AnalysisAI Models
10 days ago
User runs Wan 2.2 I2V A14B model on home 4x RTX 3090 rig
The 14B parameter model requires approximately 21.8GB of VRAM per GPU when split across two NVLink islands. The setup utilizes RayLight and XFuser to achieve full utilization on a multi-GPU workstation.
·
10 days ago