Ok thanks a lot for the feedback, I’ll take your comments to heart. > On 30 Jan 2020, at 08:34, Sitsofe Wheeler <sitsofe@xxxxxxxxx> wrote: > > On Wed, 29 Jan 2020 at 23:14, Louwrentius <louwrentius@xxxxxxxxx> wrote: >> >> Thanks a lot for your reply. I have not erased or trimmed the drives. >> I actually wanted to see how the SSDs would behave under worst >> possible circumstances. >> But maybe this causes me comparing apples to oranges. > > To do this fairly, you really would have to the same preconditioning > from an "empty" device on all of them before starting your benchmark. > See the links in the previous email for more on that. > >> So the explanation for the golden gate bridge pattern could be an >> interaction between the benchmark and the garbage collection kicking >> in. >> Maybe if I do erase the drive properly, mark all blocks as empty, the >> pattern should be gone and look more like the intel? > > That's the hope. If it is garbage collection related, starting with > all the blocks erased will hopefully make it take much longer before > garbage collection has to kick in. It could be some SSDs had more > empty cells than others or were better at garbage collecting or had > been written with a data pattern in the past more amenable to the > workload you're doing now etc. > > -- > Sitsofe | http://sucs.org/~sits/
