Generative Image Workstations
Stable Diffusion, FLUX & ComfyUI workstations — built for sustained GPU load.
Generative image workflows hammer the GPU for hours at a time. Our SD-tuned builds use the cooler, PSU and case airflow needed to hold full power across batch and video diffusion runs.
- Pre-tested with A1111, ComfyUI, Forge
- Optimised cooling for batch generation
- Up to 96GB VRAM (RTX 6000 Pro)
- 5-year warranty
Recommended configurations
Three calibrated starting points. Every component is configurable in our builder.
Most popular
SDXL Creator
SDXL & FLUX up to 1080p batch
£3,199
Starting price
- CPU
- AMD Ryzen 9 9950X
- GPU
- NVIDIA RTX 5090 (32GB)
- RAM
- 64GB DDR5-6000
- Storage
- 2TB Gen5 NVMe
- Best for
- SDXL · FLUX · Forge · ComfyUI
SD Studio
High-res, Stable Video, training
£7,999
Starting price
- CPU
- AMD Threadripper PRO 7965WX
- GPU
- NVIDIA RTX 6000 Pro Blackwell (96GB)
- RAM
- 128GB DDR5 ECC
- Storage
- 4TB Gen5 NVMe + 8TB NVMe
- Best for
- 4K img, video, LoRA training
SD Production
Dual GPU, parallel pipelines
£15,499
Starting price
- CPU
- AMD Threadripper PRO 7995WX
- GPU
- 2× NVIDIA RTX 6000 Pro Blackwell
- RAM
- 256GB DDR5 ECC
- Storage
- 4TB Gen5 + 16TB NVMe
- Best for
- Studio-scale image / video pipelines
Which Stable Diffusion model fits which GPU?
Image-diffusion VRAM use scales with resolution, batch size and base model. The figures below assume default fp16 weights with xFormers / sage-attention enabled.
| Workflow | VRAM at 1024×1024 | VRAM at 2048×2048 (Hi-Res) | Recommended GPU | |
|---|---|---|---|---|
SD 1.5 | ~5GB | ~10GB | RTX 5060 Ti 16GB+ | |
SDXL | ~9GB | ~16GB | RTX 5070 Ti / 5080 | |
FLUX.1 [dev] | ~17GB | ~28GB | RTX 5090 (32GB) | |
Stable Video Diffusion | ~16GB | ~30GB | RTX 5090 / 6000 Pro | |
SDXL LoRA training | ~24GB | ~40GB | RTX 6000 Pro Blackwell |
Numbers from public ComfyUI / A1111 benchmarks at default precision. Lower-precision (fp8 / Q4) and offload modes can cut VRAM by 30–50% at some quality cost.
Why CREATE PCs
Thermals built for batch jobs
Most generative workloads pin the GPU at 100% for an hour or more. We spec coolers and chassis that hold full clocks without throttle.
Storage tiered for model libraries
Gen5 NVMe for the OS and active workflow, secondary NVMe for your weights / LoRA / VAE library — checkpoints can easily exceed 1TB.
Quiet under load
Workstation acoustics matter when you sit next to the box for eight hours. Whisper-quiet builds are an option on every tier.