Two M3 Ultras linked by Thunderbolt RDMA for local tensor parallelism
antirez · x · 2026-07-20
A first test of DwarfStar distributed tensor parallelism over Thunderbolt RDMA on two M3 Ultra machines.
For DeepSeek V4 Flash Q2, the setup improved prefill throughput from 593 tok/s to 642 tok/s, while decode dropped from 39 tok/s to 33 tok/s. For Q4, prefill rose from 589 tok/s to 677 tok/s (+14.9%), while decode went from 35.5 tok/s on a single M3 Ultra to 31.4 tok/s with tensor parallelism, implying a small decode penalty for the gain in prefill speed.
More from Infra
- China’s AI arms race is increasingly defined by chips, data centers, and open models — BenBajarin · 2026-07-22
- Agent search bottlenecks are now about variance, not raw latency — rohanpaul_ai · 2026-07-22
- Gavin Baker argues Nvidia may be one of open source AI’s biggest supporters — GavinSBaker · 2026-07-22
- AI Power Demand Exposes US Energy Gap, Urging Shift from Scarcity to Abundance — bradneuberg · 2026-07-22
- Gavin Baker says Nvidia’s $630B figure would be system revenue, not all Nvidia’s — GavinSBaker · 2026-07-22
- A Firecracker-based platform says it can host 6,000 AI agents on one 256 GB server — maritime_sh · 2026-07-22