This is mostly an idea than anything else. The idea is make the nvidia.ko (or any other) kernel module publish information about things like temperature, fan speed, GPU memory consumed, GPU processor consumed, etc. Most of this data can be viewed with nvidia-smi or other tools (nvtop), but the idea to publish under /sys (or /proc, not sure which one is correct) is to allow more monitoring tools to get and show the information without the overload of execute the nvidia-smi command directly. I publish this thinking about a GNOME extension called Vitals. This extension collects the information available under /sys and /proc. Other tools could be created, like Zabbix external scripts to collect information about GPUs in a cluster.