Mixing 3x RTX 5090 with AMD GPUs for DeepSeek: A Local Rig Experiment
fluffywuffie90210 · reddit · 2026-08-10
An AI hardware enthusiast shares their experience expanding VRAM by integrating an AMD GPU into a 3x RTX 5090 system.
- Hardware Setup: Plans to use a PCIe riser cable to add an AMD 9700 AI Pro, aiming for a 128GB VRAM dream machine specifically to run massive DeepSeek models locally.
- Technical Bottlenecks: Highlights compatibility issues mixing CUDA and ROCm. While workarounds exist using an RPC server with llama.cpp, MoE (Mixture of Experts) models suffer terrible acceleration over multi-card RPC setups—only dense models see good speedups.
- Cost Trade-offs: The author is torn between spending heavily to maximize VRAM or settling for a cheaper 5070Ti for gaming, turning to the community for advice.
More from Infra
- MLCommons launches MLPerf Endpoints, a benchmark that measures inference endpoints the way buyers actually purchase them — TheKanter · 2026-10-16
- Neocloud M&A decoded: CoreWeave, Nebius, Crusoe, Nscale all bet margins vanish beyond GPU scarcity — AccBalanced · 2026-10-16
- AI-Driven Timing-Aware Power Optimization Uses Learned Ranking With Baseline-Preserving Repair — Amit Sharma · 2026-10-16
- Nscale AI datacenter plagued by coolant leaks, transformer fire, 10 safety violations — saranormous · 2026-10-16
- WSL 3.0 ships with native containers and Linux 6.18 kernel, syscall overhead down ~16% — solyarisoftware · 2026-10-16
- Are 22 GB modded RTX 2080 Ti cards a viable budget local LLM option? — Dave86ch · 2026-10-16