Question NVMe SSD not detected after GPU passthrough ?

[Basosa]

Hi,
I was trying GPU passthrough using IOMMU on Proxmox. When I started my VM, the screen went black, and I couldn't access Proxmox. After restarting the PC, my Samsung 990 Pro NVMe SSD was no longer detected in the BIOS or in the OS.

I tried connecting it using external adapters, but it wasn’t detected on either Windows or a Raspberry Pi. It only showed up once as "Unknown" in the partition manager, but I couldn't initialize it.

What else could I try?

Thank you!

 

[Aeacus]

What else could I try?

RMA the drive.

You can try to get it working on 2nd PC but once drive doesn't show up in BIOS, it is dead.

Same happened with my Samsung 970 Evo Plus 2TB, ~30 days after purchase and operation. So, RMAd it, got replacement (brand new one) and 2nd drive has now worked without issues for 3 years or so.

 

[Basosa]

Thanks I will try

 

[Aeacus]

NVMe drive can die during operation and not only when powering on the PC.

E.g my 970 Evo Plus died in the same manner. I powered on the PC, booted to OS, inserted my usr psw, desktop showed and i went to kitchen to make my morning tea. Once i arrived back to the PC, my profile log-in screen was again showing. So, i knew something has happened. Did log in to OS again and sure enough, something was different.

Since i kept my old OS drive (Samsung 960 Evo 500GB) as 2nd bootable option, i rebooted to BIOS and sure enough, my new 970 Evo Plus was nowhere to be found. instead PC booted from 960 Evo, once 970 Evo Plus died during operation.

I did switch around the two M.2 drives between M.2 slots, to rule out M.2 slot issue and sure enough, 960 Evo worked but 970 Evo Plus didn't. Thus, i confirmed that MoBo isn't the issue but instead the drive.

 

[evermorex76]

Very weird and bad timing for it to physically fail like that, but possibly trying IOMMU triggered it to use some internal circuitry that was actually defective and blew the whole thing.

Did you look at Device Manager or the equivalent in the Pi when you used the USB connection, or only partition manager? It could have been in DM at least, but not been readable so it wouldn't show up in partition manager. Not that it would matter if it was that busted.

 

[Aeacus]

Very weird and bad timing for it to physically fail like that

Hardware dies on daily basis and often without apparent reason. Hence why brand new hardware comes with a warranty. If none of the hardware ever dies, there would be 0 reason to have warranty in the first place.

Samsung NVMe drives (most high end ones) have ample 5 year warranty. So OP's drive still has warranty to use (990 Pro was released Q3 '22).

 

[evermorex76]

Hardware dies on daily basis and often without apparent reason. Hence why brand new hardware comes with a warranty. If none of the hardware ever dies, there would be 0 reason to have warranty in the first place.

I'm aware of that. But for it to fail at this exact moment is a coincidence with low odds, making it the last thing you'd think of when troubleshooting.

 

[das_stig]

Purely bad luck your SSD failed, Proxmox IOMMU only enables direct access and wouldn't kill the drive.