From 22f743b0da0e5e6e06c4a667ad382c2ae9edd41f Mon Sep 17 00:00:00 2001 From: Dominic Cerisano Date: Thu, 31 Aug 2023 20:08:12 -0400 Subject: [PATCH] Update gpu_usage.sh to aggregate any number of NVidia GPUs For all attached Nvidia GPUs, aggregate total vRAM available, total vRAM used, and total GPU usage. The graph is at maximum when all GPUs' vRAM is used and all GPUs are 100% busy. Works for any number of attached NVidia GPUs, from 1 to N (the same way the CPU monitor aggregates any number of cores). --- system-monitor@paradoxxx.zero.gmail.com/gpu_usage.sh | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/system-monitor@paradoxxx.zero.gmail.com/gpu_usage.sh b/system-monitor@paradoxxx.zero.gmail.com/gpu_usage.sh index 29d5919b..3e5a99df 100755 --- a/system-monitor@paradoxxx.zero.gmail.com/gpu_usage.sh +++ b/system-monitor@paradoxxx.zero.gmail.com/gpu_usage.sh @@ -28,7 +28,10 @@ checkcommand() # This will print three lines. The first one is the the total vRAM available, # the second one is the used vRAM and the third on is the GPU usage in %. if checkcommand nvidia-smi; then - nvidia-smi -i 0 --query-gpu=memory.total,memory.used,utilization.gpu --format=csv,noheader,nounits | while IFS=', ' read -r a b c; do echo "$a"; echo "$b"; echo "$c"; done + #For all attached Nvidia GPUs, aggregate total vRAM available, total vRAM used, and total GPU usage. + #The graph is at maximum when all GPUs' vRAM is used and all GPUs are 100% busy. + #Works for any number of attached NVidia GPUs, from 1 to N (the same way the CPU monitor aggregates any number of cores). + nvidia-smi --query-gpu=memory.total,memory.used,utilization.gpu --format=csv,noheader,nounits | while IFS=', ' read -r a b c; do let i++; let aa+=a; let bb+=b; let cc+=c; echo $i $aa $bb $cc; done | tail -n1 | while read i aa bb cc; do echo $aa; echo $bb; echo $((cc/i)); done; elif lsmod | grep amdgpu > /dev/null; then total=$(cat /sys/class/drm/card0/device/mem_info_vram_total)