Stable Diffusion 1.5 · 28 GPUs measured first-party · text-to-image · Updated October 2026

What GPU Do You Need for Stable Diffusion 1.5?

We ran Stable Diffusion 1.5 at 512px on 28 GPUs, from budget cards to datacenter parts, and measured image speed, power draw and peak VRAM on every one. The fastest was NVIDIA GeForce RTX 5090 at 102.4 images/min.

Benchmarked weights: stable-diffusion-v1-5/stable-diffusion-v1-5

Fastest we measured
NVIDIA GeForce RTX 5090

NVIDIA GeForce RTX 5090

102.4 images/min on Stable Diffusion 1.5, the ceiling. Measured on our bench. 32GB of VRAM, $1,999 at launch.

Pros
  • 102.4 images/min on Stable Diffusion 1.5
  • 32GB, clears the Stable Diffusion 1.5 floor
Cons
  • 575W board rating
Best consumer card
NVIDIA GeForce RTX 4090

NVIDIA GeForce RTX 4090

80.14 images/min on Stable Diffusion 1.5, fastest card you can buy at retail. Measured on our bench. 24GB of VRAM, $1,599 at launch.

Pros
  • 80.14 images/min on Stable Diffusion 1.5
  • 24GB, clears the Stable Diffusion 1.5 floor
Cons
  • 450W board rating
Cheapest card that runs it
NVIDIA GeForce RTX 5060

NVIDIA GeForce RTX 5060

23.9 images/min on Stable Diffusion 1.5, lowest launch price that still fits. Measured on our bench. 8GB of VRAM, $249 at launch.

Pros
  • 23.9 images/min on Stable Diffusion 1.5
  • 8GB, clears the Stable Diffusion 1.5 floor
Cons
  • 145W board rating
Best value
GeForce RTX 5070

GeForce RTX 5070

39.2 images/min on Stable Diffusion 1.5, most speed per dollar. Measured on our bench. 12GB of VRAM, $549 at launch. That is 71.4 images/min per $1,000 of launch price.

Pros
  • 39.2 images/min on Stable Diffusion 1.5
  • 12GB, clears the Stable Diffusion 1.5 floor
Cons
  • 250W board rating
102.4images/min
Fastest: NVIDIA GeForce RTX 5090
measured
27
Cards that run Stable Diffusion 1.5
of 27 we have data for
0
Cards that can't run it at all
published as hard gates, not omissions
544%
Fastest vs slowest that fits
102.4 vs 15.9 images/min

What GPU Do You Need for Stable Diffusion 1.5?, images/min by GPU

NVIDIA GeForce RTX 5090
102.4 images/min
NVIDIA H100 80GB HBM3
80.45 images/min
NVIDIA GeForce RTX 4090
80.14 images/min
GeForce RTX 5080
61.19 images/min
NVIDIA L40S
58.54 images/min
GeForce RTX 4080 Super
51.96 images/min
NVIDIA GeForce RTX 4080
51.23 images/min
GeForce RTX 5070 Ti
49.51 images/min
NVIDIA GeForce RTX 4070 Ti Super
46.09 images/min
NVIDIA GeForce RTX 4070 Ti
44.44 images/min
NVIDIA GeForce RTX 3090 Ti
43.77 images/min
NVIDIA RTX PRO 4000 Blackwell
40.91 images/min
NVIDIA GeForce RTX 3080 Ti
40.7 images/min
NVIDIA GeForce RTX 4070 Super
39.83 images/min
NVIDIA RTX 4000 (Ada Generation)
39.51 images/min

Top 15 shown; 13 more cards in the full table below.

Efficiency: images/min per 100W drawn

NVIDIA L4
39.72 images/min / 100W
NVIDIA RTX 4000 (Ada Generation)
32.92 images/min / 100W
GeForce RTX 5060 Ti
30.11 images/min / 100W
NVIDIA RTX PRO 4000 Blackwell
29.28 images/min / 100W
GeForce RTX 4080 Super
28.27 images/min / 100W
NVIDIA H100 80GB HBM3
27.35 images/min / 100W
NVIDIA GeForce RTX 4090
24.65 images/min / 100W
NVIDIA GeForce RTX 4070 Ti Super
24.5 images/min / 100W
NVIDIA L40S
23.83 images/min / 100W
NVIDIA GeForce RTX 4080
23.67 images/min / 100W
NVIDIA GeForce RTX 5090
22.69 images/min / 100W
GeForce RTX 5080
22.15 images/min / 100W
GeForce RTX 5070
21.61 images/min / 100W
NVIDIA GeForce RTX 4070 Ti
21.39 images/min / 100W
GeForce RTX 5070 Ti
21.26 images/min / 100W

Top 15 shown; 12 more cards in the full table below.

