M5 Ultra vs. DGX Spark: Local Compute Benchmarks
nickbaumann_ · x · 2026-08-26
Comparison of M5 Ultra Mac Studio (512GB), 4x DGX Spark, and 4x AMD Ryzen AI Halo (all priced $15-20k):
- Compute: Spark is fastest; 200GbE interconnects combine compute effectively. Mac is quick on normal prompts but slower on long ones. Halos suffer from 10GbE bottlenecks, performing worse than the Mac.
- Bandwidth: Mac's 512GB unified bus delivers 4x the tokens/sec of Spark/Halo, which are bottlenecked by sequential token passing across boxes.
- Power: Mac consumes less power than a gaming PC.
More from Infra
- NVIDIA Reveals Full AI Stack for Agents at Hot Chips — nvidia · 2026-08-26
- Data Flow is the buzzword at Hot Chips 2026 as cache fades — bookwormengr · 2026-08-26
- Hyperscalers and neoclouds cannot meet inference demand — markjeffrey · 2026-08-26
- Anonymous 1M-context model appears on OpenRouter — jonathanmast · 2026-08-26
- Lighter crypto system achieves >10x performance boost via optimization — gajesh · 2026-08-26
- Antirez on Mac Studio M5 Ultra: Compute/bandwidth insufficient for K3 tasks — antirez · 2026-08-26