I’m using “NVIDIA Nsight Systems, 2023.3.3.42-233333266658v0 Linux” on Ubuntu 20.04 with the 525.x series driver with an A6000 (device 0) and 3090 (device 1).
I have tried the following:
rebooting as suggested
upgrading to latest in 525.x series driver
Nothing, to my knowledge, changed from one session to the other when this first started. Any help would be appreciated. Thanks.
The same issue occurs after system reboot and I don’t see any nsys processes running.
NVIDIA-SMI 525.147.05 Driver Version: 525.147.05 CUDA Version: 12.0
$ nsys --version
NVIDIA Nsight Systems version 2023.3.3.42-233333266658v0
$ nsys status --environment
Timestamp counter supported: Yes
CPU Profiling Environment Check
Root privilege: disabled
Linux Kernel Paranoid Level = 1
Linux Distribution = Ubuntu
Linux Kernel Version = 5.15.0-84-generic: OK
Linux perf_event_open syscall available: OK
Sampling trigger event available: OK
Intel(c) Last Branch Record support: Available
CPU Profiling Environment (process-tree): OK
CPU Profiling Environment (system-wide): Fail
Are you able to profile a simple CUDA toolkit sample like matrixMul with the same combination of the driver version and nsys version?
What is the full nsys command line you are using to profile your app?
The 2023.3.3 nsys version is from the CUDA toolkit (CTK) version 12.3. That nsys version supports CUDA tracing fully when the driver version is <= 525.60.13. It looks like you have a slightly newer driver and that may be causing some issues. Could you try the nsys version from the web release? Nsight Systems - Get Started | NVIDIA Developer | NVIDIA Developer which is newer than the nsys found in CTK 12.3
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 525.147.05 Driver Version: 525.147.05 CUDA Version: 12.0 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 NVIDIA RTX A6000 On | 00000000:67:00.0 Off | Off |
| 30% 41C P8 20W / 300W | 13MiB / 49140MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
| 1 NVIDIA GeForce ... On | 00000000:68:00.0 On | N/A |
| 35% 45C P8 35W / 350W | 1607MiB / 24576MiB | 35% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| 0 N/A N/A 1856 G /usr/lib/xorg/Xorg 4MiB |
| 0 N/A N/A 2510 G /usr/lib/xorg/Xorg 4MiB |
| 1 N/A N/A 1856 G /usr/lib/xorg/Xorg 375MiB |
| 1 N/A N/A 2510 G /usr/lib/xorg/Xorg 626MiB |
| 1 N/A N/A 3036 G /usr/bin/gnome-shell 171MiB |
| 1 N/A N/A 3794 G ...veSuggestionsOnlyOnDemand 111MiB |
| 1 N/A N/A 5173 G ...RendererForSitePerProcess 14MiB |
| 1 N/A N/A 5583 G ...ost-linux-x64/nsys-ui.bin 102MiB |
| 1 N/A N/A 7841 G gnome-control-center 4MiB |
| 1 N/A N/A 8213 G /usr/lib/firefox/firefox 177MiB |
+-----------------------------------------------------------------------------+
I was able to profile matrixMul using the Nsight UI and default project settings.
I am using the Nsight UI to configure and launch the session. A new project was recreated and the defaults were used. Just changed command line and working directory.
Glad it is working. The web release is 2023.3.1 version which has newer code compared to the 2023.3.3 version from the CTK 12.3. The versioning scheme does not really reflect which is newer, unfortunately. Thanks for confirming that it works for you.