ComfyUI on 12GB VRAM: measured peaks show offloading turns 24s into 11 minutes
Impossible-Mail-4619 · reddit · 2026-10-27
The author systematically measured VRAM peaks and offloading behavior for 1536×864 generation on an RTX 5070 Ti Laptop (12GB) with ComfyUI 0.39.0 and PyTorch 2.9.1.
Key findings:
- PyTorch silently offloads tensors to system RAM over PCIe — it shows up as 'slowness', not errors. Your GPU isn't broken; you never budgeted VRAM
- Two biggest hogs: Control (composition/face locking) and repainting (640×360 → 1536×864)
- Extra conditioning nodes, reference images, and bigger latents are fine individually but blow past the 11.4GB line when stacked
- Timing is the giveaway: same workflow goes from 20s to 10+ minutes, with high GPU usage actually waiting on PCIe transfers — and offloaded images come out softer
Fixes:
- Cap peak VRAM explicitly; let it error out instead of silently offloading
- Strict model-loading parameters: quantize what can be quantized, preallocate what can be preallocated
Result: full repaint from degraded 640×360 JPEG in 24 seconds with peaks under 11.4GB, wet-pavement reflections and window grids intact. Includes reproducible config, VRAM peak tables for 8GB and 12GB cards, and documented mistakes.
More from Infra
- A100 rental prices still rising in 2026 per a16z data, Jevons paradox hits compute — VraserX · 2026-10-27
- Anthropic S-1 reveals $800B+ in off-books future compute commitments — GaryMarcus · 2026-10-27
- ~42% of Anthropic's ~11GW contracted compute is NVIDIA GPUs — ns123abc · 2026-10-27
- EdgeRock-4B holds KV cache at 72 MiB across 256K context, 99.1% smaller than Qwen3.5-4B — alifcoder · 2026-10-27
- RTX 5080 price jumps from $1658 to $2318 in ten days on Amazon — Frizzy-MacDrizzle · 2026-10-27
- Neon rebuilds its docs search entirely on its own stack: Lakebase Search, Functions, AI Gateway — matei_zaharia · 2026-10-27