GPU Compatibility Error: NVIDIA GeForce GT 710 with CUDA Capability 3.5 not supported

Discussion in 'PC' started by Ryck, Jun 17, 2023.

Tags:
  1. Ryck

    Ryck Guest

    Hello folks, I downloaded a repository used for voice cloning, and it works fine on Colab. However, when I try to run it locally, I encounter this error: runtime\python.exe extract_feature_print.py cuda:0 1 0 0 C:\Users\Rick\Desktop\RVC/logs/test11 ['extract_feature_print.py', 'cuda:0', '1', '0', '0', 'C:\Users\Rick\Desktop\RVC/logs/test11'] C:\Users\Rick\Desktop\RVC/logs/test11 load model(s) from hubert_base.pt C:\Users\Rick\Desktop\RVC\runtime\lib\site-packages\torch\cuda_init_.py:122: UserWarning: Found GPU0 NVIDIA GeForce GT 710 which is of cuda capability 3.5. PyTorch no longer supports this GPU because it is too old. The minimum cuda capability supported by this library is 3.7.

    From what I understand, as a newbie in this field, my GPU is not compatible with the required compute capability of 3.7. I watched a video online that explained how to install it locally, and the person said that if you have an old graphics card, you should install the version of CUDA that supports your GPU. I checked my GPU, and it supports CUDA 11.4, but I'm still getting the same error. I also downgraded to an older version of PyTorch (1.11.0+cu113), but the error persists. Do you have any ideas on how I could solve this? I'm using Windows 10 for all of this.
     
  2.  
  3. saccamano

    saccamano Audiosexual

    Joined:
    Mar 26, 2023
    Messages:
    1,288
    Likes Received:
    522
    Location:
    CBGB omfug
    GPU-z will tell you everything about your card. Under the Advanced tab is a dropdown for CUDA which will tell you exactly what your card supports. Unfortunately there doesn't seem to be any verified firmware updates to bring it into spec... Most likely if that app is a deal breaker in that you can't run it you'll most likely have to upgrade to higher capability hardware...

    Check the chart on the wiki page for individual GPU CUDA capability ratings here; https://en.wikipedia.org/wiki/CUDA
     
    Last edited: Jun 17, 2023
    • Like Like x 1
    • Agree Agree x 1
    • List
  4. Ryck

    Ryck Guest

    Thank you for the links. If the compute capability of my graphics card is 3.5 and the minimum requirement by PyTorch is 3.7 in its latest releases, then I understand why I'm facing this issue. I thought that by using an older version of PyTorch, it would work. Well, at least that's what the video's author explained, that there wouldn't be a problem with an older GPU. But you know how YouTube videos can be, and if you ask questions, they often go unanswered.
     
    • Like Like x 1
    • Love it! Love it! x 1
    • List
  5. DiRG3

    DiRG3 Producer

    Joined:
    Aug 14, 2022
    Messages:
    149
    Likes Received:
    75
    Yeah it does appear to me that you'll need to upgrade, i don't know where you're located but the website Mercari has been an absolute godsend for me when it comes to finding used components for incredibly cheap, i absolutely swear by that site if i'm on a tight budget (which is extremely often lmao).
     
  6. clone

    clone Audiosexual

    Joined:
    Feb 5, 2021
    Messages:
    7,566
    Likes Received:
    3,333
    Keep in mind that once you update a driver on a windows machine with a newer version; changing the Application back to the older version will not result in your driver or application reverting to what the old version might be looking for. You *might* be more successful using the older version of the software application by also reinstalling your older driver for the Nvidia card; should you care to tinker more. Obv a new GPU is best answer for your situation.
     
    • Like Like x 1
    • Agree Agree x 1
    • List
  7. Ryck

    Ryck Guest

    I live in Argentina, and for quite some time now, we have been experiencing constant devaluation. As a result, it is becoming increasingly difficult to acquire things. Consequently, it's impossible for me to buy a new GPU, but thanks for the information. I've been checking elsewhere, and it seems that voice cloning utilizes the GPU's RAM, which in Colab is around 12GB. Even so, it is slow in Colab. I can only imagine how long it would take if I were to use my GPU, which only has 1GB of RAM. Well, the intention of having it locally is because I have the impression that voice cloning could have a negative impact, and there might be some restrictions in Colab. I don't know, that thought came to mind because we are talking about cloning "vocal identities" in a way. So, I said to myself, 'Rick, just in case, make a copy on your PC.' But, well, it seems that it's not going to work. I've been installing all sorts of dependencies and uninstalling them for days, like Python, Librosa, PyTorch, etc. And it's also true what they mention about drivers. If I overlooked that, in fact, several times in past projects, I had blue screens due to malfunctioning drivers, and I had to uninstall everything, reinstall, and also change the paths. Anyway, I suppose I'll have to save up and buy a graphics card in the future. However, out of curiosity, if anyone knows how to make that error work, I would appreciate it
     
Loading...
Similar Threads - Compatibility Error NVIDIA Forum Date
Help with Waves older version compatibility Software Sep 26, 2024
WTF Spitfire? - Breaking compatibility with old LABS plugin. Software Jul 31, 2024
Moving from Intel to M1, need advice about compatibility. Mac / Hackintosh Apr 9, 2024
Recommended MacOS version for compatibility w. Audio Apps Mac / Hackintosh Mar 7, 2024
Amd ryzen 9 and uad2 pcie compatibility PC Sep 24, 2023
Loading...