Why the FPS is flowing on xaviie nx?

why the FPS changes so frequently, when it’s infering, the FPS can drop about 60FPS, and the original FPS is about 120FPS(not infering). But no matter it’s infering or not infering, the FPS are flowing… The same yolov5 model can run 130fps on my computer stably(Geforce 3060). and i guess the reason for the flowing FPS on xavier nx is xavier nx’s computing power is lower than my destop… Ohhh, and my camera is DAHENG

Hi,

Could you share which source are you using first?
Thanks.

sorry, i don’t know what does ‘source’ mean? do you mean the code I run? It’s my own code…

Hi,

Inference is usually a computational task so it can be the bottleneck of the pipeline.
Please try to maximize the device performance to see if it helps:

$ sudo nvpmodel -m 0
$ sudo jetson_clocks

Thanks.

Thank you, but sorry… it didn’t work, actually I had already opened MODE 20w 6CORE, and all cores are full

Hi,

Based on the screenshot, jetson_clocks is inactive and the nvpmodel is mode 8.
Could you set it to MAXN and apply the jetson_clocks?

It looks like there are still some resources in your environment (46.3% CPU and 62% GPU).
To further optimize the performance, you can try to profile the app to figure out the bottleneck.

Thanks.