NVIDIA RTX 4090 — AI Workstation Build Guide

A complete, opinionated build guide for an RTX 4090-based AI development workstation — from component selection to having Ollama and Stable Diffusion running.

**What's covered:**

**Part 1 — Component selection**

- RTX 4090 variant comparison (which cooler to buy)

- CPU pairing (Intel vs AMD for AI workloads)

- RAM configuration (why 128GB matters for large models)

- NVMe storage recommendations

- PSU sizing (850W is not enough — here's why)

**Part 2 — Assembly and BIOS**

- Step-by-step assembly with photos

- BIOS settings that matter for AI workloads

- PCIe 4.0 vs 5.0 — does it affect inference speed?

**Part 3 — Software stack**

- Ubuntu 24.04 or Windows 11 — pros and cons

- NVIDIA driver installation (avoid the common mistakes)

- CUDA, cuDNN, PyTorch setup

- Getting Ollama, Stable Diffusion, and ComfyUI running in under an hour

**Total build cost at time of writing: ~£3,200**