Monitoring NVIDIA device

[saugat]

I have windows servers that uses NVIDIA GPU for various tasks. I would like to monitor its usage using zabbix server. How can it be possible to show the NVIDIA usage of different server computers using Zabbix graph.
I tried using SNMP technique but I cannot get the usage value of NVIDIA device.

[tsalle]

Hi,
After a quick search on ddg, i found that nvidia provides a tool called nvidia-smi to retrieve some informations about GPU usage.
Maybe you can use some UserParameter in your zabbix agent config file that execute this tool with the corrects parameters.

UserParameter=nvidia.mem,nvidia-smi -q -g 0 2>&1 | grep -A 2 -i utilization | grep -i memory | tail -1 | awk '{print $3}' | sed s/\%//g
UserParameter=nvidia.gpu,nvidia-smi -q -g 0 2>&1 | grep -A 2 -i utilization | grep -i gpu | tail -1 | awk '{print $3}' | sed s/\%//g

And then add two zabbix agent items this the keys nvidia.mem and nvidia.gpu.

source :http://www.matrix44.net/blog/?p=876

[saugat]

Nvidia-smi doesnot support GTX 780

Thank you tsalle for your support but I have all GeForce GTX cards installed which are not supported by nvidia-smi.
Is there any other idea to achieve this?