Crash on executing nvidia-smi, Intel i7 6700 with C7Z170 Chipset - Tesla K40c

Hi,

I recently purchased new workstation; Intel i7-6700K 4.0GHz with Supermicro C7Z170-SQ motherboard, 16GB DDR4, SSD 480GB and 865 Watt Power supply. I had installed Centos7 with CUDA 7.5 Toolkit. The nvidia-smi works well in GTX980 and Tesla K20c. Once I changed to K40c, then it nvidia-smi crash and caused PC reboot.

I am suspecting the intel skylake arch. with tesla series of card now. Any comment?

Thank you.

So just running nvidia-smi -q causes a reboot of the machine? Are any error messages being printed before the reboot occurs? I am unclear on the hardware specifications of your system: did you replace a GTX 980 with a K20c, and then replace that with the K40c? Or are all three GPUs installed in the system simultaneously (in which case your PSU would be under-dimensioned).

I have never seen reboots caused by nvidia-smi. Also the K20c and the K40c are quite similar GPUs, and you state everything works fine with the former. Sanity checks: Make sure that the GPU is properly seated in the PCIe slot and is secured at the bracket. Make sure all required power supply cables are properly plugged in, and no power cable splitters or 6-pin to 8-pin converters are used. Install the latest NVIDIA driver package available for this platform.

I tested one card on the board at a time.

The power supply come with 6+2pins. This 8 pins cable is in the same branch. GTX980 and K40c are using the same type of cables, which is 6pin and 8pin.