FP8 (e4m3fn / e5m2) Precision
8-bit floating point weight storage that halves checkpoint memory compared to FP16.
Updated Oct 17, 2026 · 1 min read
- Term
- FP8 (e4m3fn / e5m2) Precision
- In Short
- 8-bit floating point weight storage that halves checkpoint memory compared to FP16.
- Applies To
- ComfyUI SD 1.5, SDXL, Flux & Video
Definition: FP8 (e4m3fn / e5m2) Precision
8-bit floating point weight storage that halves checkpoint memory compared to FP16.
RTX 40-series (Ada) and RTX 50-series (Blackwell) accelerate FP8 natively, while RTX 30-series cards use FP8 weight storage with fast FP16 compute.
How to tune this in your ComfyUI workflow
Every parameter in ComfyUI is exposed directly on its node rather than hidden behind global settings tabs. You can inspect or override this behavior per pass in ComfyUI Studio.
New to node graphs? Grab the Windows installer at /download and open a built-in template from the workflow menu.
FAQs
What does FP8 (e4m3fn / e5m2) Precision do in ComfyUI?
8-bit floating point weight storage that halves checkpoint memory compared to FP16. RTX 40-series (Ada) and RTX 50-series (Blackwell) accelerate FP8 natively, while RTX 30-series cards use FP8 weight storage with fast FP16 compute.
Where do I adjust this setting on the canvas?
Check the corresponding loader or KSampler widget in your active workflow graph.
Run on your hardware. Free forever.
The full ComfyUI engine: open source, fast, extensible, and yours to run however you want. Download the desktop app and open this workflow on your own machine.
