How much VRAM does Stable Diffusion XL 1.0 need?
Stable Diffusion XL 1.0 needs about 3.3 GB at 8-bit, or 3.1 GB as a 4-bit GGUF, at its default output size.
- Type
- Image model
- Parameters
- 2.6B
- Text encoder
- CLIP ViT-L/14 + OpenCLIP ViT-bigG/14
- Released
- Jul 2023
- License
- CreativeML OpenRAIL++-M
Memory at each precision
| Precision | Weights | Peak at 1024 × 1024 |
|---|---|---|
| 16-bit BF16 | 4.8 GB | 5.6 GB |
| 8-bit FP8 · Q8_0 | 2.5 GB | 3.3 GB |
| 4-bit GGUF Q4_K_M · NVFP4 | 1.4 GB | 3.1 GB |
The CLIP ViT-L/14 + OpenCLIP ViT-bigG/14 text encoder (782 MB at FP8) runs first and then waits in system RAM.
What runs it
| Machine | Best fit | Per image |
|---|---|---|
| RTX 3060 12 GB | 16-bit | 25 s |
| RTX 4060 Ti 16 GB | 16-bit | 15 s |
| RTX 3090 | 16-bit | 9.6 s |
| RTX 4090 | 16-bit | 4.7 s |
| RTX 5090 | 16-bit | 3.9 s |
| RTX PRO 6000 Blackwell | 16-bit | 2.2 s |
| M4 Max 128 GB | 16-bit | 41 s |
| M5 Max 128 GB | 16-bit | 12 s |
| Ryzen AI Max+ 395 128 GB | 16-bit | 19 s |
| DGX Spark 128 GB | 16-bit | 11 s |
| M3 Ultra 256 GB | 16-bit | 23 s |
| M5 Ultra 512 GB | 16-bit | 8.0 s |
| H100 SXM | 16-bit | 1.6 s |
| DGX Station GB300 | 16-bit | 1.2 s |
1024 × 1024, 30 steps. Times are rough, plus or minus 30%.
Where to get it
Questions
How much VRAM does Stable Diffusion XL 1.0 need?
Stable Diffusion XL 1.0 needs about 3.3 GB at 8-bit, or 3.1 GB as a 4-bit GGUF, at its default output size.
Can Stable Diffusion XL 1.0 run on an RTX 4090?
Yes. At 16-bit it takes about 4.7 s per image.