CrystalDiskMark performance variations

Garido · 21 Apr 2011 at 18:47

Greetings folks. I've got several Samsung 1TB drives. All are the same model and are of the same batch but upon running them on the same board I've noticed they each seem to perform differently in CrystalDiskMark 2.2 benchmarks.

I just wanted to confirm if the amount of space being used on the drive actually impacts performance?

Each drive has varying levels of performance and is plugged into the same motherboard (I've also tried on an AMD board, so seperate controller and had similar results). The settings I have used is to run 5 tests/500MB each.

Drive 1
SEQ READ 145.2
SEQ WRITE 80.78
512K READ 50.64
512K WRITE 48.09
4K READ 0.635
4K WRITE 1.227

Drive 2 (has files on it)
SEQ READ 113.3
SEQ WRITE 109.5
512K READ 26.90
512K WRITE 55.49
4K READ 0.354
4K WRITE 1.041

It seems a significant variation? Should I be concerned?

Garido · 21 Apr 2011 at 21:02

Okay in some futher insight I have just formatted the drive which did have files on it so that it can be compared to the previous result and that is as follows.

2 identical models on the same motherboard both using the same SATA controller as a result yet we see different performance according to CrystalDiskMark

Drive 2 (now with no files on it)
SEQ READ 151.5
SEQ WRITE 139.5
512K READ 52.45
512K WRITE 67.28
4K READ 0.547
4K WRITE 1.251

readings aren't similar at all nor consistent :/

Garido · 23 Apr 2011 at 10:04

Anyone got an idea why the performance is varied. C'mon guys.

Surveyor · 23 Apr 2011 at 11:17

The benchmark may well vary depending on what your system is doing at the time.

If you benchmarked the same drive 10 times you would probably get 10 different results.

Garido · 23 Apr 2011 at 19:56

Surveyor said:
The benchmark may well vary depending on what your system is doing at the time.

If you benchmarked the same drive 10 times you would probably get 10 different results.

Is there any way to benchmark a drive and get a consistent result. I have benchmarked my SSD and consistently got the same results that reviewers have. This gives me the belief that maybe it's a problem with the drive, rather than the software or my controller as they seem to be fine with my SSD.

Surveyor · 23 Apr 2011 at 20:36

But which drive do you think has a problem?

If you went by these results it looks like that if there is a problem it's with Drive 2:

Drive 1
SEQ READ 145.2
SEQ WRITE 80.78
512K READ 50.64
512K WRITE 48.09
4K READ 0.635
4K WRITE 1.227

Drive 2 (has files on it)
SEQ READ 113.3
SEQ WRITE 109.5
512K READ 26.90
512K WRITE 55.49
4K READ 0.354
4K WRITE 1.041

But once you add these results it looks like that if there is a problem it's with Drive 1:

Drive 2 (now with no files on it)
SEQ READ 151.5
SEQ WRITE 139.5
512K READ 52.45
512K WRITE 67.28
4K READ 0.547
4K WRITE 1.251

:confused:

Garido · 23 Apr 2011 at 20:54

That's the problem. I've also switched the SATA cables as well as controllers (from an AMD board to an Intel board) and I've never received consistent results. Should I RMA the drives? Seatools seems to think the drives are fine though :/

I'm just incredibly confused :/

Surveyor · 23 Apr 2011 at 21:36

Well there isn't anything wrong with the results themselves it's just that they're different from each other but not different in a consistent way!

If the drives perform OK, and don't have errors, then I would leave things as they are.

What are you going to say if you RMA them? Drive 1 isn't as quick as drive 2, no, hang on a sec., drive 2 isn't as quick as drive 1 :confused:

Garido · 23 Apr 2011 at 22:24

Surveyor said:
Well there isn't anything wrong with the results themselves it's just that they're different from each other but not different in a consistent way!

If the drives perform OK, and don't have errors, then I would leave things as they are.

What are you going to say if you RMA them? Drive 1 isn't as quick as drive 2, no, hang on a sec., drive 2 isn't as quick as drive 1

While you raise a valid point I'm still unsure whether the drives performs correctly given the results. What I may do is a more lengthy test and publish some results. Are you able to download Crystal Disk Mark and perform a similar test on your drives?

