Budget
Budget
Starter AI Desktop
$582
12 GB VRAM250W TDP~35dB7 components
Inside This Build
Components
| Category | Component | Price | Rationale | Buy |
|---|---|---|---|---|
| gpu |  NVIDIA GeForce RTX 3060 12GB | $269 | 12GB VRAM at the absolute lowest price. Runs 3B-8B models at Q4-Q5 comfortably. The used market makes this even cheaper. | |
| cpu |  Intel Core i3-12100F | $79 | 4-core budget CPU. Inference is GPU-bound, so CPU matters less. Enough for system tasks while GPU handles AI. | |
| motherboard |  MSI PRO B660M-A WiFi | $89 | Basic B660 mATX board with WiFi. PCIe 4.0 x16 for the GPU. No frills needed. | |
| ram | 16GB DDR4-3200 (2x8GB) | $32 | 16GB is tight but workable for entry-level. Models load to GPU VRAM, not system RAM for inference. | |
| storage | WD500GB WD Blue SN580 NVMe | $35 | 500GB holds 10-15 quantized models. Upgrade later if needed. | |
| psu | EVGAEVGA 500 BR 500W 80+ Bronze | $39 | 500W is sufficient for RTX 3060 (170W TDP). Budget-friendly with decent efficiency. | |
| case | ThermaltakeThermaltake Versa H18 | $39 | Cheap, functional micro-ATX case with decent airflow. Gets the job done. | |
| Total | $582 | |||
Shop Components
Search links — prices and availability vary by retailer.
Buy Used GPU
Prices and availability vary. Inspect hardware before purchasing.
What This Runs
AI models tested on this build's hardware.
| Model | Quant | Speed |
|---|---|---|
| Llama 3.2 3B Instruct | Q8_0 | 80 tok/s |
| Llama 3.2 1B Instruct | Q8_0 | 120 tok/s |
| Phi-4 Mini | Q8_0 | 75 tok/s |
| Llama 3.1 8B Instruct | Q4_K_M | 30 tok/s |
| Mistral 7B Instruct v0.3 | Q4_K_M | 30 tok/s |
| Whisper Large V3 | FP16 | — |
Upgrade Path
Add 16GB more RAM ($32) for system headroom. Swap GPU to RTX 4060 Ti 16GB ($449) for a massive VRAM upgrade. The platform supports it without other changes.
Target Use Cases
ChatCoding
Last updated: 2026-03-15.