review of the Surface Book, but there was no update listed there and the quoted sections appear to be reviewing the drives under another article since it refers to the SB drives in a "3rd party" type of manner ("the drives used in that machine also ...")

Does anyone know where the original article is?

 

[flar]

Ah, I think I found the link:

Microsoft Surface Pro 4 (Core m3) Tablet Review

And here is the full quote:

*Update (11/13/15): A solution to this problem was discovered after this article was written and the data was compiled. Installing the Samsung NVMe driver from this official downloads pagecorrects the 4K speeds in AS SSD and also provides a small boost to performance in CrystalDiskMark and other tests. We have updated the AS SSD results in our database to reflect this development.

And here is the Samsung download link they refer to in the article:

Samsung SSD Downloads | Samsung SSD

I'm going to install the drivers and see what results I get...

 

[GreyFox7]

More importantly in all that was the fact they got much better results on the Surface Book using the same Samsung drive with the same default driver. Why? What happened during the test? Is it valid or is there a problem with the test? Do we even know if that Samsung driver is the right driver? I'm guessing at the least it's not power optimized.

 

[leeshor]

That driver software says it supports the 950 Pro only.

 

[flar]

I tried installing those Samsung NVMe drivers (confirmed in Device Manager) and my numbers in CrystalDiskMark were less than 10% better than their best previous run (and some numbers were off from their previous best). How significantly did their results improve and on what benchmark?