Spinrite Readspeed Samsung 840 EVO SSD results

[eng3]

I am not a spinrite owner myself YET because my main computer is NVME only. However, I recently discovered that my friend is an owner (6.0) but he has not used it for years. He has a desktop and laptop with SATA SSD drives. Both SSD's are Samsung 840 EVO 500GB drives. I mentioned to him that I listen to the podcast and that spinrite could improve SSD performance. I thought I'd provide a report.

It took a while to get it booted on his desktop. He has a Gigabyte B150M-D3H. We enabled CSM, set everything to legacy but it kept wanting to boot UEFI. Finally we set optimized defaults and it gave the legacy boot option for the USB. We still don't know what setting did it.

On the desktop, here is the intial RS output: 72.5 58.4 543.6 543.4 543.4

After Level 2 there was no change so we ran Level 3. Here is the after:

533.3 529.3 533.0 530.2 530.4



On his laptop (Lenovo T440p), it was easier to boot the USB.

Initial RS output: 73.1 542.9 83.9 87.9 542.8

Level 2 made no difference. However, spinrite completely froze (the animation stopped) at a certain point. We continued slightly before the percentage point and it completed ok.
We ran Level 3. It "hung" at around 85%. The animation was still going and I was able to switch back to "Graphic Status Screen" from the "Screen Blanker". I noticed the megabytes counter stopped. On the sector display, the part it was stuck on was on "analyzing". However certain parts of the drive had random text with random colors, most did not match the sector status key.
Then continuing, close to the 95% it complexly froze (animation stopped) again.
Eventually, we were able to complete it. Not sure what is going on here.
Anyways, here are the after RS results: 532.1 528.8 532.3 529.1 528.8


To summarize, Level 2 did nothing. Level 3 restored them. The laptop had some strange issues but resuming slightly before, it did eventually finish.

It would be interesting to measure every day/week/month and plot out the drop in speed over time to try to determine if there is an "optimal" frequency to re-run level 3. Though it might vary across drives and usage habits.

I'm excited for NVME support to try this on my own computer. I was thinking of seeing if I could buy a SATA or USB to NVME adapter but wasnt sure if it would give raw access for spinrite. I also found that someone put virtualbox on a live USB and put spinrite on there and configured for rawdisk access. I might try this but I need to figure out how to setup a live USB. I've only ever downloaded ISOs and loaded them. I also wonder if running badblocks (in non-destructive mode) would do the same. Even if it did, I still would need a way to run readspeed to see if it worked.

 

[eng3]

I assumed that part of the drive was just empty. The first desktop drive was not very full and I don't think has every been too full so probably quite a bit of untouched space. The second laptop drive was fairly full. He said he uses his laptop alot more

 

[Steve]

Level 2 made no difference. However, spinrite completely froze (the animation stopped) at a certain point. We continued slightly before the percentage point and it completed ok.
We ran Level 3. It "hung" at around 85%. The animation was still going and I was able to switch back to "Graphic Status Screen" from the "Screen Blanker". I noticed the megabytes counter stopped. On the sector display, the part it was stuck on was on "analyzing". However certain parts of the drive had random text with random colors, most did not match the sector status key.
Then continuing, close to the 95% it complexly froze (animation stopped) again.
Eventually, we were able to complete it. Not sure what is going on here.

SpinRite's misbehavior is certainly a concern. I believe that I have a Lenovo of the same or similar model, so I'll check to see whether I'm able to recreate that misbehavior.

Otherwise, what you're seeing is what pretty much everyone sees, which is a significant improvement in performance after solid state memory is freshly rewritten. And I agree that it's going to be interesting to learn more about the decrease in performance over time. We have lots to learn. Thanks for sharing your results and experience.

 

[Steve]

I always forget what these levels do, but assuming lvl 3 just wrote to every LBA address they're the opposite of trimmed.

Right. @fzabkar was referring to the initial pre-Level3 performance. And I agree with him that those were likely reflecting pseudo-reads. But it's also interesting to note that after the Level3, ReadSpeed only showed a slight drop in maximum read performance... The the SSD was able to nearly keep up with its link speed.

