cuda question on tx2

I just ran some CUDA samples on TX2, seems like lower performance than TX1.
Such as some samples in 2_Graphics root, bindlessTexture/simpleGL just 60fps, but on TX1 about 100-200 fps.
What is difference between CUDAon TX1/TX2?

have you tried to run ./ which enables acceleration?
sudo nvpmodel -m 1 will also give max power on CPU’s as far as I remember

Both enabled acceleration, and these examples use GPU operation.

Existence of additional cores doesn’t cause performance increase on itself;
It is due to parallel code architecture if it utilizes more cores in a way that it leads to performance increase, as it seems to me.

Is this a difference of vsync versus not?


Thanks for your question.
We will check this issue and update information to you later.


Thanks for your patience.

We will keep tracking why simpleGL on tx2 yields lower framerate and update information to you later.



Thanks for your patience.

First, there is some error message in comment #7, please ignore it.

In this sample, the refresh rate of glutMainLoop() is controlled by display fps.
The samples logic of displaying fps is not perfect, there can be false positives.

CUDA Samples are not meant for performance benchmarking, its purpose is educating how to use CUDA.
If the intention is to analyze performance than we would suggest to try appropriate GL benchmarks for this.