FearL0rd manages hardware resources with ComfyUI-AnyDeviceOffload

    
        By vramkickedin    
     | 
    
            February 25, 2026 at 1:20 pm        
    
     | 
    
        2 min read

FearL0rd introduces ComfyUI-AnyDeviceOffload, a new custom node for ComfyUI designed to give users precise control over hardware resource management. This tool allows users to manually assign models, VAEs, and CLIPs to specific devices, including cuda:0, cuda:1, CPU, or Mac MPS.

It specifically targets and resolves common 'Out of Memory' (OOM) errors while enabling complex multi-GPU workflows. The node also addresses persistent crashes often encountered when running modern workflows like Flux on CPUs or secondary graphics cards.

Technical capabilities of the new node

Manual Device Selection.
Intelligent VRAM Management includes a "Kill Switch" that automatically unloads models.
VAE Patcher fixes RuntimeErrors and CuDNN errors.
CPU 'Safe Mode' intercepts xFormers calls on CPU to prevent crashes.

Development focused insights

The development of this node focused on resolving specific hardware bottlenecks in existing AI workflows. The GitHub repository states,

'This custom node gives you total control over where your models run and where they live in memory.'

FearL0rd explains that the project

'solves common "Out of Memory" (OOM) errors, enables multi-GPU workflows, and fixes persistent crashes when trying to run modern workflows (like Z_Image/Flux) on CPUs or secondary GPUs.'

This provides a practical solution for users working with limited or complex hardware setups.