I recently upgraded from an RTX 3060 to an RTX 5080, but it's noticeably slower than my previous GPU. I suspect the Blackwell architecture isn't fully supported yet. Is there a solution for this? I’ve heard that ComfyUI users are experiencing similar issues, with reports stating that the current stable version of PyTorch doesn’t support Blackwell yet.
For context, on my RTX 3060, deepseek-r1:7b typically took 2 seconds to generate a response. However, on the new RTX 5080, it now takes 22 seconds.
Originally created by @ChuweePappap on GitHub (Feb 20, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/10405
I recently upgraded from an RTX 3060 to an RTX 5080, but it's noticeably slower than my previous GPU. I suspect the Blackwell architecture isn't fully supported yet. Is there a solution for this? I’ve heard that ComfyUI users are experiencing similar issues, with reports stating that the current stable version of PyTorch doesn’t support Blackwell yet.
For context, on my RTX 3060, deepseek-r1:7b typically took 2 seconds to generate a response. However, on the new RTX 5080, it now takes 22 seconds.
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
Originally created by @ChuweePappap on GitHub (Feb 20, 2025).
Original GitHub issue: https://github.com/open-webui/open-webui/issues/10405
I recently upgraded from an RTX 3060 to an RTX 5080, but it's noticeably slower than my previous GPU. I suspect the Blackwell architecture isn't fully supported yet. Is there a solution for this? I’ve heard that ComfyUI users are experiencing similar issues, with reports stating that the current stable version of PyTorch doesn’t support Blackwell yet.
For context, on my RTX 3060, deepseek-r1:7b typically took 2 seconds to generate a response. However, on the new RTX 5080, it now takes 22 seconds.