Skip to main content
AMD 6th Gen EPYC Processors Now Available With Zen 6 Architecture For Agentic AI & Enterprise Workloads - Pokde.Net

AMD 6th Gen EPYC Processors Now Available With Zen 6 Architecture For Agentic AI & Enterprise Workloads - Pokde.Net

AMD Processors
12 lis 2026, 15:01Hardware News
Oryginał
1 min read

AMD has announced the general availability of its 6th Gen EPYC server processors based on the new Zen 6 architecture, with systems now shipping worldwide.

The 6th Gen EPYC portfolio is designed to address expanding compute demands across agentic AI deployments, cloud infrastructure, high-performance computing, and critical enterprise applications. Cloud instances from Microsoft Azure and Oracle Cloud Infrastructure are expected to become available soon.

Performance Claims

AMD's flagship EPYC 9996 CPU delivers up to 3.7x more performance compared to an NVIDIA Vera CPU-based system and up to 2.2x the performance of an Intel Xeon 6980P, according to SPECrate 2017 integer benchmark results. For GPU host node workloads, high-frequency processors reaching up to 5.3GHz provide up to 1.8x the LLM output throughput compared to the previous 5th Gen EPYC processors.

Embedded Solutions

AMD also introduced the EPYC Embedded 9006 Series SP7 processors, extending the Zen 6 architecture to high-performance embedded applications. These processors offer up to 256 cores and 512 threads, delivering up to 1.8x higher throughput compared to the previous 5th Gen EPYC 9005 Series.

The embedded processors support 16 DDR5 memory channels with RDIMM speeds up to 8,000 MT/s and MRDIMM speeds up to 12,800 MT/s. They also feature up to 96 PCIe 6.0 lanes operating at 64 GT/s, with support for high-speed connectivity to various devices.

Helion Platform

AMD introduced the Helion agentic AI rack platform alongside the processor launch. The Helion family comes in three configurations: Helion Liquid for large-scale liquid-cooled deployments, Helion Air for air-cooled data centers, and Helion Air Inference for combined CPU execution and local GPU inference. Some configurations can reach up to 73,728 threads per rack.

Z.ai
Article written by AI
Z.ai GLM-4.5-FlashFree
Free AI articles150 of 150
No limit after signing in with Telegram