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

NVIDIA GeForce RTX 5060
95.98 images/min / $1k
GeForce RTX 5070
71.4 images/min / $1k
GeForce RTX 4060
67.59 images/min / $1k
NVIDIA GeForce RTX 4070 Super
66.49 images/min / $1k
GeForce RTX 5070 Ti
66.1 images/min / $1k
GeForce RTX 5080
61.25 images/min / $1k
NVIDIA GeForce RTX 4070
59.03 images/min / $1k
GeForce RTX 5060 Ti
58.6 images/min / $1k
NVIDIA GeForce RTX 4070 Ti Super
57.68 images/min / $1k
NVIDIA GeForce RTX 4070 Ti
55.62 images/min / $1k
NVIDIA GeForce RTX 4060 Ti 16GB
52.57 images/min / $1k
GeForce RTX 4080 Super
52.01 images/min / $1k
NVIDIA GeForce RTX 5090
51.21 images/min / $1k
NVIDIA GeForce RTX 3060 Ti
51.18 images/min / $1k
NVIDIA GeForce RTX 3070 Founders Edition
50.4 images/min / $1k

Top 15 shown; 13 more cards in the full table below.

Launch price, not street price, so it ages. A speed leaderboard always crowns the most expensive card; this is the counterweight.

Stable Diffusion 1.5. Measured image speed by GPU

NVIDIA GeForce RTX 5090102.4
NVIDIA H100 80GB HBM380.45
NVIDIA GeForce RTX 409080.14
GeForce RTX 508061.19
NVIDIA L40S58.54
GeForce RTX 4080 Super51.96
NVIDIA GeForce RTX 408051.23
GeForce RTX 5070 Ti49.51
NVIDIA GeForce RTX 4070 Ti Super46.09
NVIDIA GeForce RTX 4070 Ti44.44
NVIDIA GeForce RTX 3090 Ti43.77
NVIDIA RTX PRO 4000 Blackwell40.91
NVIDIA GeForce RTX 3080 Ti40.7
NVIDIA GeForce RTX 4070 Super39.83
NVIDIA RTX 4000 (Ada Generation)39.51
GeForce RTX 507039.2
NVIDIA GeForce RTX 309038.24
NVIDIA GeForce RTX 407035.36
NVIDIA RTX A400029.26
NVIDIA L428.64
NVIDIA GeForce RTX 3070 Ti27.62
NVIDIA GeForce RTX 4060 Ti 16GB26.23
NVIDIA GeForce RTX 3070 Founders Edition25.15
GeForce RTX 5060 Ti25.14
NVIDIA GeForce RTX 506023.9
NVIDIA GeForce RTX 3060 Ti20.42
GeForce RTX 406020.21
NVIDIA GeForce RTX 306015.9
GPUImages/mins per imageimg/W·minAvg power
NVIDIA GeForce RTX 5090102.40.5860.227451.1 W
NVIDIA H100 80GB HBM380.450.750.273294.2 W
NVIDIA GeForce RTX 409080.140.7490.247325.1 W
GeForce RTX 508061.190.9810.222276.2 W
NVIDIA L40S58.541.020.238245.7 W
GeForce RTX 4080 Super51.961.1550.283183.8 W
NVIDIA GeForce RTX 408051.231.1710.237216.4 W
GeForce RTX 5070 Ti49.511.2120.213232.9 W
NVIDIA GeForce RTX 4070 Ti Super46.091.3020.245188.1 W
NVIDIA GeForce RTX 4070 Ti44.441.350.214207.8 W
NVIDIA GeForce RTX 3090 Ti43.771.3710.13337.8 W
NVIDIA RTX PRO 4000 Blackwell40.911.4670.293139.7 W
NVIDIA GeForce RTX 3080 Ti40.71.4740.122334.4 W
NVIDIA GeForce RTX 4070 Super39.831.5060.209190.6 W
NVIDIA RTX 4000 (Ada Generation)39.511.5190.329120.0 W
GeForce RTX 507039.21.5310.216181.4 W
NVIDIA GeForce RTX 309038.241.5690.116329.7 W
NVIDIA GeForce RTX 407035.361.6970.204173.0 W
NVIDIA RTX A400029.262.0510.212138.2 W
NVIDIA L428.642.090.39772.1 W
NVIDIA GeForce RTX 3070 Ti27.622.1720.108256.6 W
NVIDIA GeForce RTX 4060 Ti 16GB26.232.2880.194135.2 W
NVIDIA GeForce RTX 3070 Founders Edition25.152.3850.129194.7 W
GeForce RTX 5060 Ti25.142.3870.30183.5 W
NVIDIA GeForce RTX 506023.92.5110.175136.7 W
NVIDIA GeForce RTX 3060 Ti20.422.9380.108188.3 W
GeForce RTX 406020.212.969——
NVIDIA GeForce RTX 306015.93.7740.108147.3 W

