Seems correct.
Can you lower the driver version to 470.63 and try again?
Just select the third one.
Update the system X config file to use the NVIDIA X driver.
Everything is fine except the login issue(can not login the gdm3)…
I used sudo nvidia-xconfig command,here is my xorg.conf file:
update my debug info:
sudo nvidia-xconfig --query-gpu-info
Section “Device”
Identifier “Device0”
Driver “nvidia”
VendorName “NVIDIA Corporation”
BusID "PCI:*" //change to you got from above command like PCI:131:0:0
and restart gdm3, try again.
I added the BusID
restart the gdm can’t enter system,then reboot the server got this(ssh server is ok,so I can still listen to you)
Hang in here?
Can you enter into console mode?
yes Sir,delete the BusID that will recover
ok.
Did you install cuda driver and display driver both in the system?
Nope，I remove all drvier before install cuda drvier
You’d better use nvidia display driver instead of cuda driver. i am not sure. we do not have such condition to do the test.
Did you plugin monitor into nvidia GPU card?
I install nvidia display driver.
But the login issue happened again,so I delete the xorg.conf file then I can enter the system,Deepstream app still no work
There is no monitor on my GPU card，for now nvidia xserver seems to be the key to the issue
You need to use nvidia X system to make deepstream app using sink type nveglglessink to work. i am not sure why add PCI into xorg.conf cause can not enter into system. it works on my side, but we use another series GPU P4.
Can you use RTSPStreaming for output view?
OK，I undestood，nvidia X system is the key of my issue。
Is it convenient for me to refer your xorg.conf file? Also I will try rtsp and a V100 GPU to see the difference.
Thank you,Sir！
xorg.conf (1.3 KB)
I checked my Jetson Xavier’s xorg.conf file, found this Option different
So I add this on my server.and the xorg.conf issue has been solved! I reboot the server and login normally!
Now,how do I know nvidia X system works?
Try to run deepstream app, which use nveglglessink type, like deepstream-app or test1 sample.
When CUDA is installed, the login issue occurs again. I continue to install CUDA and TRT, and then run sample in ssh,Deepstrem seems still not work
I found why PCI 23 into xorg.conf cause can not enter into system, I changed that to PCI 17:0:0 got a gdm login interface
Great. after login issue solved, deestream app still can not run?
noticed that, in the picture, having words TensorRT linked against cublas 11.6.3 but loaded cublas 11.5.4, which TensorT you are using?