Dual R9700 local inference: how much does PCIe Gen4 actually cost vs Gen5?
IngwiePhoenix · reddit · 2026-09-30
A builder who waited over two years before buying two AMD Radeon AI Pro R9700 GPUs (64GB VRAM total) plans to run llama.cpp with ROCm behind llama-swap for dual-GPU inference and VRAM pooling.
The open question: since DDR5 RDIMMs are prohibitively expensive, would a cheaper DDR4 + PCIe Gen4 platform significantly hurt inference performance compared to Gen5? The post is a help-seeking thread, useful for anyone planning budget local inference rigs.
More from Infra
- Dumping GPUs and tokens can meaningfully speed up AI development — menhguin · 2026-09-30
- 9B Open-Weight Model Drops Agent Accuracy From 96% to 62.3% — TheZachMueller · 2026-09-30
- Jensen Huang: AI data centers add 10-20GW a year and roughly a million jobs — victor_explore · 2026-09-30
- First SGLang Summit set for Nov 12-13 in SF, with Intel CEO and Lilian Weng speaking — BanghuaZ · 2026-09-30
- Interview with Richard Ho, leader of OpenAI's in-house chip project — bookwormengr · 2026-09-30
- Tenstorrent opens bio-model training: OpenFold3 on Blackhole Galaxy nears DGX H200 at quarter the cost — MoAlQuraishi · 2026-09-30