What the numbers show. Across 27 GPUs measured on our own bench, RTX 5090 is fastest at 102 images/min. L4 is the most efficient, 28.6 images/min at 72W (0.397 images/min per watt). Per dollar of launch price, RTX 5060 Ti gives the most (76.4 images/min per $1,000). The fastest card with 16GB or less is RTX 5080 at 61.2 images/min. The measured peak was ~5GB, so it runs on 6GB cards and up. That floor is a property of the model, so it applies to every GPU, measured or not.

About Stable Diffusion 1.5. Stable Diffusion 1.5: from Lykon, 0.9B parameters, on Hugging Face since August 2023, creativeml-openrail-m licence. 67,836 downloads in the last 30 days.

How it compares. L40S: Stable Diffusion 1.5 58.54 images/min, PixArt-Sigma XL 26.95, Stable Diffusion 3.5 Medium 11.58 (2B), Playground v2.5 10.85 (3B), SDXL Turbo 423.5 (3B). 1 of 4 beat Stable Diffusion 1.5 here.

Cost on a rented GPU. 1,000 images of Stable Diffusion 1.5: $0.038 on a RTX 3060 ($0.036/hr, 63 min), $0.063 on a RTX 5090 ($0.39/hr, 10 min, 1.7x the cost).

Stable Diffusion 1.5: cost per 1,000 images on rented GPUs

NVIDIA GeForce RTX 3060$0.036/hr
NVIDIA RTX A4000$0.078/hr
NVIDIA GeForce RTX 3070 Founders Edition$0.075/hr
GeForce RTX 5070 Ti$0.15/hr
NVIDIA GeForce RTX 3090$0.12/hr
GeForce RTX 5080$0.21/hr
NVIDIA GeForce RTX 4070$0.12/hr
NVIDIA GeForce RTX 4070 Ti Super$0.16/hr
NVIDIA GeForce RTX 5060$0.090/hr
NVIDIA GeForce RTX 5090$0.39/hr
NVIDIA GeForce RTX 4080$0.20/hr
NVIDIA GeForce RTX 4070 Ti$0.18/hr
NVIDIA GeForce RTX 4090$0.34/hr
GeForce RTX 4080 Super$0.22/hr
NVIDIA GeForce RTX 3080 Ti$0.18/hr
GeForce RTX 5070$0.19/hr
NVIDIA RTX PRO 4000 Blackwell$0.20/hr
NVIDIA RTX 4000 (Ada Generation)$0.20/hr
GeForce RTX 5060 Ti$0.14/hr
NVIDIA GeForce RTX 3090 Ti$0.27/hr
NVIDIA L40S$0.79/hr
NVIDIA L4$0.44/hr
NVIDIA H100 80GB HBM3$2.14/hr
GPUCheapest rateSpeed (images/min)Cost per 1,000 images
NVIDIA GeForce RTX 3060$0.036/hr15.9$0.038
NVIDIA RTX A4000$0.078/hr29.26$0.044
NVIDIA GeForce RTX 3070 Founders Edition$0.075/hr25.15$0.050
GeForce RTX 5070 Ti$0.15/hr49.51$0.050
NVIDIA GeForce RTX 3090$0.12/hr38.24$0.053
GeForce RTX 5080$0.21/hr61.19$0.057
NVIDIA GeForce RTX 4070$0.12/hr35.36$0.058
NVIDIA GeForce RTX 4070 Ti Super$0.16/hr46.09$0.059
NVIDIA GeForce RTX 5060$0.090/hr23.9$0.063
NVIDIA GeForce RTX 5090$0.39/hr102.4$0.063
NVIDIA GeForce RTX 4080$0.20/hr51.23$0.066
NVIDIA GeForce RTX 4070 Ti$0.18/hr44.44$0.066
