Improving SSD performance

[ctitanic]

Recently I found this thread:

http://answers.microsoft.com/en-us/...st-fixed/790dc9b9-455d-4216-af64-c57c17d65d10

It seems that running this program makes a difference.

Be sure to run it in Read-Only mode.

 

[Randy31416]

I believe the DiskFresh program must be run in read-write mode to have an effect on the subsequent speed of reads from the SSD.

 

[Jesse San Nicolas]

Very interesting find. I figured the SP3 was a variant of the 840 EVO SSD, as was surprised when I tried to run Samsung Magician software on the SP3 and it said it could not find a Samsung drive on my machine, even though my SP3 has a Samsung SSD.

Also, the SP3 read/performance at times seems to get bogged down quite often, and after monitoring the CPU and the SSD this past month I'm still not sure which is the bottleneck.

It does surprise me that the Samsung SSD in the SP3 shows very poor read/write speeds compared to my Samsung 830, 840, and 850 SSD's I have. Not sure why it shows horrible benchmarks as all Samsung SSD's in the past 4 years get way better numbers than the one in the SP3. I'm guessing it is being limited my firmware, or even the Samsung SSD driver for the SP3.

Either way, seems that Samsung really needs to get this issue fixed ASAP.

 

[Sven]

Something does seem off.

I noticed improved performance turning bitlocker off as in this SP3 SSD test video - but still get nowhere close to the write speeds (granted I suspect the 256gb SSD is faster in this area):

 

[Fazer Rider]

I read in reviews (before buying) the SP3 had a Samsung SSD on board. I actually have a Hynix SSD in my I5/256/8Gb SP3. Upgraded to the latest firmware version (thanks to a Dell tool, Dell uses the same SSD in some laptops).

 

[plop28]

I disabled Bitlocker since the first day.


Everytime i receive a new computer, i bench it :

here are the SSD Bench on my 3 Surface Pro (1-2-3)

Surface Pro 1 (64GB) :
Seq 470,8 / 110,5
512K 384,5 / 102,9
4K 20,10 / 38.93
4K QD32 134.6 / 81.66

Surface Pro 2 (128GB)
Seq 506.4/320.8
512K 380,2/320,4
4K 27.81/65.33
4K QD32 311/291,8

Surface Pro 3 (256GB)
Seq 419.8/262,1
512K 271.5/198.8
4K 19.02/58.69
4K QD32 303.5/233.1

It's not as good as the video for my SP3 and Bitlocker is disable..

edit: i have a SK Hynix SSD (HFS256G3AMNB-2200A) not a samsung

 

[ctitanic]

I believe the DiskFresh program must be run in read-write mode to have an effect on the subsequent speed of reads from the SSD.

That's not what the program documentation says.

 

[GreyFox7]

Rather than chasing our tails it would be better to have MS respond to the report. Support Case??? OR perhaps an insider can comment .

 

[Randy31416]

I believe the DiskFresh program must be run in read-write mode to have an effect on the subsequent speed of reads from the SSD for three reasons.

First, the DiskFresh documentation that I find online http://www.fact-reviews.com/info/diskfresh.aspx without doing an install (likely not official documentation) says "In order to keep the data signal from fading, you need to re-write the data. This is often known as “hard disk maintenance”, and should be done 3 or 4 times a year. While it does not prevent data from being corrupted or deleted, it can go a long way towards ensuring that the magnetic signal does not fade away completely. The way it works is to read every sector of the drive, and then re-write the data found there, provided the drive reported no errors. If this is done on a regular basis, the magnetic signal of every part of the drive will be refreshed long before the signal fades or becomes ambiguous." Now the documentation is talking about its use on spinning hard drives, but I did not see anything that indicated that DiskFresh distinguished among storage media types.

Secon, in that other forum where this issue is discussed http://answers.microsoft.com/en-us/surface/forum/surfpro3-surfhardware/sp3-ssd-possible-firmware-bug-same-as-just-fixed/790dc9b9-455d-4216-af64-c57c17d65d10 and the user DavidN has provided his graphs, he says "Then I ran the DiskFresh tool that reads and writes back the whole drive. The performance now looks like this" and shows the vast improvement between before and after.

Finally, in all the reading I have seen about the Samsung fix to the 840 EVO SSDs, the discussion has been unambigous that the fix reqires reading and rewriting the sectors of the SSD in order to recover performance. (There has been speculation that the fixed firmware may simply periodically do this under the covers to maintain performance, although others hope that there is an algorithm change that actually reads the data as written after the refresh.)

 

[leeshor]

I'm not convinced. The wording of what I have read does in fact sound like it's for a spinning disk. Modern SSD drives do wear leveling and Windows does maintenance in the background. Sounds like hocus pocus.

 

[ctitanic]

I'm not convinced. The wording of what I have read does in fact sound like it's for a spinning disk. Modern SSD drives do wear leveling and Windows does maintenance in the background. Sounds like hocus pocus.

Agree, but there is an issue with Samsung SSD

 

[ctitanic]

Rather than chasing our tails it would be better to have MS respond to the report. Support Case??? OR perhaps an insider can comment .

working on that.