TX2_ Why is the performance of NX lower than that of Xavier NX

Compared with Xavier NX, the execution speed of the algorithm has increased from 50ms to 200ms. According to the evaluation of hardware computing capability, is the current speed of TX2 NX reasonable?

Now the CPU has 6 cores, 1.4g and GPU has 1.1g. Is there any place to speed limit or reduce the frequency?image image

Please provide more details of the execution speed of the algorithm, how do you do the measurement. What’s the tools, or applications?

The program is developed by ourselves, mainly using GPU

With the same program, the computing speed of the two platforms is about four times faster

One 1.33tflop, one is top

21tops

Uhm … aren’t you supposed to add some kind of heatsink and fan? :O Maybe the Tx2 NX just gets too hot?

Now, it is almost 70 degrees, the program runs at a temperature of more than 90, and then the program is hot and dead. It can not be tested in detail

Hi,
Do you mean you see worse performance on Xavier NX when comparing to TX2 NX? Xavier is newer than TX2 and should perform better.

Do see this:
https://forums.developer.nvidia.com/t/jetson-tx2-nx-and-xavier-nx-accelerated-performance-compare/176015/2