Check my GPU →

FP8 vs GGUF on a 16 GB RTX 5060 Ti: I timed FLUX.1 dev

Same prompt, same settings, three files. FP8 took 36.4 s per image, GGUF Q8_0 40.6 s and Q4_K_S 47.3 s. Full numbers and setup.

MeasuredBy Tomasz · aboutUpdated 2026-11-15

Most advice about FP8 and GGUF is copied from forum posts written for older GPUs and older ComfyUI versions. So I timed it on my own card: an RTX 5060 Ti 16 GB, the card I use every day for FLUX images.

The result

Seconds per 1024×1024 image, FLUX.1 [dev], 20 steps. Shorter is better.

FP836.4 s
GGUF Q8_040.6 s
GGUF Q4_K_S47.3 s
File typeFileSizePer imageRangePeak VRAM
FP8flux1-dev-fp8.safetensors11.9 GB36.4 s36.0–37.1 s14.8 GB
GGUF Q8_0flux1-dev-Q8_0.gguf12.7 GB40.6 s40.5–40.7 s15.1 GB
GGUF Q4_K_Sflux1-dev-Q4_K_S.gguf6.8 GB47.3 s47.2–47.4 s8.8 GB

Average of 3 timed runs after 1 warm-up run. Peak VRAM is the memory in use on the whole card during the run, including Windows and the text encoder.

The three pictures

Same prompt, same seed (42), same settings; only the file changed. The prompt asked for a cabin at dusk with a small wooden sign that says “VRAM READY”.

FLUX.1 dev FP8: a log cabin at dusk with a VRAM READY sign
FP8 11.9 GB file · 36.4 s
FLUX.1 dev GGUF Q8_0: almost the same cabin and sign
GGUF Q8_0 12.7 GB file · 40.6 s
FLUX.1 dev GGUF Q4_K_S: a different cabin, the sign on one post
GGUF Q4_K_S 6.8 GB file · 47.3 s

FP8 and Q8_0 drew almost the same picture: the same cabin and the same sign, with small differences in the trees and the snow. Q4_K_S drew a different picture from the same seed. The cabin is turned, and the sign stands on one post and is out of focus. With a smaller file it is not only the fine detail that changes; the whole image can come out different.

The VRAM READY sign enlarged from the FP8, Q8_0 and Q4_K_S images
The sign, enlarged FP8 · Q8_0 · Q4_K_S

On the Q4_K_S sign the R at the start of READY is smudged. One image is not a rule, but if your pictures need clean text, I would stay with an 8-bit file.

What it means

  • FP8 is the fastest on an RTX 50 card. Q8_0 is almost the same size but was 12% slower. RTX 40 and RTX 50 GPUs compute FP8 in hardware; GGUF weights have to be unpacked at every step.
  • Q4_K_S saves a lot of memory: about 6 GB less in use on the card. It costs 30% more time per image, and some fine detail.
  • On a 16 GB card, take FP8 for FLUX.1 [dev]. Use Q4_K_S (or Q5/Q6) only if you need the memory for something else at the same time: LoRAs, ControlNet, a bigger resolution, or a second model.

On RTX 30 or older, and on AMD RX 7000 or Intel Arc, there is no FP8 compute, so this result does not carry over. There, Q8_0 is usually the better 8-bit choice. See FP8, GGUF or 16-bit for the full picture.

For comparison, FLUX.1 [schnell] Q8_0 at 4 steps took 9.4 s per image on the same card.

My setup

  • GPU: NVIDIA GeForce RTX 5060 Ti 16 GB, desktop. System RAM: 32 GB.
  • ComfyUI 0.25.0 portable (Windows), PyTorch 2.12 with CUDA 13.0, default launch options.
  • Text encoders: T5-XXL FP8 + CLIP-L. VAE: ae.safetensors. Sampler euler, scheduler simple, guidance 3.5, CFG 1.
  • GGUF files loaded with the ComfyUI-GGUF Unet Loader (GGUF) node; the FP8 file with the normal Load Diffusion Model node.
  • Time is measured by ComfyUI from start to end of the job: sampling plus VAE decode. The prompt stays the same, so the text encoder output is reused between runs.
  • Date: 2026-09-27.

More results from my card are on the RTX 5060 Ti 16 GB page and on all measured and reported results. If you have run the same test on another GPU, send me your numbers.