self-hosted/ai
§01·compatibility · /check

Krea 2 on RTX 3090

Yes — Krea 2 runs on the RTX 3090 (24 GB). Fastest community-measured result: 14.75 s/img, with 18.75 GB peak VRAM.

✓ runsimageactive30 series24GB VRAM
step-by-step recipe for this pair

Krea 2 Turbo (FP8) on RTX 3090 via ComfyUI: 8-Step Text-to-Image, with the Raw Tier in Reach

imageintermediate16GB+
model
name
Krea 2
slug
krea-2
vertical
image
status
active
open detail ↗
gpu
name
RTX 3090
slug
rtx-3090
vram
24 GB
series
30
open detail ↗
§02·benchmarks
TaskQuantSpeedVRAMWorksConfidenceSourceVerified
t2ifp814.75s/img18.75GB✓github.com· github2026-06-24
§03·how it was measured
  • t2ifp814.75 s/img

    Firsthand community benchmark by Raigor2 (local-inference-benchmarks), also posted to r/StableDiffusion. Mean of 9 clean runs (1 warmup discarded) on one RTX 3090, timed over the ComfyUI websocket progress stream, fixed prompt/seed. Krea 2 Turbo FP8 (fp8_scaled), 8 steps, 1024x1024: 14.75 s/img, 0.65 it/s, 4.07 img/min, 18.75 GB peak. Author+commenter note int8 (not fp8) is the format Ampere can hardware-accelerate (~2x). Reddit discussion: https://www.reddit.com/r/StableDiffusion/comments/1udzugx/

    recorded from github.com ↗ · verified 2026-06-24

§04·more Krea 2 recipes
§05·common questions
Can you run Krea 2 on RTX 3090?

Yes — Krea 2 runs on the RTX 3090 (24 GB). Fastest community-measured result: 14.75 s/img, with 18.75 GB peak VRAM.

How much VRAM does Krea 2 need on RTX 3090?

Measured peak VRAM is 18.75 GB — about 5.3 GB of headroom on the 24 GB card.

Which quantizations have been tested for Krea 2 on RTX 3090?

fp8 — measured in community benchmarks.

How fast is Krea 2 on RTX 3090?

14.75 s/img (t2i) in the fastest community-measured run.

Are there step-by-step instructions for Krea 2 on RTX 3090?

Yes — a step-by-step recipe documents Krea 2 on the RTX 3090, linked at the top of this page.