Hi everyone,
I measured the power consumption for TensorRT Imagenet running with FP32 and FP16 with MAX-Q. My measurements tell me that FP16 consumes more power than FP32. What could a reason for this
I thank you for the insights.
Hi everyone,
I measured the power consumption for TensorRT Imagenet running with FP32 and FP16 with MAX-Q. My measurements tell me that FP16 consumes more power than FP32. What could a reason for this
I thank you for the insights.
Hi,
It may be related to the percentage of GPU utilization.
You can get GPU utilization with tegrastats. The higher GPU utilization will cause more power consumption.
sudo ./tegrastats
Thanks.
hi AastaLLL,
thank you very much.
Coudl you tell me if it is possible to measure the memory bandwidth also? I wanna measure it while my application is running.
Thanks!
Hi,
There is a CUDA sample to measure the bandwidth of TX2:
/usr/local/cuda-9.0/bin/cuda-install-samples-9.0.sh .
cd NVIDIA_CUDA-9.0_Samples/1_Utilities/bandwidthTest/
make
./bandwidthTest
Thanks.