bvaisvil
Add GPU usage monitor
Task List for GPU Support. For linux, GPU support will be compiled using feature flags.
- [x] Support for NVIDIA GPUs on Linux #30
--features nvidia - [ ] Support for AMD GPUs on Linux #31
--features amd - [ ] Support for GPUs on MacOS #32
Would it be possible to have one graph per GPU? I have multiple nvidia GPUs on my machine and it would be very useful to see their historic usage individually as opposed to only one of them. Great app btw.
Thanks for the feedback! Yes, that's certainly possible. My thought was being to switch between which video card was being displayed. However, to clarify, you'd rather have a separate element for each card?
@larrybrit I've got a preview of multiple GPU support in the branch multiple-card-support.
Just tab over the to the GPU section and the ↑ and ↓ switch between cards.
@bvaisvil Apologies for the slow response. It's already useful as it is! But, yes, as you guessed I'd love to see the historic per card, all at once. Side note, the memory usage report is incorrect. I'm using the GPU for 3d rendering and they are using 90-95% of the ram per the GPU renderer / Nvidia-smi report. In zenith, they show a much lower value.
your arrow up / down control would be great to toggle between the history graph and the list of processes on the GPU, instead of toggling through the various GPUs.
@larrybrit No apology necessary! Thank you very much for the feedback.
Looks like the memory usage reported in structnvmlProcessUtilizationSample__t is the Frame Buffer utilization as opposed to Total Memory utilization. Difference between those two values wasn't clear to me until this moment!
I'll work on changing the indicators to correctly reflect the values zenith is presenting.