CyrstalDiskMark - 100MB versus 1GB file tests

Hey guys,

I was wondering whether there's much different between the read/write rates noted by CrystalDiskMark when testing 100MB files versus the 1000MB files option. I performed the test on my 1 year old (heavily used) SSD and at 100MB files I'm still getting reasonable read/write speeds - 200MB/s and 130MB/s respectively. But when testing with 1000MB files I'm getting 200MB/s read rates and 12MB/s write rates.


The reason I'm opting for 1000MB tests is that I seem to be having stalls and slow downs when larger chunks of data have to be written in one go. That's both when transferring data, extracting/compressing large files, etc. The first few 100MB's are ok, happening at around 100-150MB/s, but then the system seems to stall and that can last for 10seconds up to several minutes, before continuing the next chunk.

And my 1000MB test results seem to confirm that:

 

The CDM test sizes (100MB vs. 1000MB) allows you to choose how much of the drive you're testing at one time.

With the larger sizes and the higher repetitions (up to 4000MB x 9 times), I feel that is a more accurate reflection of real world usage (but unfortunately, CDM still only reports the highest 'score' and does not show the lowest or average score at all once it finishes its testing.

Worse, it is actually damaging your drive by doing these tests - do so infrequently!

Which SSD are you running, btw? I'd also be interested in how you've used your drive in the last year (eg. video editing, torrenting, etc.).

What will probably be recommended by most here is to backup your data (maybe even clone the whole drive) and do a Secure Erase, then re-install your system again (I recommend a clean install, but I'll concede a clone will be much, much faster).

Good luck.

 

Thanks.

Yea I did a factory restore from the recovery partition on the SSD itself, last year that took <10mins, this time it took almost 2 hours. So definitely something wrong with it.

I'm running the Samsung PM800 256GB drive with the old non-TRIM-supporting firmware.

And yes I've used my machine quite intensely. I'm guessing I'm averaging at least 20GB of data being written to it per day, and moving a lot too (extracting and compressing large files, etc.). The drive has also been close to full at certain times. I also used the machine heavily for torrents.

Mind you performance was still great for a long time, until recently when I found things going downhill quite fast...


Oh and also, yea..writing those test bytes on the drive during a benchmark may have a small effect on drive performance but I think that should be negligible..especially in my situation where that amount of data written is probably less than my daily movement of data on the disk.

 

Well, I don't know if it is neglibible - each test you do is easily up to your daily normal usage - I think that is a lot.

 

In the past, I remember hearing that it may have something to do with the cache size. But I don't know if that's the case.

I didn't notice much of a difference with my C300:

 

Thanks for the testing. That's indeed how I thought a normal-functioning SSD should perform and confirms what I initially thought about my SSD (after several 100MB the writing/performance stalls and pauses).

Yes I heard it may be something to do with the cache too...in any case I do believe the SSD is due for a replacement...