Hello NVIDIA team,
I recently installed a NVIDIA GeForce RTX 5070 on a Windows system.
The GPU is correctly detected by Windows and Device Manager, but it is not usable in CUDA-based frameworks.
System details:
-
OS: Windows 10 / 11 (64-bit)
-
GPU: NVIDIA GeForce RTX 5070
-
Driver: latest available Game Ready / Studio driver
-
CUDA Toolkit: latest available version
-
Frameworks affected:
-
PyTorch
-
ComfyUI (Stable Diffusion / SDXL)
-
Other CUDA-based applications
-
Problem:
GPU is visible in Device Manager
-
NVIDIA driver installs successfully
-
CUDA applications fail with errors such as:
-
no kernel image is available for execution on the device
-
GPU not detected by PyTorch / CUDA
-
-
Applications fall back to CPU mode
What I suspect:
It looks like full CUDA support for RTX 5070 is not yet enabled in the current public drivers / CUDA toolkit.
Because of this, AI frameworks cannot compile or run kernels for this GPU architecture.
Question:
-
Is RTX 5070 officially supported in the current CUDA / driver stack?
-
If not, is support planned in an upcoming driver or CUDA release?
-
Is there a recommended workaround or beta driver to enable CUDA functionality on RTX 5070?
Thank you for your time and support.
I’m happy to provide logs or additional system details if needed.
Best regards.