Parallel Device List (1-4x)

If chaining three Parallel Device Config nodes feels like more wires than a static machine deserves, Parallel Device List (1-4x) is the one-stop version: every device slot and its percentage lives in a single node, and it outputs the exact same DEVICE_CHAIN as the chained approach.

What it is

It's a convenience wrapper over the same idea - build the device list that Parallel Anything splits work across. Where Parallel Device Config gives you one node per device that you chain together, this one gives you four dropdowns and four sliders in a single box.

The inputs

Required:

• device_1 / pct_1 and device_2 / pct_2 - your first two devices and their workload percentages (defaults to 50/50).

Optional:

• device_3 / pct_3 and device_4 / pct_4 - third and fourth slots, defaulting to 0%.

The dropdowns auto-populate from your hardware just like Parallel Device Config does - cpu, cuda:0, cuda:1, plus mps/xpu when present. Any slot left at 0% is simply dropped from the chain, so a two-GPU box sets slots 3 and 4 to 0 and never thinks about them again.

Output

One device_chain output, wired straight into Parallel Anything's device_chain input. From that node's point of view, the list node and a chain of Config nodes are indistinguishable - same data, same behavior, same split logic.

Which one should you use?

Genuinely a matter of taste. The List node wins when your hardware is fixed and you want the fewest widgets on the canvas; the chained Config nodes win when you want each GPU visible as its own node, or when you like rewiring a single link to swap a device. There is no performance difference.

Install

Same pack, same steps:

cd ComfyUI/custom_nodes
git clone https://github.com/FearL0rd/ComfyUI-ParallelAnything.git

Restart ComfyUI (or ComfyUI Manager, search "ComfyUI-ParallelAnything"). No extra dependencies.

One thing worth knowing before you bother: like the rest of the pack, this only does anything on a machine with at least two usable devices. On a single-GPU rig the whole chain is dead weight.