Dual Radeon AI Pro R9700 vs. Two Used RTX 3090s at $1600 Each for Local LLM Inference
Current-Ticket4214 · reddit · 2026-09-15
A Redditor is building a dual-GPU box targeting 30B at FP8 or 70B at Q4. The original plan was two AMD Radeon AI Pro R9700s, but they found a pair of new-in-box RTX 3090s nearby at $1600 each with a discount for buying both. RTX 40/50 series and the RTX PRO 6000 are off the table due to budget. They currently run a Hermes agent on a 7900 XT AM4 box alongside an M1 Max Studio with 64GB RAM, and aim to cancel most frontier subscriptions—keeping only one for coding. The thread weighs AMD vs. NVIDIA dual-card setups for local LLM inference.
More from Infra
- Leaker Claims Nvidia RTX Rubin 6090 Launching Next Year — max_paperclips · 2026-09-15
- Pareta routes cheap LLM tasks to small models, 620x cheaper than GPT-5.5 — D33B · 2026-09-15
- Prefill and Decode: why asking an LLM for three takeaways from a long document still takes minutes — dotey · 2026-09-15
- Training from scratch on a single H100 hits 76% on ARC-AGI-1 in ~4 hours — GregKamradt · 2026-09-15
- From Ollama to vLLM: a roadmap for scaling LLM deployment — kalyan_kpl · 2026-09-15
- Kimi K3 is live and free on NVIDIA NIM with OpenAI-compatible API — airesearch12 · 2026-09-15