Fixed now! It was faulty riser…
I have bought ASUS M.2 board to use 4 additional NVMe’s on my X570D4U-2L2T. I am having some issues with my setup. It seems to be somewhat fighting GPU in x1 slot (FCH), leading to -1 NVMe (I think it’s always last one of those 4) or non existent GPU. Also system takes long time to post, rebooting during post (in A0 – IDE Initialization Is Started)
Sometimes I can see that one NVMe is not detected in BIOS, sometimes it’s there. Sometimes it shows up even in lspci but not working, ZFS is in degraded state, so it can not be read/working.
Board is giving me 5 long beeps on power on (only when power cycled PSU, it’s GPU not recognized) but I got this even when GPU is physically disconnected.
I have tried to enable above 4G decoding, without luck. also it seems to be issue only when there’s GPU and all 4 NVMe’s connected at the same time.
Bifurcation is set to x4,x4,x4,x4.
Resizable BAR is disabled, SR-IOV enabled, SVM enabled, primary video out onboard D-Sub.
Last known good: HP Z Turbo Drive Duo - 2x NVMe + nVidia GPU in x1 slot (trough x16 to x1 adapter with additional power), bifurcation x4,x4,x4,x4 in this setup it was working just fine. before I had gpu in x16 slot and bifurcation set to x8,x4,x4, working fine too.
I had to take off 1mm from upper side of that x16>x1 adapter for gpu (to fit it under motherboard in bottom 2 slots…anyway it should not be the case, since it’s one or another scenario. GPU is working fine, detected in OS, appliance is using GPU without issues…
Is there something else to change or try? I need it working.
Asrock Rack X570D4U-2L2T (X570 - 4 lanes)
AMD R9 5950x (24 lanes should be enough for all)
4x32GB Kingston ECC 3200MHz
nVidia GTX960 6GB (1 lane on FCH)
4x KC3000 1TB (4x4lanes on CPU, ZFS 2 mirrors in 1 pool)
2x Samsung 980 500GB (1 on CPU - 4 lanes, other one trough FCH, ZFS mirror on partition, proxmox boot drive)
6x Toshiba MG08 16TB (on onboard sata controller, ZFS RAID-Z2)
Proxmox VE 8