Stable Diffusion 2.1 · 4 GPUs measured first-party · text-to-image · Updated October 2026
What GPU Do You Need for Stable Diffusion 2.1?
Stable Diffusion 2.1 on 4 GPUs, measured first-party: NVIDIA GeForce RTX 5090 leads at 65.53 images/min, RTX 3060 trails at 17.7 images/min, and it peaked at 4GB of VRAM.
Benchmarked weights: Manojb/stable-diffusion-2-1-base
NVIDIA GeForce RTX 5090
65.53 images/min on Stable Diffusion 2.1, the ceiling. Measured on our bench. 32GB of VRAM, $1,999 at launch.
- 65.53 images/min on Stable Diffusion 2.1
- 32GB, clears the Stable Diffusion 2.1 floor
- 575W board rating
NVIDIA GeForce RTX 4090
60.58 images/min on Stable Diffusion 2.1, fastest card you can buy at retail. Measured on our bench. 24GB of VRAM, $1,599 at launch.
- 60.58 images/min on Stable Diffusion 2.1
- 24GB, clears the Stable Diffusion 2.1 floor
- 450W board rating
NVIDIA GeForce RTX 3060
17.71 images/min on Stable Diffusion 2.1, lowest launch price that still fits. Measured on our bench. 12GB of VRAM, $329 at launch.
- 17.71 images/min on Stable Diffusion 2.1
- 12GB, clears the Stable Diffusion 2.1 floor
- 170W board rating
NVIDIA GeForce RTX 4060 Ti 16GB
28.74 images/min on Stable Diffusion 2.1, most speed per dollar. Measured on our bench. 16GB of VRAM, $499 at launch. That is 57.6 images/min per $1,000 of launch price.
- 28.74 images/min on Stable Diffusion 2.1
- 16GB, clears the Stable Diffusion 2.1 floor
- 165W board rating
What GPU Do You Need for Stable Diffusion 2.1?, images/min by GPU
Efficiency: images/min per 100W drawn
Power is the average pulled during the run, sampled at 1Hz. The fastest card is often not the one here, and for anything left running this is the number that shows up on the bill.
Value: images/min per $1,000 of MSRP
Launch price, not street price, so it ages. A speed leaderboard always crowns the most expensive card; this is the counterweight.
Stable Diffusion 2.1. Measured image generation speed by GPU
| GPU | Images/min | s per image | img/W·min | Avg power |
|---|---|---|---|---|
| NVIDIA GeForce RTX 5090 | 65.53 | 0.916 | 0.212 | 309.2 W |
| NVIDIA GeForce RTX 4090 | 60.58 | 0.99 | 0.22 | 275.0 W |
| NVIDIA GeForce RTX 4060 Ti 16GB | 28.74 | 2.088 | 0.214 | 134.6 W |
| NVIDIA GeForce RTX 3060 | 17.71 | 3.388 | 0.109 | 162.4 W |
What the numbers show. Across 4 GPUs measured on our own bench, RTX 5090 is fastest at 65.5 images/min. The slowest, RTX 3060, manages 17.7, so the spread is 3.7x from top to bottom. RTX 4090 is the most efficient, 60.6 images/min at 275W. Per dollar of launch price, RTX 4060 Ti 16GB gives the most (57.6 images/min per $1,000). The fastest card with 16GB or less is RTX 4060 Ti 16GB at 28.7 images/min.
About Stable Diffusion 2.1. Stable Diffusion 2.1: from Manojb, 0.9B parameters, on Hugging Face since June 2025, OpenRAIL++ licence. 70,635 downloads in the last 30 days and 3 community quantizations.
How it compares. RTX 4090: Stable Diffusion 2.1 60.58 images/min, SD Turbo 786.1, Stable Diffusion 1.5 80.14, LCM DreamShaper v7 260.9, PixArt-Sigma XL 24.27. 3 of 4 beat Stable Diffusion 2.1 here.
Cost on a rented GPU. 1,000 images of Stable Diffusion 2.1: $0.034 on a RTX 3060 ($0.036/hr, 56 min), $0.099 on a RTX 5090 ($0.39/hr, 15 min, 2.9x the cost).
Stable Diffusion 2.1: cost per 1,000 images on rented GPUs
| GPU | Cheapest rate | Speed (images/min) | Cost per 1,000 images |
|---|---|---|---|
| NVIDIA GeForce RTX 3060 | $0.036/hr | 17.71 | $0.034 |
| NVIDIA GeForce RTX 4090 | $0.34/hr | 60.58 | $0.092 |
| NVIDIA GeForce RTX 5090 | $0.39/hr | 65.53 | $0.099 |
Cheapest hourly rate we track on RunPod and Vast.ai, divided by the measured speed. Startup time and storage are extra.
Speed tiers for Stable Diffusion 2.1. 30+ images/min: 2 (RTX 5090, RTX 4090); 6-30 images/min: 2 (RTX 4060 Ti 16GB, RTX 3060). 30 images/min means two seconds or less per picture.
Time per image. Stable Diffusion 2.1 at 512px and 30 steps: 0.9s per image on the RTX 5090, 3.4s on the RTX 3060. A batch of 100 takes 2 min on the fastest card and 6 min on the slowest.
VRAM for Stable Diffusion 2.1. Measured peak 3.7GB, so 8GB is the smallest common card size; smallest card it ran on: RTX 3060 (12GB).
Power on Stable Diffusion 2.1. Most efficient: RTX 4090, 275W, 75.7 Wh per 1,000 images. Hungriest: RTX 5090, 309W, 78.6 Wh. At $0.15/kWh: $0.011 per 1,000 images.
Our verdict
Fastest on Stable Diffusion 2.1: NVIDIA GeForce RTX 5090, 65.53 images/min. Cheapest consumer card that ran it: NVIDIA GeForce RTX 3060 ($329, 17.71 images/min). Cheapest to rent per job: NVIDIA GeForce RTX 3060, $0.034 per 1,000 images.
FAQ
- What GPU do I need to run Stable Diffusion 2.1?
- About 4GB. Cheapest consumer card that ran it: NVIDIA GeForce RTX 3060 (12GB, 17.71 images/min).
- How fast is Stable Diffusion 2.1 on the NVIDIA GeForce RTX 5090?
- 65.53 images/min at 309W, faster than every datacenter card we ran it on.
- How much does it cost to run Stable Diffusion 2.1 in the cloud?
- $0.034 per 1,000 images on a NVIDIA GeForce RTX 3060 at $0.036/hr, cheapest of 3 rentable cards we measured.
- Can I run Stable Diffusion 2.1 on a 12GB, 16GB or 24GB card?
- It used 3.7GB at the precision we tested. 12GB: yes; 16GB: yes; 24GB: yes.
- Is the RTX 5090 or the RTX 4090 faster for Stable Diffusion 2.1?
- The RTX 5090: 65.53 vs 60.58 images/min, 8% faster on our bench.
- Should I buy or rent a RTX 5090 for Stable Diffusion 2.1?
- Its $1,999 launch price buys 5,139 rented hours at $0.39/hr, enough for about 20,205x 1,000 images of Stable Diffusion 2.1. Buy only if you'll run more than that.
How we test
Stable Diffusion 2.1 at 512px in diffusers, bf16, native precision with no offload, timed over three generations after a warmup, with power and VRAM sampled throughout. Image speed tracks tensor compute and architecture generation more than memory bandwidth, so the order here differs from our LLM boards.