I’m used Nvidia xavier nx developkit
There seems to be a problem with the jetpack
The performance of the jetpack 4 version and the performance of the jetpack 5 version are different
And after installing jetpack 5, if you insert the sd card with jetpack 4 installed, it will not work.
Is jetpack5 installed in parallel to jetson core emmc and sd card?
April 20, 2023, 11:55pm
Waht does this mean? Any data?
The devkit verson is the module with SD card slot, so you use another eMMC module on devkit carrier board, right?
Insert the SD card into the jetson xavier nx developer kit, install jetpack 5.1 using sdkmanager, remove the sd card, and insert the sd card with jetpack 4.5.1 version installed, it will not boot.
I think it should work, but why doesn’t it work?
And when the test was conducted with the RealSense camera
There is a clear performance difference between jetpack5 and jetpack4.
Hi, bootloader is burnt into the QSPI memory on your device, not SD card, so you could expect that bootloader designed for JetPack 5 would not boot an SD card flashed with JetPack 4.
For the performance difference between JetPack 4 and JetPack 5, do you have any table or log showing it?
The picture above is the result of running on Jetpack 5. All 6 CPUs show high usage.
Below is Jetpack 4 version. Both CPUs show stable usage.
I have 2 jetson xavier nx one is jetpack 5.1 and the other is 4.5.1.
However, the problem shows the same phenomenon from jetpack 5 or later.
Both installed Opencv with Cuda=on and the Ros installation files are identical for both.
jetson with jetpack 5 can’t output image topic even at 15hz
jetpack 4 outputs normally up to 30hz
There is a big difference in performance between the two.
Looks like you device runs at different power mode under JP4 (15W 6CORE) and JP5 (20W 6CORE).
Can you adjust them so both version run at the same power mode? Click the NVIDIA icon on the top-right corner of the screen to make change.
If the issue is still the same, please provide us the code and the environment you are using, so we can try if we can re-produce it.
Power mode doesn’t matter. And with 20W output, shouldn’t the jetpack5 have better performance?
There is no separate code that I am using.
Except for building Opencv4.5.1 from source with Cuda=ON, everything is a program to install on ROS.
In jetpack 5, you can often see the phenomenon that all 6 CPUs are used more than 20% even if the program is not running.
I think this is a problem caused by Bootloader installation.
Hi, can you try run a clean install of JetPack 5 without ROS, and see if the CPU usage is still over 20%?
Also, do you experience difference in terms of time when building OpenCV from source on these two versions?
There is a clear performance difference in building Opencv.
jetpack 4 takes 1 hour to 1 hour 30 minutes
jetpack 5 takes over 2 hours. Performance is degraded than rpi4
So let’s test it with the newly released version 5.1.1.
What I’m talking about is installing via SDKmanager.
I recently bought a Jetson Orin nano and tested jetpack 5.1.1.
Same high CPU usage.
I don’t know what’s wrong
This isn’t just a phenomenon I’m experiencing.
I am attaching a link
we have upgraded some of our Jetsons to JetPack 5.1 from JetPack 4.6 and noticed substantial latency spikes when running our pytorch-based object detector.
Since we are unable to share our detector, we used Yolov7 for this post.
detector.zip (204.9 KB)
Base docker image on JetPack 4.6 was: nvcr.io/nvidia/l4t-pytorch:r32.6.1-pth1.9-py3
and on Jetpack 5.1 we tried: nvcr.io/nvidia/l4t-ml:r35.2.1-py3, nvcr.io/nvidia/l4t-pytorch:r35.2.1-pth2.0-py3, as well as building pytorch for 5.1.
Memory performance has declined severely with the new kernel.
Could you please advice to kernel settings that restore performance to the JetPack 4.4.1 level (4.9.140-tegra)?
sudo mount tmpfs -ttmpfs -osize=4G /mnt/
sudo dd if=/dev/zero of=/mnt/zero.bin bs=1M count=4000
JP441: 4194304000 bytes (4.2 GB, 3.9 GiB) copied, 1.55166 s, 2.7 GB/s
JP502: 4194304000 bytes (4,2 GB, 3,9 GiB) copied, 3,9952 s, 1,0 GB/s
sudo dd if=/mnt/zero.bin of=/dev/null bs=1M