frequent-errors
Python Out of Memory in ComfyUI
How to respond to 20 Python-side out-of-memory telemetry events, including MemoryError and CPU allocator failures.
Telemetry snapshot queried on April 17, 2026: 20 events across 8 installs for comfyui_fatal_error_detected.
What telemetry shows
- These incidents point to system RAM or CPU-side allocation failures rather than GPU VRAM exhaustion.
- Sample telemetry lines included
MemoryErrorandDefaultCPUAllocator: not enough memory.
What users usually notice
- Workflows fail mid-run even when GPU memory does not look maxed out.
- The machine becomes sluggish or heavily paged before the failure.
What to do next
- Lower resolution, frame count, batch size, or workflow parallelism.
- Close RAM-heavy background apps and increase page file limits if needed.
- Split large workflows into smaller stages and persist intermediate outputs.
Escalate when
- Medium-size workflows fail on machines that should have enough RAM.
- A launcher update causes the same workflow to use much more memory than before.
Документация Wonderful Launcher