amdgpu-fancontrol VS nvtop

I had the same problem on Fedora with RX 5500 XT. To my surprise, GPU fans just didn't turn on, no matter what. I use this script: https://github.com/grmat/amdgpu-fancontrol and it works perfectly, but it just boggles my mind, that there isn't any native solution.

I was gonna use this, but CoreCtrl looks really good! Thanks

https://github.com/grmat/amdgpu-fancontrol <-- i am running the fancontrol-service on my office-pc. no issues so far, but you need some basic linux skills :)

> NVTOP stands for Neat Videocard TOP, a (h)top like task monitor for GPUs and accelerators.

I have been using this on AMD for a long time now.

[0]: https://github.com/Syllo/nvtop

If you're certain your GPU has deeper power saving states than P8, I would start by checking why it's not using them. Maybe tlp, powertop or nvtop (or their documentation) can help.

Did you run any of the "dashboards" while executing the test? Like [NvTop](https://github.com/Syllo/nvtop)

If you plan on making a system monitor/task manager, could you please make it display actual useful information and not just extremely basic stats like most do now. GNOME and KDE both have overly simple applications for this, that just show the CPU and memory utilisation, and it makes finding information like the processor frequency and temperature so much more difficult. On Windows, for example, the task manager performance tab shows a much greater amount of information for each system component, not just total utilisation. GPU usage is also never shown in any of the popular system monitors, something that's probably important for a lot of users. It's not even necessarily impossible since projects such as nvtop are able to. I do hope we'll eventually get a much better GUI program for monitoring our system in Linux.

Can't nvtop do that?

I'm looking for an alternative to nvtop. It's a Ncidia monitoring tool for Linux. You can easily monitor or kill processes via terminal. I used the nvidia-smi command on windows commanline, but it gives very little information.

Try monitoring your VRAM using nvtop (or something), then run training, if the VRAM maxed, training stopped, then VRAM dropped, then it's 100% not enough VRAM.