Chroma1-HD · 3 GPUs measured first-party · text-to-image · Updated October 2026
What GPU Do You Need for Chroma1-HD?
Chroma1-HD on 3 GPUs, measured first-party: NVIDIA GeForce RTX 5090 leads at 1.29 images/min, RTX A6000 trails at 0.62 images/min, and it peaked at 33GB of VRAM.
Benchmarked weights: lodestones/Chroma1-HD
NVIDIA GeForce RTX 5090
1.29 images/min on Chroma1-HD, the ceiling. Measured on our bench. 32GB of VRAM, $1,999 at launch.
- 1.29 images/min on Chroma1-HD
- 32GB, clears the Chroma1-HD floor
- 575W board rating
NVIDIA RTX A6000
0.62 images/min on Chroma1-HD, lowest launch price that still fits. Measured on our bench. 48GB of VRAM, $4,650 at launch.
- 0.62 images/min on Chroma1-HD
- 48GB, clears the Chroma1-HD floor
- Rentable by the hour rather than bought
- 300W board rating
- Datacenter or workstation hardware, not a retail purchase
What GPU Do You Need for Chroma1-HD?, 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.
Chroma1-HD. Measured image generation speed by GPU
| GPU | Images/min | s per image | img/W·min | Avg power |
|---|---|---|---|---|
| NVIDIA GeForce RTX 5090 | 1.29 | 46.611 | 0.002 | 520.7 W |
| NVIDIA A100 40GB PCIe | 1.07 | 56.01 | 0.004 | 248.4 W |
| NVIDIA RTX A6000 | 0.62 | 96.577 | 0.002 | 298.9 W |
What the numbers show. Across 3 GPUs measured on our own bench, RTX 5090 is fastest at 1.29 images/min. The slowest, RTX A6000, manages 0.62, so the spread is 2.1x from top to bottom. A100 40GB PCIe is the most efficient, 1.07 images/min at 248W.
About Chroma1-HD. Chroma1-HD: from lodestones, 8.9B parameters, on Hugging Face since August 2025, Apache 2.0 licence. 61,943 downloads in the last 30 days.
How it compares. RTX 5090: Chroma1-HD 1.29 images/min, Z-Image 1.72 (6B), FLUX.1 dev 2.04 (12B), FLUX.2 klein 4B 59.33 (4B), Playground v2.5 13.28 (3B). All 4 beat Chroma1-HD here.
Cost on a rented GPU. 1,000 images of Chroma1-HD: $5.03 on a RTX 5090 ($0.39/hr, 12.9 hours).
Chroma1-HD: cost per 1,000 images on rented GPUs
| GPU | Cheapest rate | Speed (images/min) | Cost per 1,000 images |
|---|---|---|---|
| NVIDIA GeForce RTX 5090 | $0.39/hr | 1.29 | $5.03 |
| NVIDIA A100 40GB PCIe | $0.45/hr | 1.07 | $6.99 |
| NVIDIA RTX A6000 | $0.33/hr | 0.62 | $8.87 |
Cheapest hourly rate we track on RunPod and Vast.ai, divided by the measured speed. Startup time and storage are extra.
Speed tiers for Chroma1-HD. under 6 images/min: 3 (RTX 5090). 30 images/min means two seconds or less per picture.
Time per image. Chroma1-HD at 1024px and 40 steps: 46.5s per image on the RTX 5090, 96.8s on the RTX A6000. A batch of 100 takes 78 min on the fastest card and 161 min on the slowest.
VRAM for Chroma1-HD. Measured peak 30.8GB, so 48GB is the smallest common card size; smallest card it ran on: RTX 5090 (32GB).
Power on Chroma1-HD. Most efficient: A100 40GB PCIe, 248W, 3.87 kWh per 1,000 images. Hungriest: RTX 5090, 521W, 6.73 kWh. At $0.15/kWh: $0.58 per 1,000 images.
Our verdict
Fastest on Chroma1-HD: NVIDIA GeForce RTX 5090, 1.29 images/min. Cheapest to rent per job: NVIDIA GeForce RTX 5090, $5.03 per 1,000 images.
FAQ
- What GPU do I need to run Chroma1-HD?
- About 32GB. Cheapest consumer card that ran it: NVIDIA GeForce RTX 5090 (32GB, 1.29 images/min).
- How fast is Chroma1-HD on the NVIDIA GeForce RTX 5090?
- 1.29 images/min at 521W, faster than every datacenter card we ran it on.
- How much does it cost to run Chroma1-HD in the cloud?
- $5.03 per 1,000 images on a NVIDIA GeForce RTX 5090 at $0.39/hr, cheapest of 3 rentable cards we measured.
- Can I run Chroma1-HD on a 12GB, 16GB or 24GB card?
- It used 30.8GB at the precision we tested. 12GB: no; 16GB: no; 24GB: no.
- Should I buy or rent a RTX 5090 for Chroma1-HD?
- Its $1,999 launch price buys 5,139 rented hours at $0.39/hr, enough for about 398x 1,000 images of Chroma1-HD. Buy only if you'll run more than that.
How we test
Chroma1-HD 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.