Recent content by joerambo

Intel/Solidigm 5336 and 5316 are not exactly the fastest drives due to them being QLC. In fact 5316 is slower than good SATA SSD in random 4K writes and that might become a problem, if You have multiple threads writing many files, that need syncing, metadata updates etc, each of those writes...

I would not care about write amplification at all. Setup classical 6 SSD drive RaidZ2 and thats it. For bulk storage, the fact that ZFS writes data twice (in fact due to bulk writes, much less so) does not matter at all. The users that killed SSDs were abusing them with DBMS or some crazy VM...

And the same can be said about initial tests with FIO. Is screaming about the test where the following happens => 10GB of data is bursted into cache and then fsynced. So devices go from 0% to 100% utilization in bursts. The fsync in this case does not have to be explicit...

They do more in proper setup, for example 905P's 1.5TB connected directly to CPU PCIE lanes on X299 platform has the following: iops-test-job: (g=0): rw=randread, bs=(R) 4096B-4096B, (W) 4096B-4096B, (T) 4096B-4096B, ioengine=io_uring, iodepth=1 fio-3.33 Starting 1 process Jobs: 1 (f=1)...

That's what i mean by "vendor" magic, but i'd start with basics, like what LLM recommend: nvme create-ns /dev/nvme0 -s $(nvme id-ctrl /dev/nvme0 | grep tnvmcap | awk '{print $3}') -b 4096

What is the full command You are trying to execute? Sadly create-ns is usually riddled with magic, vendor specific things, might need to google deeper to solve them.

I think You need to start from "basics" of nvme devices. 1) Is there already namespace on device? from above we see this device supports one ns and that is perfectly fine: nvme list-ns /dev/nvme0 if it does, proceed to format it (will loose data) with format desired nvme format --lbaf=1...

@John T Davis I was able to flash my PM893 only with Toolkit version 3.0, so it might be worth for You to try it, before going deeper into debugging the issue. What i did was download samsung tool, set it executable, enumerate my desired disk number and then flash it with latest ( at the time )...

Is smartctl returning the same, without 5 digits after? smartct -x /dev/nvmX where X is your device should show that info in lines at the top.

WOW, this forum and its users are incredible value. Thanks for nothing Samsung! The steps i took to update FW were: downloaded latest Samsung toolkit 3 Linux version as file samsung-tool chmod +x samsung-tool ./ samsung-tool -L | 0 | /dev/sda | SAMSUNG MZ7L37T6HBLA-00A07 |...

Thanks a lot, tested a binary downloaded from DarkServant's on 3.84TB M.2 drive and everything seems to be fine. Now if only someone would find latest FW JXTC404Q for PM893 :(

@dreamkass for "vanilla" Samsung's i was using ( m.2 variant, mind the fw ver ): nvme fw-download -f gdc7502q.bin /dev/nvme0 nvme fw-commit -s 2 -a 3 /dev/nvme0 I think You are trying to commit to slot 1, that is already taken by factory FW. Instead try comminting to slot 2 and activating.

JXTC404Q seems to be newer than what Dell has (21 Apr 2023 date). But that's a viable idea, once/if DELL posts update.

Also a bit of highjack of a thread, but maybe someone already has: PM893 2.5" : JXTC404Q firmware Need to update some drives, as always Samsung and resellers are very "helpful".

Any specifics what was wrong with slot mount PCBs? I think M.2 has an advantage of full speed PCIE 4 or 5 depending on CPU, but if PCIE slot is connected directly to CPU ( as Threadrippers and workstation Xeons are ) they should be full speed? Or is something inherently wrong about PCB mounts...