 

[eng3]

SpinRite's misbehavior is certainly a concern. I believe that I have a Lenovo of the same or similar model, so I'll check to see whether I'm able to recreate that misbehavior.

Otherwise, what you're seeing is what pretty much everyone sees, which is a significant improvement in performance after solid state memory is freshly rewritten. And I agree that it's going to be interesting to learn more about the decrease in performance over time. We have lots to learn. Thanks for sharing your results and experience.

I thought I saw some reports that level 2 brought back performance. It didnt quite make sense though.

I think the lenovo is nearly 10yrs old. So I suppose it is possible something in there is broken. It is interesting that I saw two different types of freeze/hang.

Any comment on the live USB + Virtualbox approach to running readspeed/spinrite in order to get USB and NVME support? In thoery, a minimial linux live USB that has virtualbox and freedos could be created (if it doesnt already exist). Then one would need to setup the "rawdisk" to point to their drive and copy over spinrite/readspeed. If it really works, it's a decent workaround to allow spinrite to work on newer systems before 7.x. And a way to get more sales of spinrite

Also, do you know if running "badblocks -n" would be equiv to level 3 in terms of refreshing an ssd? I know its not equiv to spinrite but it sounds like it would do the same thing that is needed to refresh an ssd if my understanding is correct.

 

[ColbyBouma]

SpinRite's misbehavior is certainly a concern. I believe that I have a Lenovo of the same or similar model, so I'll check to see whether I'm able to recreate that misbehavior.

I believe they used 6.0, not 6.1.

However, I recently discovered that my friend is an owner (6.0) but he has not used it for years.

 

[Steve]

I thought I saw some reports that level 2 brought back performance. It didnt quite make sense though.

It might do that only to the extent that it would re-write sectors that may have been slowing things down. But in general, it's not expected to make much difference.

 

[Steve]

I believe they used 6.0, not 6.1.

Oh!!! Good point, Colby! I'm so "6.1-focused" that I just assumed!

 

[eng3]

I believe they used 6.0, not 6.1.

Correct, I was limited to what I could get my friend to try. The prospect of improving SSD speeds dramatically was enough. I think his interest in continued investigation is much lower now that he's gotten the result. If I still had an SATA SSD, I'd go buy my own copy now

 

[Steve]

Correct, I was limited to what I could get my friend to try.

Oh thank god!! I've been worrying about what might have been missed. Whew! Thanks for the follow-up!

 

[eng3]

Oh thank god!! I've been worrying about what might have been missed. Whew! Thanks for the follow-up!

I'll see if I can convince him to try 6.1. I don't think he'll want to bring his computer offline for several hours again. Then again, it something he'll have to do at some point in the future once the data becomes stale again. On the otherhand, he said he didn't really notice any difference in speed (ie. windows load time)

 

[Steve]

On the otherhand, he said he didn't really notice any difference in speed (ie. windows load time)

One thing we don't know (yet) is the distribution of read performance. If, rather than a widespread slow down, it only comparatively few locations were very slow, then ReadSpeed's inherent performance averaging across a large range would not be reflective of most of the drive's sectors. A primary focus of SpinRite 7 will be to see whether we can nail this down to individual 4Kbyte, 8-sector blocks so that SpinRite can be highly selective with its rewriting and users will be able to get a much more accurate appraisal of drive state.

 

[eng3]

One thing we don't know (yet) is the distribution of read performance. If, rather than a widespread slow down, it only comparatively few locations were very slow, then ReadSpeed's inherent performance averaging across a large range would not be reflective of most of the drive's sectors. A primary focus of SpinRite 7 will be to see whether we can nail this down to individual 4Kbyte, 8-sector blocks so that SpinRite can be highly selective with its rewriting and users will be able to get a much more accurate appraisal of drive state.

Yes I agree. I was just mentioning that although he was impressed with the readspeed improvement, he was slightly disappointed with the "real" improvement. I'll have to ask him next time I see him if he's notice any difference in program load times.