Wan 2.2 I2V on a 12GB GPU uses only 8-9GB VRAM—system RAM is the real bottleneck
noise_to_motion · reddit · 2026-10-19
The author benchmarked Wan 2.2 I2V A14B fp8 three times on an RTX 5070 12GB (ComfyUI 0.37 dynamic loading + lightx2v LoRA + RIFE interpolation):
- VRAM: the models total 34GB (high 13.9 + low 13.6 + UMT5 6.4), yet GPU VRAM peaked at 9GB—ComfyUI streams weights from system RAM
- RAM: 40GB committed during renders, so 32GB will page and 16GB will hurt; system RAM is the real bottleneck
- Timing: 832×480/49 frames took 8.5 min; 1.65× frames (49→81) cost only 2.07× time, and RIFE interpolating 80 extra frames took 10s—short clips + interpolation is much cheaper
- Oddities: the high-noise stage ran 2.7× slower per step than low-noise in every run; first render after launch was 3–4× slower while models pinned
Author's guess: 8GB cards can do 480p with enough RAM, but PCIe x8 cards (4060) and DDR4 systems will be notably slower. Next up: same tests with MiniMax H3.
More from Infra
- d-Matrix CEO on why 3D-stacked memory compute could break the AI memory wall — firstadopter · 2026-10-19
- Rubick: an Open-Source Gateway Serving Chat, Embeddings and Calibrated Decisions from One Model — wFXx · 2026-10-19
- Mixing a 9070 XT with an RTX 3080 for local LLMs: 30-50 tok/s and hard lessons — BrightCandle · 2026-10-19
- LLM Quantization Explained: Bit Width Alone Doesn't Determine Accuracy or Speed — techNmak · 2026-10-19
- 303 companies mapped on Nebius: from 1X robots to Mastercard, tracked in independent atlas — pdamodaran · 2026-10-19
- vLLM details MiniMax M3.1 Flash serving stack: 1.7x faster indexer decode via NVFP4 — vllm_project · 2026-10-19