Jetson xavier nx is slow

[dxt]

Hello,

I have been using jetson xavier nx devkit for quite some time. But all of sudden i feel the speed of device has become slow. Like every small action takes time either it may be opening a simple terminal or any other application. My main worry is i am running the realtime STT (speech to Text) on jetson xnx and rpi5 paralelly. When i run the simple speech to text (whisper) model, the rpi5 started to listening within 2 seconds, but jetson took 30 sec start listening after running the code. So i am doubtful if anything is mismatched here .

I am using:
Jetpack 5.1.4
python3.8
Boot : SD card (128 GB)
swap memory: 11GB
i checked the number of cpu cores running its 6.

Anyone has any idea like what could have gone wrong

[AastaLLL]

Hi,

Could you share more details about how you run the inference?
Do you run it with PyTorch? If so, is CUDA enabled?

Thanks.

[dxt]

@dxt

l, the rpi5 started to listening within 2

I have installed the torch 2.1.0… wheel supported for our current jetpack mentioned from nvidia site and torch with cuda is also enabled( torch.cuda.is_available() is True)

[AastaLLL]

Hi,

Please also try the below commands to maximize the device performance:

$ sudo nvpmodel -m 0
$ sudo jetson_clocks

Then if the performance doesn’t improve, please run the tegrastats concurrently and share the log with us.

$ sudo tegrastats

Thanks.