GPU clock

Is there any way to set the GPU frequency?
Or does it always run at 1.3 Ghz?

deviceQuery from CUDA returns a Memory Clock rate of 13 Mhz.
Is this true?

A code of mine ran slower on the TX2 than on a GeForce 9600 GT.
It performs several copies to/from GPU memory (and also to shared mem from there).

Does anyone here had a similar experience?

The glmark2 benchmark yielded a score something slightly above 4000.

Hi enthusi,

We provide a script named It is installed by Jetpack.
This script would raise both cpu and gpu clock.

Thank you Wayne! Is there any way I can see that or additional information on the GPU part?
Will it automatically rise under use?

Hi Wayne,

I tried to start the script at boot, but it works only if I run it “by hand”.
I tried calling it in rc.local and putting it in “Startup applications” in Unity without success.

Which is the best way to have max performances automatically at boot?

Thank you


You can’t run in rc.local because there’s a lock that requires the system to have run for at least 60 seconds before it can be execute. To test run immediately after boot. You’re best beat is to write a separate script (make executable) that loads the appropriate settings and run that script in rc.local. There’s an example on for the TK1.


Hi enthusi,

What gpu info do you want to know? We have application called tegrastats for clock info but not in detail.


“sudo nvpmodel -q --verbose” would work well, right?


Yes, nvpmodel still works.

I solved in a less invasive way, this is my “rc.local” script:

#Maximize performances 
( sleep 60 && /home/ubuntu/ )&

exit 0

I wanted to know about the current GPU clock speed (range).
–verbose was what I was missing. Thank you!