SSD testing - unchached performance

[Chomiq]

Hey @W1zzard any chance of updating the SSD test methodology to include the drive performance with cache disabled. Apparently this matters when dealing with SQL databases as they use uncached writes.
Some illustration:

This basically comes from one of the devs at my workplace. He contacted IThardware.pl about it and they have now updated theirs to include this scenario:

Samsung 980 - test niedrogiego dysku SSD M.2 PCIe 3.0 z aspiracjami

Testujemy budżetowy dysk Samsung 980 z interfejsem PCIe 3.0 NVMe. Tani i szybki?

ithardware.pl

First graph shows performance drop when cache is filled to the max and the second one has cache disabled for the test.

 

[arczi19]

This scenario is very specific to SQL though - data is lost if there is a power outage unless your SSD controller has a battery-backup for the cache (I think Microsoft recommends keeping the write cache disbaled). 99% of use cases would want the cache to remain enabled.

 

[W1zzard]

Is it possible you are confusing pseudo-SLC caching with the Windows disk write cache? These are two completely different concepts

I already have testing for the SLC cache size and performance drops, and also a dedicated page to MySQL performance

 

[Chomiq]

Thx @W1zzard, that dev must have missed it. Pointed him to TPU reviews and he's happy now.