Title nails it but specifically: Say a SN850 2TB has ~256GB of SLC cache. To get those amazing benchmarks. But she’s a dog after she hits the MLC. How to determine the exact LBA of the SLC. Or even better, a function of say nvme-cli to create a namespace for the SLC and MLC seperately. I’d rather I get to decide what goes in the SLC cache vs the drive. Anyone else gone down this rabbit hole?
you could probably provision at the size it does SLC at and then get that perf
Pretty sure most drives just use the cells as SLC not MLC till they need to switch them to cheat the tests
I’ve tested a few 980 PRO 2 TB and wasn’t able to trigger an extreme performance throttle while writing with a little over 1 GB/s from 0 to 100 % full (the software I’m using that fills a volume up with test files to read them back later is quite old) while connected with PCIe Gen4 x4.
Are your concerns of a practical nature (=load scenarios you actually encounter) or more motivated by reviews that mention a slow down after the pSLC cache has been used up?
On the other hand while doing the same with a WD SN 770 2 TB the drive slowed down significantly to about 400 MB/s write and a highly increased latency after a couple of hundred gigabytes written to it sequentially. Even a WD SN550 2 TB didn’t display such low absolute performance numbers while doing the same.
A buddy and I were talking about similar last night. His use case is his NAS and PC where he doesn’t want performance drops. He doesn’t want to get a 980 or 990 Pro as they seem to actually use TLC instead of MLC. Any thoughts appreciated and he would prefer gen 4x4.
If you want consistent performance, you’ll need to get an enterprise drive along with a way to connect U2 to your motherboard.
1 Like
This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.