I put a 4th GPU on one of my motherboard’s PCIe x1 slots using a riser and when I start up Ubuntu 17.04, it seems like the kernel recognizes the 4th GPU because all 4 show up in lspci. However, when I run nvidia-smi, it only recognizes three. When I run dmesg | grep nvidia, I get:
[ 1.010904] [drm] [nvidia-drm] [GPU ID 0x00000100] Loading driver
[ 1.010975] [drm] [nvidia-drm] [GPU ID 0x00000200] Loading driver
[ 1.011021] [drm] [nvidia-drm] [GPU ID 0x00000300] Loading driver
[ 1.011321] [drm] [nvidia-drm] [GPU ID 0x00000500] Loading driver
[ 4.949631] nvidia-uvm: Loaded the UVM driver in 8 mode, major device number 240
[ 7.718236] nvidia-modeset: Allocated GPU:0 (GPU-4924e62f-ac32-93ac-ef48-3d96671d6455) @ PCI:0000:01:00.0
[ 8.370621] nvidia-modeset: Allocated GPU:1 (GPU-91f82e1c-0a3d-132c-4c94-28b3395c7e7e) @ PCI:0000:02:00.0
[ 9.097813] nvidia-modeset: Allocated GPU:2 (GPU-3d1dace1-cebe-b66d-47a8-0e112ecfb029) @ PCI:0000:03:00.0
I’ve been using Linux for a bit, but I’m not a superuser. If I know the address of the missing GPU (PCI:0000:05:00.0), is there a way for me to manually do what nvidia-modeset is doing for the other three GPUs during start-up? Thanks.
nvidia-bug-report.log.gz (271 KB)