$1,818
| Category | Component | Price | Rationale | Buy |
|---|---|---|---|---|
| gpu | $849 | 24GB VRAM at ~$500 less than an RTX 4090. ROCm support has improved significantly. The value play for large model inference. | ||
| cpu | $339 | 8-core AM5 powerhouse. V-Cache helps with system tasks. Great all-around CPU. | ||
| motherboard | $179 | Solid B650 board with WiFi 6E. Full PCIe 4.0 x16 for the GPU. Good VRM for the 7800X3D. | ||
| ram | 32GB DDR5-5600 (2x16GB) | $79 | DDR5 for AM5. 32GB minimum for AI work — models load into system RAM before GPU. | |
| storage | $89 | Fast NVMe for model loading. Samsung reliability. 1TB holds plenty of models. | ||
| psu | $129 | 850W for the 355W TDP RX 7900 XTX. Fully modular. Plenty of headroom. | ||
| case | Fractal DesignFractal Design North | $119 | Beautiful mid-tower with excellent airflow. Fits the 7900 XTX (287mm). Premium without being expensive. | |
| cooler | ThermalrightThermalright Peerless Assassin 120 SE | $35 | Excellent budget tower cooler. Keeps the 7800X3D cool without noise. | |
| Total | $1,818 | |||
Search links — prices and availability vary by retailer.
Prices and availability vary. Inspect hardware before purchasing.
AI models tested on this build's hardware.
| Model | Quant | Speed |
|---|---|---|
| Llama 3.1 8B Instruct | Q8_0 | 60 tok/s |
| Qwen 2.5 Coder 32B Instruct | Q4_K_M | 18 tok/s |
| Gemma 3 27B | Q4_K_M | 15 tok/s |
| Llama 3.1 70B Instruct | Q3_K_M | 4 tok/s |
| Stable Diffusion XL 1.0 | FP16 | — |
| Stable Diffusion 3.5 Large | FP16 | — |
| Whisper Large V3 | FP16 | — |
AMD roadmap leads to RX 8000 series with improved ROCm. Alternatively, keep the 7900 XTX and add a second one when multi-GPU ROCm support matures. The 850W PSU can handle it.
Last updated: 2026-03-15.