Does "Bridged" GPUs get read as GPU 0 in conky

[tmick2]

So, I'm sort of baffled. I have 2 Nvidia 1080 GPUs bridged together,
My code for both is
${color maroon}CPU Usage:${color maroon} $cpu% ${cpubar 4} ${color green} GPU:${nvidia modelname} ${nvidia driverversion} ${color orange}GPU0 Usage: Temp: Performance: ${nvidia gpuutil 0} ${nvidia gputemp} ${nvidia perflevel} ${nvidiagraph gpuutil 0 } ${color orchid}GPU1 Usage: Temp: Performance: ${nvidia gpuutil 1} ${nvidia gputemp} ${nvidia perflevel} ${nvidiagraph gpuutil 1 }

As you can see the only differences are the GPU numbers, but I get this view on my Conky:

Could some kind soul explain why that is how it looks to me? My primary questions are:

1. Why does the graph for GPU1 show as a solid block where GPU0 has fluctuation in the way it displays?
2. Why does the GPU Usage change for GPU 0 But not for GPU1, GPU1 always shows 0 usage but the graph is a solid line?
3. Why does neither of them show performance levels?
4. Is this still used to ask questions or has it been moved to Discord or something?

[tmick2]

Well, I added this to my code:
${color green} GPU:${nvidia modelname} ${nvidia driverversion} ${color orange}GPU0 Usage: Temp: Performance: ${nvidia gpuutil}% ${nvidia gputemp}C ${nvidia perflevelcur}% ${nvidiagraph gpuutil 28,C8 8e59faef-cee9-258c-a5e4-0b525715b9d2 -l -m 0 } ${color orchid}GPU1 Usage: Temp: Performance: ${nvidia gpuutil 1}% ${nvidia gputemp 1}C ${nvidia perflevelcur 1}% ${nvidiagraph gpuutil 28,C8 bfce9776-2218-66a3-2ae1-85b5e52077eb -l -m 1}
and it doesn't appear to have made much of a difference.

I'm kind of wondering if nvidia gpuutil and nvidia perflevelcur even work since they don't seem to show anything. It'll show the lines moving but they show as the colors of the graph outline (Orange & Orchid) not the gradient colors of Green& Red.

Maybe I should submit them as bugs?
Anyway, it now looks like this:

NVidia-smi shows this:

You'll notice the appearance of the network graphs and how they show a reading of the KiBs where nvidiagraph just shows line.

[tmick2]

It seems nobody knows the answer to my question or is even bothering to read it. I'll just mark it as solved.

[Caellian]

Related issue: #1615

Conky currently uses XNVCtl, which is somewhat dated and no longer maintained in favor of NVML. Nvidia-smi is essentially a cli frontend for NVML. This is likely the cause for incorrectly reported values which don't match nvidia-smi.

Replacing XNVCtl with NVML is a bit tricky because the headers are provided in different packages in different distros, so just ensuring the headers and library path are consistent for builds is difficult. Using the library is probably not much more difficult though. The issue is stuck because it just takes a week of time to get it to build and I wasn't too enthusiastic about doing that ig.

[tmick2]

Thank you for explaining that, if you ever get "enthusiastic" let me know , you have a willing tester ;-)