GPU cuda compute stress program available ?

trying to understand some commercial software I’m running at work in linux, where nvidia-smi does not report a change in gpu utilization nor power draw on some P100 and V100 cards. I just played around with unigine heaven in centos (at home on gtx970) and nvidia-smi works in linux.
I don’t expect unigine to work on p100 or v100 when I have a p620 installed in a server, especially if i have the monitor off the server motherboard vga.

Is there a CUDA compute gpu stress program available?
right now i’m looking for something really basic… let me set an environment variable or recompile from C/C++ code with a N_GPU_CORES hard coded variable to just do some basic arithmetic (whatever) on all gpu cores. I just want to see power draw, utilization, and temperature change as reported by nvidia-smi.
would anyone be willing to post some basic cuda code to accomplish this?