Hi,
I've been installing some small older mSATA drives (which use SATA II I believe) into laptops that I rebuild to sell or give away to people who need them. The drives are mostly 32GB (one is 128GB) and I install Windows 10 on these drives to increase boot speed and to create extra space on the rotating HDD.
Once RS became available, I thought it would be useful to benchmark the mSATA drives and preferentially install the fastest ones.
I obtained some surprising (to me) and inconsistent results when I ran RS several times on the same drive, which I will attempt to document here. I kept re-running RS on each drive until the results were consistent.
These 32GB tests were done using RS-PR8, but, unfortunately, I reformatted the USB stick and lost the original log files when I installed the official RS. So, these rounded results are taken from my hand-written notes of the experience. Each line is a new RS run on the same drive.
Let's start with 32GB mSATA drive #2 for a "new" laptop build ...
Now, drive #3 ...
Drive #1 ...
Drive #4 ...
The overall speed results were a little disappointing, but I guess still a lot faster than HDD speeds.
@Steve, do the above gradual improvements show some sort of drive self-healing?
Now for the mysterious results on the larger (and faster) 128GB drive which I use on my regular everyday laptop . I *do* have the logs for these tests, but it's easier to see the changes if I just tabulate the runs in order, one per line (and I'm just looking at only the first two columns) ...
What is going on here?? On a few of the runs, I had requested more details, so I went from my hand-written notes to the logs and found things like this ...
So, it seemed to me that maybe there were "bad spots" on the drive which were throwing out the average values because of the way RS handles that condition.
BTW, @Steve, how *does* RS handle that condition?
Because this is a GPT drive, I cannot run SR on it ... so I though maybe CHKDSK would do a reasonable job of fixing it? Wrong!!
The RS results after running "chkdsk /b" on the drive in Windows 10 were worse ...
... but it does look like some self-healing occurred by the end of the last run. This is some of the detail from the start of the final run ...
@Steve, to my eye, it looks like most of the 511 results in the final column, apart from the one bad spot I'm about to list, are around 660MB/s ... I don't understand why the average is presented as 512.4 (when I do the rough maths myself I get about 651MB/s for the average)? Here's the one bad spot from the 100% column ...
I will attach the log file of this final run, but have kept the others in case any interest in this topic is generated.
Cheers, Peter.
I've been installing some small older mSATA drives (which use SATA II I believe) into laptops that I rebuild to sell or give away to people who need them. The drives are mostly 32GB (one is 128GB) and I install Windows 10 on these drives to increase boot speed and to create extra space on the rotating HDD.
Once RS became available, I thought it would be useful to benchmark the mSATA drives and preferentially install the fastest ones.
I obtained some surprising (to me) and inconsistent results when I ran RS several times on the same drive, which I will attempt to document here. I kept re-running RS on each drive until the results were consistent.
These 32GB tests were done using RS-PR8, but, unfortunately, I reformatted the USB stick and lost the original log files when I installed the official RS. So, these rounded results are taken from my hand-written notes of the experience. Each line is a new RS run on the same drive.
Let's start with 32GB mSATA drive #2 for a "new" laptop build ...
Code:
0% 25% 50% 75% 100
115 145 211 211 209
124 146 211 211 209
124 145 211 211 209
125 146 211 211 209
125 146 211 211 209
Code:
0 25% 50% 75% 100
108 134 210 211 207
109 135 210 211 208
110 136 210 211 208
110 136 210 211 208
Code:
0 25% 50% 75% 100
121 143 211 211 208
121 142 211 211 208
121 142 211 211 208
Code:
0 25% 50% 75% 100
101 133 210 208 208
107 133 210 208 208
107 133 210 208 208The overall speed results were a little disappointing, but I guess still a lot faster than HDD speeds.
@Steve, do the above gradual improvements show some sort of drive self-healing?
Now for the mysterious results on the larger (and faster) 128GB drive which I use on my regular everyday laptop . I *do* have the logs for these tests, but it's easier to see the changes if I just tabulate the runs in order, one per line (and I'm just looking at only the first two columns) ...
Code:
0% 25%
419 540
368 490
327 499
432 544
312 497
386 538
372 500What is going on here?? On a few of the runs, I had requested more details, so I went from my hand-written notes to the logs and found things like this ...
Code:
522.5 590.7 604.5 604.3 604.4
450.1 604.1 605.2 606.4 604.9
472.7 604.1 604.7 600.5 604.5
5.0 604.1 604.5 604.8 604.5
602.6 605.0 604.3 604.9
77.8 604.3 604.6 603.3 604.9
78.2 606.6 604.7 603.6 604.6
79.6 600.7 604.5 604.2 604.4
78.5 604.7 607.4 604.3 605.1
79.5 603.3 601.8 603.2 604.7
484.1 604.3 602.7 604.7 604.8
603.5 604.6 603.8 603.4 604.6
437.9 604.4 602.9 604.4 604.7
405.7 603.2 603.6 604.3 604.9So, it seemed to me that maybe there were "bad spots" on the drive which were throwing out the average values because of the way RS handles that condition.
BTW, @Steve, how *does* RS handle that condition?
Because this is a GPT drive, I cannot run SR on it ... so I though maybe CHKDSK would do a reasonable job of fixing it? Wrong!!
The RS results after running "chkdsk /b" on the drive in Windows 10 were worse ...
Code:
0% 25%
256 469
328 488
318 373
229 368
301 547
428 536
Code:
Driv Size Drive Identity Location: 0 25% 50% 75% 100
---- ----- ---------------------------- ------- ------- ------- ------- -------
81 120GB SSD128G 428.7 536.6 532.6 538.9 512.4
---- ----- ---------------------------- ------- ------- ------- ------- -------
92.3 571.7 654.3 579.7 654.9
90.9 555.1 665.2 606.1 664.6
89.5 574.3 664.7 626.5 664.7
89.7 594.5 664.5 619.3 664.7
531.6 564.5 664.8 583.4 664.9
664.1 664.2 664.4 664.2 664.4
664.7 664.3 664.8 662.8 665.0
663.1 664.1 664.9 664.8 664.5
481.6 663.1 664.2 663.5 664.9@Steve, to my eye, it looks like most of the 511 results in the final column, apart from the one bad spot I'm about to list, are around 660MB/s ... I don't understand why the average is presented as 512.4 (when I do the rough maths myself I get about 651MB/s for the average)? Here's the one bad spot from the 100% column ...
Code:
457.7 662.9 665.1 664.5 664.1
505.9 663.4 660.2 664.4 5.5
628.9 664.6 664.3 664.3
572.7 663.7 661.8 664.3 88.7
540.2 663.8 664.7 665.0 87.1
552.1 664.2 663.0 663.9 89.9
549.2 662.7 664.3 665.2 89.7
658.5 664.8 664.0 664.5 89.7
579.3 662.7 664.0 664.6 545.3
496.6 667.4 663.3 664.0 664.2
556.5 661.1 664.8 665.0 663.7I will attach the log file of this final run, but have kept the others in case any interest in this topic is generated.
Cheers, Peter.