EDIT: test run

Drive 1 (now with some files on it 140GB)
SEQ READ 149.3
SEQ WRITE 122.8
512K READ 52.24
512K WRITE 51.58
4K READ 0.610
4K WRITE 1.096

Again, different results. I've changed SATA cables too :/

Surveyor · 23 Apr 2011 at 23:13

Garido said:
Are you able to download Crystal Disk Mark and perform a similar test on your drives?

I'm on a RAID10 setup with 4x1TB Samsung F3's.

Results 1000MB:

SEQ READ 162.9
SEQ WRITE 170.5
512K READ 30.07
512K WRITE 38.00
4K READ 0.446
4K WRITE 1.337

Results 500MB:

SEQ READ 150.6
SEQ WRITE 158.1
512K READ 31.32
512K WRITE 53.27
4K READ 0.510
4K WRITE 1.746

Results 50MB:

SEQ READ 695.2
SEQ WRITE 380.4
512K READ 718.3
512K WRITE 329.0
4K READ 25.09
4K WRITE 90.48

It all makes a lot of sense :confused:

Garido · 23 Apr 2011 at 23:33

Surveyor said:
I'm on a RAID10 setup with 4x1TB Samsung F3's.

Results:

SEQ READ 162.9
SEQ WRITE 170.5
512K READ 30.07
512K WRITE 38.00
4K READ 0.446
4K WRITE 1.337

Try and make sense of that.

The above test was with the default setting of 1000MB.

I've just had another go with a 50MB setting and got this:

SEQ READ 695.2
SEQ WRITE 380.4
512K READ 718.3
512K WRITE 329.0
4K READ 25.09
4K WRITE 90.48

It all makes a lot of sense

Wow that's bizzare. I did a 1000MB test 9 times for the final test I did and got something similar to the first test you did. Is CrystalDiskMark only good for SSDs?

Surveyor · 23 Apr 2011 at 23:42

I've amended my post now with a 500MB test.

Garido said:
Wow that's bizzare. I did a 1000MB test 9 times for the final test I did and got something similar to the first test you did. Is CrystalDiskMark only good for SSDs?

No idea, but it doesn't look very consistent.

Garido · 24 Apr 2011 at 00:14

Surveyor said:
I've amended my post now with a 500MB test.

No idea, but it doesn't look very consistent.

I'm going to test the other drives but I'm guessing when it comes to this program that results are very hit and miss. I bet if you ran the same test twice, you would find it unlikely to get 2 of the same results.

Surveyor · 24 Apr 2011 at 00:29

Garido said:
I bet if you ran the same test twice, you would find it unlikely to get 2 of the same results.

Same conclusion I came to at post #4.

That's why I don't benchmark, it'll drive you nuts.

Unless there's some obvious issue leave it alone.

Garido · 24 Apr 2011 at 01:05

Surveyor said:
Same conclusion I came to at post #4.

That's why I don't benchmark, it'll drive you nuts.

Unless there's some obvious issue leave it alone.

Understood though can anyone give an explaination on why the drives perform so inconsistently on such a regular basis. It should be black and white/clear cut.

Garido · 24 Apr 2011 at 06:48

Garido said:
While you raise a valid point I'm still unsure whether the drives performs correctly given the results. What I may do is a more lengthy test and publish some results. Are you able to download Crystal Disk Mark and perform a similar test on your drives?

EDIT: test run

Drive 1 (now with some files on it 140GB)
SEQ READ 149.3
SEQ WRITE 122.8
512K READ 52.24
512K WRITE 51.58
4K READ 0.610
4K WRITE 1.096

Again, different results. I've changed SATA cables too :/

In an interesting development I am testing the drive on my parents machine. It's also the same model (Samsung 1TB F3).

Drive 3 (now with some files on it 12GB)
SEQ READ 151.1
SEQ WRITE 138.8
512K READ 49.89
512K WRITE 66.17
4K READ 0.596
4K WRITE 1.312

Seems the results are now a bit more consistent. I think I have to run the larger/longer tests as it eliminates more of the inconsistency. Still not an exact result but the closest I've seen so far.