DreamShaper XL Lightning · 6 GPUs measured first-party · text-to-image · Updated October 2026
What GPU Do You Need for DreamShaper XL Lightning?
DreamShaper XL Lightning on 6 GPUs, measured first-party: NVIDIA GeForce RTX 5090 leads at 122.8 images/min, RTX 3060 trails at 18.3 images/min, and it peaked at 12GB of VRAM.
Benchmarked weights: Lykon/dreamshaper-xl-lightning
NVIDIA GeForce RTX 5090
122.8 images/min on DreamShaper XL Lightning, the ceiling. Measured on our bench. 32GB of VRAM, $1,999 at launch.
- 122.8 images/min on DreamShaper XL Lightning
- 32GB, clears the DreamShaper XL Lightning floor
- 575W board rating
NVIDIA GeForce RTX 4090
91.52 images/min on DreamShaper XL Lightning, fastest card you can buy at retail. Measured on our bench. 24GB of VRAM, $1,599 at launch.
- 91.52 images/min on DreamShaper XL Lightning
- 24GB, clears the DreamShaper XL Lightning floor
- 450W board rating
NVIDIA GeForce RTX 3060
18.28 images/min on DreamShaper XL Lightning, lowest launch price that still fits. Measured on our bench. 12GB of VRAM, $329 at launch.
- 18.28 images/min on DreamShaper XL Lightning
- 12GB, clears the DreamShaper XL Lightning floor
- 170W board rating
GeForce RTX 5060 Ti
28.35 images/min on DreamShaper XL Lightning, most speed per dollar. Measured on our bench. 16GB of VRAM, $429 at launch. That is 66.08 images/min per $1,000 of launch price.
- 28.35 images/min on DreamShaper XL Lightning
- 16GB, clears the DreamShaper XL Lightning floor
- 180W board rating
What GPU Do You Need for DreamShaper XL Lightning?, 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.
DreamShaper XL Lightning. Measured image generation speed by GPU
| GPU | Images/min | s per image | img/W·min | Avg power |
|---|---|---|---|---|
| NVIDIA GeForce RTX 5090 | 122.8 | 0.488 | 0.296 | 414.5 W |
| NVIDIA GeForce RTX 4090 | 91.52 | 0.656 | 0.299 | 306.0 W |
| NVIDIA GeForce RTX 3090 | 47.54 | 1.262 | 0.143 | 332.5 W |
| GeForce RTX 5060 Ti | 28.35 | 2.117 | 0.322 | 88.0 W |
| NVIDIA GeForce RTX 4060 Ti 16GB | 27.95 | 2.147 | 0.211 | 132.2 W |
| NVIDIA GeForce RTX 3060 | 18.28 | 3.283 | 0.124 | 147.2 W |
What the numbers show. Across 6 GPUs measured on our own bench, RTX 5090 is fastest at 123 images/min. The slowest, RTX 3060, manages 18.3, so the spread is 6.7x from top to bottom. RTX 5060 Ti is the most efficient, 28.4 images/min at 88W. Per dollar of launch price, RTX 5060 Ti gives the most (86.2 images/min per $1,000). The fastest card with 16GB or less is RTX 5060 Ti at 28.4 images/min.
About DreamShaper XL Lightning. DreamShaper XL Lightning: from RunDiffusion, on Hugging Face since February 2024, creativeml-openrail-m licence. 24,349 downloads in the last 30 days.
How it compares. RTX 4090: DreamShaper XL Lightning 91.52 images/min, SDXL-Lightning 86.7, Stable Diffusion 1.5 80.14, Stable Diffusion 2.1 60.58, DreamShaper XL Turbo 54.35 (3B). DreamShaper XL Lightning beats all 4 here.
Cost on a rented GPU. 1,000 images of DreamShaper XL Lightning: $0.033 on a RTX 3060 ($0.036/hr, 55 min), $0.053 on a RTX 5090 ($0.39/hr, 8 min, 1.6x the cost).
DreamShaper XL Lightning: 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 | 18.28 | $0.033 |
| NVIDIA GeForce RTX 3090 | $0.12/hr | 47.54 | $0.043 |
| NVIDIA GeForce RTX 5090 | $0.39/hr | 122.8 | $0.053 |
| NVIDIA GeForce RTX 4090 | $0.34/hr | 91.52 | $0.061 |
| GeForce RTX 5060 Ti | $0.14/hr | 28.35 | $0.080 |
Cheapest hourly rate we track on RunPod and Vast.ai, divided by the measured speed. Startup time and storage are extra.
Speed tiers for DreamShaper XL Lightning. 30+ images/min: 3 (RTX 5090, RTX 4090, RTX 3090); 6-30 images/min: 3 (RTX 5060 Ti, RTX 4060 Ti 16GB, RTX 3060). 30 images/min means two seconds or less per picture.
Time per image. DreamShaper XL Lightning at 1024px and 4 steps: 0.5s per image on the RTX 5090, 3.3s on the RTX 3060. A batch of 100 takes 1 min on the fastest card and 5 min on the slowest.
VRAM for DreamShaper XL Lightning. Measured peak 11.3GB, so 12GB is the smallest common card size; smallest card it ran on: RTX 3060 (12GB).
Power on DreamShaper XL Lightning. Most efficient: RTX 5060 Ti, 88W, 51.7 Wh per 1,000 images. Hungriest: RTX 5090, 414W, 56.2 Wh.
Our verdict
Fastest on DreamShaper XL Lightning: NVIDIA GeForce RTX 5090, 122.8 images/min. Cheapest consumer card that ran it: NVIDIA GeForce RTX 3060 ($329, 18.28 images/min). Cheapest to rent per job: NVIDIA GeForce RTX 3060, $0.033 per 1,000 images.
FAQ
- What GPU do I need to run DreamShaper XL Lightning?
- About 12GB. Cheapest consumer card that ran it: NVIDIA GeForce RTX 3060 (12GB, 18.28 images/min).
- How fast is DreamShaper XL Lightning on the NVIDIA GeForce RTX 5090?
- 122.8 images/min at 414W, faster than every datacenter card we ran it on.
- How much does it cost to run DreamShaper XL Lightning in the cloud?
- $0.033 per 1,000 images on a NVIDIA GeForce RTX 3060 at $0.036/hr, cheapest of 5 rentable cards we measured.
- Can I run DreamShaper XL Lightning on a 12GB, 16GB or 24GB card?
- It used 11.3GB at the precision we tested. 12GB: yes; 16GB: yes; 24GB: yes.
- Is the RTX 4090 or the RTX 3090 faster for DreamShaper XL Lightning?
- The RTX 4090: 91.52 vs 47.54 images/min, 93% faster on our bench.
- Should I buy or rent a RTX 5090 for DreamShaper XL Lightning?
- Its $1,999 launch price buys 5,139 rented hours at $0.39/hr, enough for about 37,878x 1,000 images of DreamShaper XL Lightning. Buy only if you'll run more than that.
How we test
DreamShaper XL Lightning at 1024px 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.