Include all available metrics in scripting output

[LukeLR]

Hi everyone,

#358 is actually really awesome - just today I thought about writing a prometheus exporter for GPU metrics using nvtop, as my current exporter based on radeontop is not able to provide power metrics, and now I see that this has just been merged 5 days ago. Thanks for the great work!

However, the scripting output doesn't provide all information nvtop has available. For example, the memory usage is only given in percent, the actual memory size of the card is not given. PCIe stats are also omitted. Would it make sense to add these metrics to the scripting output as well?

Thanks!

[Kellsen]

I second this!
I just spent hours to try to get the actual memory usage rather then the percentage into my Conky. I want to use nvtop because nvidia-smi causes lag in my entire system whenever it's queried while nvtop does not. Also the way the memory usage is calculated differs significantly between nvidia-smi and nvtop (i.e. nvidia-smi shows 0.40G usage and nvtop 9% of 8GB total in my 3060 Ti which obviously doesn't add up. I found some explanation but disregard for now, different topic)

Anyhow...my motivations aside: that would be great to have 🙂