# Dual-GPU 48 GB

> Illustrative build. Two used 3090s split a 70B model by layers. Capacity doubles; single-stream bandwidth does not.

Canonical: https://thelocalbench.com/builds/dual-gpu-48gb · Data as of 2026-10-12 · The Local Bench (thelocalbench.com), data as of 2026-10-12, CC BY 4.0

Flags: [verified] checked against primary source · [sourced] cited · [estimate] formula · [sample] placeholder

Total: $4,044 [sample] · Load 780 W · Idle 110 W · Electricity $23.94/month

## Parts

| Part | Item | USD | Flag |
|---|---|---|---|
| GPU | 2× GeForce RTX 3090 24 GB (used) | 2400 | sample |
| CPU | Ryzen 9 7900 | 349 | sample |
| Motherboard | X670E ATX (x8/x8) | 299 | sample |
| Memory | 128 GB DDR5-5600 (4×32) | 359 | sample |
| Storage | 4 TB NVMe Gen4 | 229 | sample |
| PSU | 1 | 249 | sample |
| Case | Full tower | 159 | sample |

## Expected speed

- Llama 3.3 70B Instruct: 12.4 tok/s [estimate]
- Qwen3 32B: 25.6 tok/s [estimate]
- Gemma 3 27B IT: 31.8 tok/s [estimate]