We want to measure the power consuption on the PX2 AutoChauffeur on max load on the CPU’s iGPUs and dGPUs. Is there a way to monitor the CPU, i/dGPU usage, eventually a benchmarking tool for the PX2? And how do we enable both the GPU’s on the same time?
Is there any specific documentation about the usage of iGPU vs dGPU and power consumption?
Dear didrik.galteland,
Please use “$sudo tegrastats” for checking iGPU & dGPU usage.
GR3D : iGPU
GR3D_PCI : dGPU.
And please use this option when you run an application.
CUDA_VISIBLE_DEVICES=0 : dGPU
CUDA_VISIBLE_DEVICES=1 : iGPU
In CUDA samples for dGPU, ./bandwidthTest CUDA_VISIBLE_DEVICES=0
Hi Steve.
I cannot seem to get the dGPU to work. Whenever I try to run any sample of any kind, the GR3D_PCI always stays on 0%. I do not know if I either do something wrong, or I have to do something in addition to setting the CUDA_VISIBLE_DEVICES option.
It seems that I am able to use the dGPU, but I do not get any indication of it being used, which is my main problem, as I need to benchmark and use the PX2 at its full capacity, so that I can measure it actual power consumption. The power consumption is important for our use, and will be a deciding factor of some other parts of the system. I know that the maximum power usage is around 250-300W, but I need to know the real usage of this very unit
Do you also have any form of benchmarking I can do, to really use all the power on both the Tegra’s and the dGPUs? The bandwith test lasted for about 3 seconds, but what I really need is a benchmarking that lasts for a couple of minutes, or even up to an hour.