I’m using GeForce RTX 3070 Ti Laptop GPU on windows11 with integrated graphics card. Recently, when i try to get system info using command nvidia-smi --query-gpu=index,timestamp,power.draw,clocks.sm,clocks.mem,clocks.gr --format=csv -l 1, the reports are always: [No data], 210 MHz, 7000 MHz, 210 MHz. At the same time my gpu core freq sticks on 1.4GHz, memory freq 7GHz. (showd in GeForce Experience)
However, when I open GeForce Experience, the power info turns to be corrected, while the frequency infos are still wrong. And after a short while, my gpu core freq drops to 210MHz, and memory freq to 405MHz.
During all test above, my gpu load is 0 with 160MB memory reserved for hardware.
I had no idea what was going on. It’s my fault? And how can I fix it?
Hello @zqaptx, welcome to the NVIDIA developer forums.
What is the output of a simple nvidia-smi? The fact that you don’t see the GPU index (should be 0) and no timestamp would indicate that the driver might not be installed correctly.
What CPU do you use and what OS?
If you are using Linux, please make sure to follow this post:
The nvidia-smi output looks quite normal to me, it might just be that something on the system side interferes with the GFE process reading HW information. It does not work exactly like nvidia-smi does.
You could double-check with another tool like HWInfo or Afterburner to see if the values are consistent. If the GPU is otherwise working correctly I would not worry about slightly different values in GFE.
Thanks for your advice. I used to inspect the GPU infomation using LibreHardwareMonitor. It shows that my GPU core frequency keeps to be 1.4GHz, and memory frequency keeps to be 7GHz all the time. And only a short while after I open GFE, the frequence drop to 210MHz and 405MHz. If GFE is closed, frequence will increase back to 1.4G and 7G again.
I think my gpu is working uncorrectly, but i’m not sure.
Actually, GFE works correctly on my PC, but nvidia-smi a little strange. All the frequency information reported by nvidia-smi is always constant.
good news first, I think your GPU is fine, I see the same effect on one of my systems and the GPU is completely ok.
It might actually be a real issue with nvidia-smi and I will contact engineering to check if this is the case.
I’m new to deep learning with nvidia gpu, and curious about what happened on my gpu. My gpu works excellently before I train a deep learning model and open my computer the next day( I haven’t checked frequency using nvidia-smi before). And I don’t know what happened.
I’m glad to help, and I’m looking forward to knowing the reason.