AMD Radeon R7 SSD several years old, slow up front

  • Release Candidate 6
    Guest:
    We are at a “proposed final” true release candidate with nothing known remaining to be changed or fixed. For the full story, please see this page in the "Pre-Release Announcements & Feedback" forum.
    /Steve.
  • Be sure to checkout “Tips & Tricks”
    Dear Guest Visitor → Once you register and log-in:

    This forum does not automatically send notices of new content. So if, for example, you would like to be notified by mail when Steve posts an update to his blog (or of any other specific activity anywhere else), you need to tell the system what to “Watch” for you. Please checkout the “Tips & Tricks” page for details about that... and other tips!

    /Steve.

rfrazier

Well-known member
Sep 30, 2020
569
194
@Steve I love the utilities you write, with their 20 KB, etc. executables. :cool:

Here are results on an AMD Radeon R7 SSD which is several years old as well as a couple of "spinners". The SSD is a little slow up front, so I'm running SpinRite Level 2 on it to see if that improves things. I noticed the progress bar pausing a few times when running readspeed. If not, I'll run a SpinRite Level 3 on it and see what happens. That will, unfortunately, use up a complete write cycle on the drive, so I don't want to do that often.

If my spinners are typical, it looks like the speed at the front of the drive is twice the speed at the back of the drive.

Code:
Driv Size  Drive Identity     Location:    0      25%     50%     75%     100
---- ----- ---------------------------- ------- ------- ------- ------- -------
 81  240GB Radeon R7                     319.1   323.9   326.0   329.9   331.1
 82  1.0TB WDC WD1002FAEX-00Z3A0         128.8   121.3   108.1    90.6    63.0
 83  2.0TB HGST HUS726020ALE614          207.5   192.8   175.2   143.2   104.5

                  Benchmarked: Wednesday, 2020-12-30 at 16:31

Ron
 
@rfrazier : Ron, Yes to all of that. That this is why a future SpinRite will be selective in what to rewrites. If you still see a slowdown, try running with the /1 option which will breakdown and enumerate the speeds of the benchmark's 64 individual transfers. It will tend to better show you the slow spots.
 
  • Like
Reactions: rfrazier
@Steve Thanks for the tip. I'm not really having trouble with the drive. Mainly doing this for research and to tinker with your utility. But, it's good to know that it can give more detailed results.

Here's part 2, results after running SpinRite Level 2 read only test. I didn't run the spinners.

Code:
Driv Size  Drive Identity     Location:    0      25%     50%     75%     100
---- ----- ---------------------------- ------- ------- ------- ------- -------
 81  240GB Radeon R7                     321.9   323.5   326.0   333.1   328.4
 82  1.0TB WDC WD1002FAEX-00Z3A0       
 83  2.0TB HGST HUS726020ALE614       

                  Benchmarked: Wednesday, 2020-12-30 at 17:47

Some of the numbers went up, although I don't know if it's enough to imply a cause and effect. I'm now running a SpinRite Level 3 read / write test on the drive just to see if I can get all the numbers up around that nice tasty 333 level. I won't be running full writes often, but it will be interesting to see what happens.

Ron
 
@Steve Should be interesting. I'll post when it's done. Level 3 takes longer of course. It's 2.5 hr into the test and predicting 4.5 hr more. I probably won't post tonight.

FYI (and everyone else) as part of a reply I posted to this thread:


I googled endurance on an 860 evo. I found this:


That says a 500 GB 860 evo is rated at 300 TB TBW. So, if you write the whole drive which is .5 TB, you're using up 1/600th of the life of the drive. I don't know if similar ratios apply to other drives.

Ron
 
@Steve and all. Part 3. Here are the results after mostly finishing the SpinRite Level 3 run. There was a 20 GB section of unpartitioned space that was only about half done. But, all of my main NTFS partition was done.

Code:
Driv Size  Drive Identity     Location:    0      25%     50%     75%     100
---- ----- ---------------------------- ------- ------- ------- ------- -------
 81  240GB Radeon R7                     315.4   313.1   313.5   312.7   333.2
 82  1.0TB WDC WD1002FAEX-00Z3A0
 83  2.0TB HGST HUS726020ALE614

                  Benchmarked: Thursday, 2020-12-31 at 01:52

Doing this actually slightly reduced the read speed, although I'm not sure it's statistically significant. As I mentioned, the original drive wasn't doing that bad. The numbers do seem a bit more consistent. We're right up against the 300 MB/s bandwidth limit of the SATA 3 Gbps interface, so I strongly suspect that this is the limiting factor. If one section had been much lower, I suspect that this would have brought it up to par. I'm not worried about these numbers, and I suspect that the drive / PC is doing about all it's capable of. Regardless, thanks for the cool utility.

Ron
 
  • Like
Reactions: Steve and DarkwinX
@Steve and all. Part 4. I went back and finished the SpinRite Level 3 pass of the 1/2 of the 20 GB of unpartitioned space, just to be thorough. Now, the whole drive has been rewritten by SpinRite. Here are the results.

Code:
Driv Size  Drive Identity     Location:    0      25%     50%     75%     100
---- ----- ---------------------------- ------- ------- ------- ------- -------
 81  240GB Radeon R7                     308.2   308.5   305.7   305.1   308.8

Everything is now very consistent. I don't really understand why the speeds seem to be decreasing a bit. That's above my pay grade. Maybe when things were less consistent, some uber ram caching was going on or something. Or maybe it has something to do with the readspeed averaging algorithm. But, as I said before, all these numbers are up pretty close to the speed limit on SATA 3 Gbps. So, I'm cool with it. Use the info for what it's worth.

Ron
 
@rfrazier hmmmm.. I also noticed that in the results from my drive, which Steve referenced in SN799 the 75% and 100% areas took a hit for the 0-50% to improve.

Perhaps this is where wear levelling is kicking in. When we try and write to the drive using Level 3, the FTL is receiving information from the EZ NAND chip to indicate problem areas and is performing some redistribution from other areas. It would make sense for something to keep track of known good flash pages (or just known bad areas) and when we trigger a write, it determines the ECC is too high and it re maps that location to somewhere in a better part of the disk and also moving what was a better part of the disk to a slightly more troublesome area (but still within acceptable parameters).

Milton did mention that when EZ NAND reports CSP then it takes the area out of service - so this may not fit the theory. Perhaps were dealing with drives that were made without following this standard.

FYI Milton made some great explanation posts in the newsgroups this morning (morning for me anyway) related to FTL and EZ NAND.
 
@DarkwinX Those are some interesting speculations. Frankly, the under the covers magic of how all this gets done is beyond my level of expertise. I'm proceeding under the assumption that if one segment of the readspeed diagnostic is dramatically slower for an SSD, that the data may be in danger. So, consistent across the board is probably good. My SSD wasn't in much trouble at all as far as I know. In the other thread we've been interacting on https://forums.grc.com/threads/slow-ssd-up-front-improved-by-spinrite.362/ , I found it interesting that @thompsondn did get an improvement in the readspeed scores with a SpinRite Level 2 pass. It obviously did something, but I certainly don't know what. In my case, the Level 3 scan made everything all nice and even. I'm assuming that's a good thing. But, as to what exactly it's doing, I don't know.

May your bits be stable and your interfaces be fast. :cool: Ron