AMD MI355X sustains 2x the concurrent agents of MI300X in Signal65 PINNACLE tests, up to 6.5x on MiniMax-M3
ryanshrout · x · 2026-09-19
- Signal65's PINNACLE GPU benchmark added the AMD Instinct MI355X for the first time, comparing it to MI300X on identical fully loaded 8-GPU nodes under the same agentic workload and service level (every agent decoding at 10 tokens/s or better, 95% first token within 60 seconds).
- Results: MI355X sustains a median 2x the concurrent agents of MI300X across seven models, and up to 6.5x on MiniMax-M3; it is faster at every shared load, roughly 2x tokens on Qwen3.5-35B-A3B and 1.8x on Qwen3.5-397B-A17B.
- New initial results include Kimi-K3, GLM-5.2 and an MXFP8 build of MiniMax-M3; the board now covers 12 models across 4 GPU platforms, with numbers to be updated as serving stacks are tuned.
Related event: Signal65 tests show MI355X far ahead of MI300X in concurrent AI agents(2 posts)→
More from Infra
- Why Jev-class models could become a near-free judgment primitive running on-device — signulll · 2026-09-19
- Pedro Domingos: Opposing Data Centers Means Keeping Your Country Stupid — pmddomingos · 2026-09-19
- Pedro Domingos: OpenAI and Anthropic's real moat is their massive secured compute — pmddomingos · 2026-09-19
- Fully Local, Private Computer-Use AI Arrives — Runs on a 3060 Ti 12GB — WolframRvnwlf · 2026-09-19
- AWS ships 13 SageMaker inference updates in 2026, cutting cold-start latency up to 65% — AWS ML Blog · 2026-09-19
- Dev Runs World Model LingBot-World 2.0 at 16 FPS on One RTX 5090, 2.5x Faster Than SGLang — Kaarel_Kaarelson · 2026-09-19