NVIDIA GeForce RTX 4090$0.34/hr80.14$0.070
GeForce RTX 4080 Super$0.22/hr51.96$0.071
NVIDIA GeForce RTX 3080 Ti$0.18/hr40.7$0.074
GeForce RTX 5070$0.19/hr39.2$0.081
NVIDIA RTX PRO 4000 Blackwell$0.20/hr40.91$0.083
NVIDIA RTX 4000 (Ada Generation)$0.20/hr39.51$0.084
GeForce RTX 5060 Ti$0.14/hr25.14$0.090
NVIDIA GeForce RTX 3090 Ti$0.27/hr43.77$0.10
NVIDIA L40S$0.79/hr58.54$0.22
NVIDIA L4$0.44/hr28.64$0.26
NVIDIA H100 80GB HBM3$2.14/hr80.45$0.44

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 1.5. 30+ images/min: 18 (RTX 5090, RTX 4090, RTX 5080); 6-30 images/min: 10 (RTX 3070 Ti, RTX 4060 Ti 16GB, RTX 3070 Founders Edition). 30 images/min means two seconds or less per picture.

Time per image. Stable Diffusion 1.5 at 512px and 30 steps: 0.6s per image on the RTX 5090, 3.8s on the RTX 3060. A batch of 100 takes 1 min on the fastest card and 6 min on the slowest.

VRAM for Stable Diffusion 1.5. Measured peak 3.9GB, so 8GB is the smallest common card size; smallest card it ran on: RTX 3070 Ti (8GB).

Power on Stable Diffusion 1.5. Most efficient: L4, 72W, 42.0 Wh per 1,000 images. Hungriest: RTX 5090, 451W, 73.4 Wh.

Our verdict

Fastest on Stable Diffusion 1.5: NVIDIA GeForce RTX 5090, 102.4 images/min. Cheapest consumer card that ran it: NVIDIA GeForce RTX 5060 ($249, 23.9 images/min). Cheapest to rent per job: NVIDIA GeForce RTX 3060, $0.038 per 1,000 images.

FAQ

What GPU do I need to run Stable Diffusion 1.5?
About 4GB. Cheapest consumer card that ran it: NVIDIA GeForce RTX 5060 (8GB, 23.9 images/min).
How fast is Stable Diffusion 1.5 on the NVIDIA GeForce RTX 5090?
102.4 images/min at 451W, faster than every datacenter card we ran it on.
How much does it cost to run Stable Diffusion 1.5 in the cloud?
$0.038 per 1,000 images on a NVIDIA GeForce RTX 3060 at $0.036/hr, cheapest of 23 rentable cards we measured.
Can I run Stable Diffusion 1.5 on a 12GB, 16GB or 24GB card?
It used 3.9GB at the precision we tested. 12GB: yes; 16GB: yes; 24GB: yes.
Is the RTX 4090 or the RTX 3090 faster for Stable Diffusion 1.5?
The RTX 4090: 80.14 vs 38.24 images/min, 110% faster on our bench.
Should I buy or rent a RTX 5090 for Stable Diffusion 1.5?
Its $1,999 launch price buys 5,139 rented hours at $0.39/hr, enough for about 31,561x 1,000 images of Stable Diffusion 1.5. Buy only if you'll run more than that.

How we test

Stable Diffusion 1.5 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.