Adding CUDA version documentation

Author: justinwlin

docs.json

@@ -206,6 +206,7 @@
               {
                 "group": "Troubleshooting",
                 "pages": [
+                  "references/troubleshooting/cuda-version-issue",
                   "references/troubleshooting/leaked-api-keys",
                   "references/troubleshooting/storage-full",
                   "references/troubleshooting/troubleshooting-502-errors",

images/additional-filter-cuda-version.png

@@ -0,0 +1,26 @@
+---
+title: "CUDA version mismatch"
+---
+
+When using RunPod templates or community templates, such as "runpod/pytorch:2.8.0-py3.11-cuda12.8.1-cudnn-devel-ubuntu22.04", you might occasionally encounter errors such as:
+
+```bash
+RunPod Pytorch 2.8.0 runpod/pytorch:2.8.0-py3.11-cuda12.8.1-cudnn-devel-ubuntu22.04 start container for runpod/pytorch:2.8.0-py3.11-cuda12.8.1-cudnn-devel-ubuntu22.04: begin error starting container: Error response from daemon: failed to create task for container: failed to create shim task: OCI runtime create failed: runc create failed: unable to start container process: error during container init: error running prestart hook #0: exit status 1, stdout: , stderr: Auto-detected mode as 'legacy' nvidia-container-cli: requirement error: unsatisfied condition: cuda>=12.8, please update your driver to a newer version, or use an earlier cuda container: unknown
+```
+
+This error occurs when there's a mismatch between the CUDA version required by your container template and the CUDA driver version installed on the host machine. In this example, the template requires CUDA 12.8 or higher, but the host machine has an older CUDA driver installed.
+
+## Resolution
+
+To resolve this issue when launching a Pod:
+1. Click "Additional filters" in the Pod creation interface
+2. Click "CUDA Versions" filter dropdown
+3. Select a CUDA version filter that matches or exceeds your template's CUDA requirements
+
+<Frame>
+  <img src="/images/additional-filter-cuda-version.png" />
+</Frame>
+
+In this example, the RunPod template name includes `cuda12.8.1`, indicating it requires CUDA 12.8 or higher. You should select a CUDA version filter of at least 12.8.
+
+**Note:** If you're unsure about your template's CUDA requirements (especially for community templates), check the template documentation or error message. You can also just choose the latest CUDA drivers as they are backward compatible - newer drivers can run containers with older CUDA versions installed.