Sub-second image generation is no longer theoretical. The RTX 5090 generates a complete 512×512 Stable Diffusion 1.5 image in 0.98 seconds — 25 denoising steps, full quality, no shortcuts. We ran the full benchmark suite on GigaGPU to document what peak SD 1.5 performance actually looks like.
61 Images Every Minute
| Metric | Value |
|---|---|
| Iterations/sec | 25.5 it/s |
| Seconds per image | 0.98 sec (25 steps) |
| Images per minute | 61.2 |
| Resolution | 512×512 |
| Sampler | Euler a / DPM++ 2M Karras |
| Performance rating | Excellent |
Benchmark conditions: 25-step generation at 512×512, batch size 1, FP16 precision. Using A1111 WebUI or ComfyUI backend.
Massive VRAM Surplus
| Component | VRAM |
|---|---|
| Model weights | 3.2 GB |
| Sampling buffer | ~0.6 GB |
| Total RTX 5090 VRAM | 32 GB |
| Free headroom | ~28.8 GB |
SD 1.5 uses barely a tenth of the 5090’s memory. The remaining 28.8 GB is overkill for SD 1.5 alone, but it unlocks serious multi-model scenarios: run SD 1.5 alongside Flux.1 (12 GB) for parallel generation, or pair it with a 7B LLM for automated prompt generation. You could even host three distinct image models simultaneously.
Cost at Maximum Throughput
| Cost Metric | Value |
|---|---|
| Server cost | £1.50/hr (£299/mo) |
| Cost per 1K images | £0.41 |
| Images per £1 | 2439 |
Despite being the most expensive card in the range, the 5090’s raw throughput keeps per-image costs competitive at 41p per thousand. For teams generating 50,000+ images daily, the hourly cost is easily justified. See the full benchmark table for side-by-side GPU comparisons.
Is It Worth It for SD 1.5?
Honestly, if SD 1.5 is your only workload, the RTX 5080 at £189/mo delivers better cost-per-image. The 5090 earns its keep when you combine workloads: SD 1.5 generation plus real-time Whisper transcription plus an LLM chatbot, all on one server. That is where the 32 GB VRAM and raw compute become irreplaceable. Read the full analysis in our GPU selection guide.
Quick deploy:
docker run --gpus all -p 7860:7860 ghcr.io/ai-dock/stable-diffusion-webui:latest --sd15Explore: SD hosting guide, all benchmarks.
Deploy SD 1.5 on RTX 5090
Order this exact configuration. UK datacenter, full root access.
Order RTX 5090 Server
