No kernel image is available: check GPU architecture support in each binary
This occurs when a GPU operation lacks a compatible compiled kernel. A device may be too new, omitted as an older architecture, or unsupported by a particular extension. It is not the same as running out of VRAM.
Symptoms and scope
This occurs when a GPU operation lacks a compatible compiled kernel. A device may be too new, omitted as an older architecture, or unsupported by a particular extension. It is not the same as running out of VRAM.
Error fragments for search; IDs, values and filenames may vary:
no kernel image is available for execution on the device
Source-supported context
Public ComfyUI reports cover both RTX 5060 Ti and GTX 1080 Ti scenarios; the latter includes device capability and build support details. xFormers also requires compatible build architectures. Source 1 Source 2 PyTorch architecture APIs Source 3 Source 4
Cases to distinguish
1. The installed Torch wheel does not include the target device architecture.
2. Native Torch works but an attention or custom CUDA extension lacks the required kernel.
3. Asynchronous execution reports an earlier failure at a later node.
Diagnostic sequence
The sequence below is an editorial procedure based on the cited context, not a diagnosis already confirmed for your environment.
Step 1. In ComfyUI's active Python, record GPU model, Torch version and installation source. If the GPU initializes, compare torch.cuda.get_device_capability() with torch.cuda.get_arch_list() for the core build; use python -m xformers.info for xFormers kernels. Then test a native path without third-party attention.
Step 2. If native execution fails, compare the device capability with the core build's architecture list and select an officially documented compatible build. Newest is not automatically best for an older GPU.
Step 3. If only an extension fails, use a supported fallback or a specifically compatible extension build. Evaluate source compilation only when appropriate support and prerequisites exist.
Step 4. Use synchronous debugging in an isolated test if needed to locate the first failing operation, then remove the temporary setting.
Completion check
The actual failing operator executes on that device. Printing the GPU name is insufficient; record core and extension checks separately.
Limitations and cautions
The core build's architecture list does not describe third-party extension kernels. Do not spoof an architecture to bypass compatibility checks. No guide can promise that future binaries will retain every older GPU.
Original sources
- PyTorch local installation — original record checked 2026-09-21.
- xFormers README — original record checked 2026-09-21.
- PyTorch CUDA architecture list and device capability — checked 2026-09-25.
- No kernel image on RTX 5060 Ti — original record checked 2026-09-21.
- No kernel image on GTX 1080 Ti — original record checked 2026-09-21.
Localization prepared: 2026-09-22. The source-check dates above were inherited from the existing article; this translation does not claim they were all rechecked today. No GPU run, minimum-memory measurement or runtime guarantee is supplied.
Related troubleshooting and guides
Was this page helpful?
Anonymous. Only a yes/no counter is stored; no account, IP address or device details.
Sources & references
Localized from the existing Chinese research draft. Original source-check dates are retained; this translation is not new technical verification or a runtime test.
01PyTorch local installationSource checked: 2026-09-2102xFormers READMESource checked: 2026-09-2103No kernel image on RTX 5060 TiSource checked: 2026-09-2104No kernel image on GTX 1080 TiSource checked: 2026-09-2105PyTorch CUDA architecture listSource checked: 2026-09-2506PyTorch CUDA device capabilitySource checked: 2026-09-25