AMD launches 256-core Zen 6 EPYC Venice: agent tool work 1.2x-1.6x faster per core
ryanshrout · x · 2026-11-11
AMD launched its sixth-generation EPYC processors, code-named Venice, and Signal65 got early access to test two SKUs: the 256-core EPYC 9996 (compact Zen 6c cores) and the 96-core EPYC 9686F (full Zen 6 cores boosting to 5.3 GHz).
Key results:
- On CPU-bound agent tool workloads (Terminal-Bench 2), both Venice parts ran 22%-60% faster per core than the Zen 5 EPYC 9755
- A preview of the Signal65 PINNACLE benchmark shows core count paying off as agent concurrency climbs: at 128 concurrent agents, median tool time per agent was 83s on the 9686F and 96s on the 9996, both far ahead of Xeon 8559C's 187s
- The Venice lineup spans GPU host nodes (Venice HF/Verano), high-core agent sandboxes (Venice 256c), and general-purpose servers
Testing recorded whole-system wall power in a shared chassis — agent concurrency scaling is becoming a server CPU selling point.
More from Infra
- SpaceX teams with Qualcomm to build AI-native base stations, eyeing first AI carrier — elonmusk · 2026-11-11
- Prism capability extraction: specialists carved from Qwen 27B at <20% memory match full model — tokenbender · 2026-11-11
- Apple reportedly building Harbor app to turn Macs into private AI inference servers with Thunderbolt 5 clustering — BenBajarin · 2026-11-11
- Agent Arena: Gemini 4 Carbon Gets More Cost-Competitive as Tasks Get Harder — rseroter · 2026-11-11
- AMD Zen 6 deep dive: cheaper branch mispredicts and a return to Zen 3/4 integer scheduling — jedisct1 · 2026-11-11
- Together AI SVP Dan Fu on kernels, serving 3T-param models day zero, and self-improving inference — togethercompute · 2026-11-11