Stable Diffusion 1.5 VRAM Requirements
Stable Diffusion 1.5 needs about 1.8 GB of VRAM with FP8 weights at 512×512, loading one model at a time. In FP16 it needs 2.6 GB (3.6 GB with everything loaded), and GGUF Q4 brings it to 1.7 GB. The smallest consumer GPU that runs it is the RTX 3050 8GB with Q4 weights, or the RTX 3050 8GB in FP8.
1.8 GB peak VRAM
Stable Diffusion 1.5 at 512×512 with FP8 weights, loading one model at a time (ComfyUI default on smaller cards). Keeping everything loaded needs 2.7 GB.
- Diffusion model 0.8 GB
- Text encoders 0.1 GB
- VAE 0.2 GB
- Peak activations 0.8 GB
- Overhead 0.8 GB
GPU fit
- RTX 4060 8GBFitsall loaded
- RTX 3060 12GBFitsall loaded
- RTX 4060 Ti 16GBFitsall loaded
- RTX 4090 24GBFitsall loaded
- RTX 5090 32GBFitsall loaded
- RTX 6000 Ada 48GBFitsall loaded
- RTX PRO 6000 Blackwell 96GBFitsall loaded
- Apple M4 Max 128GBFitsall loaded
The VAE decode is the peak here. Turn on tiled VAE decoding to cut it to about 0.4 GB.
Stable Diffusion 1.5 VRAM by precision
At 512×512. Text encoders in FP8 except for the FP16 row.
| Weights | Diffusion model | Peak (swapping) | All loaded | Smallest tier |
|---|---|---|---|---|
| FP16 / BF16 | 1.6 GB | 2.6 GB | 3.6 GB | 4 GB |
| FP8 | 0.8 GB | 1.8 GB | 2.7 GB | 4 GB |
| GGUF Q8_0 | 0.9 GB | 1.8 GB | 2.7 GB | 4 GB |
| GGUF Q5_K_S | 0.6 GB | 1.7 GB | 2.4 GB | 4 GB |
| GGUF Q4_K_S / NF4 | 0.5 GB | 1.7 GB | 2.3 GB | 4 GB |
VRAM by resolution
FP8 weights, with model swapping.
| Resolution | Latent tokens | Activations | Peak VRAM |
|---|---|---|---|
| 512×512 | 4,096 | 0.2 GB | 1.8 GB |
| 768×768 | 9,216 | 0.4 GB | 2.7 GB |
| 1024×1024 | 16,384 | 0.6 GB | 4.1 GB |
About Stable Diffusion 1.5
The model behind most early community checkpoints and LoRAs. Weights are small (about 2 GB in FP16), so the VAE decode step is often the memory peak, especially when upscaling.
| Component | Size |
|---|---|
| Diffusion model | 0.86B parameters (UNet) |
| Text encoders | CLIP ViT-L (0.123B) |
| VAE | 0.084B parameters |
| Classifier-free guidance | Yes (2 passes per step) |
| Released | 2022-10 |
| License | CreativeML Open RAIL-M |
Stable Diffusion 1.5 VRAM: common questions
Can Stable Diffusion 1.5 run on a 4GB GPU?
Yes at 512×512 in FP16. Use tiled VAE decoding for higher resolutions to avoid out-of-memory errors at the final step.
How much VRAM does Stable Diffusion 1.5 need?
At its default 512×512: about 2.6 GB in FP16 with model swapping (3.6 GB with everything loaded), 1.8 GB in FP8 and 1.7 GB with GGUF Q4 weights.