AMD EPYC 9996 Benchmarks Show Unparalleled Performance With 256 Zen 6C Cores Review

AMD announced the general availability of its 6th Gen EPYC server processors, codenamed "Venice" and based on the new "Zen 6" architecture, on November 10, 2026. The company said systems powered by the new CPUs are shipping from leading server OEMs and ODMs, with cloud instances from Microsoft Azure and Oracle Cloud Infrastructure (OCI) expected to be generally available soon. Dan McNamara, senior vice president and general manager for compute and enterprise AI at AMD, stated that "AI is significantly expanding the role of the CPU across every layer of the data center" and that the new portfolio allows customers to "standardize on the proven x86 foundation while choosing a workload-optimized solution."

The flagship EPYC 9996 features 256 cores and operates at up to 5.3 GHz within a 400-watt TDP, doubling the core count from the previous generation's 128 cores at 5.0 GHz. Memory bandwidth increases from 12 channels of DDR5 at 614 GB/sec to 16 channels at 1.6 TB/sec, while I/O bandwidth doubles with the move from PCI-Express Gen 5 to Gen 6. According to AMD, the Venice chips deliver 1.9x the throughput of the prior "Turin" EPYC 9005 processors.

AMD positioned the new processors specifically for the emerging agentic AI market, which The Next Platform reported is forecast to grow nearly 43 percent annually through 2034. Mike Clark, AMD corporate fellow and lead Zen architect, told journalists that while the company "didn't have a foresight that agentic was going to be what it has become," the investment in the Core Complex Die for AI capability "is paying off as agentic has taken off." Madhu Rangarajan, corporate vice president for compute and enterprise AI products, said agentic AI workloads require CPUs for orchestration, policy, security, and code execution, shifting the expected GPU-to-CPU ratio from 1:4–1:8 toward 1:1.

AMD claimed the EPYC 9996 delivers up to 3.7x more performance than an Nvidia Vera CPU-based system and up to 2.2x the performance and 2.0x the performance per dollar of an Intel Xeon 6980P system. For agent sandbox workloads, the 256-core chip offers 3.3x the agents-per-watt over Nvidia Vera and 2.5x over Intel Xeon 6980P. AMD also introduced Helion agentic AI racks in three configurations—Liquid, Air, and Air Inference—with some reaching up to 73,728 threads per rack. Meta is working with AMD as a lead customer validating the platforms for deployment at scale, while Microsoft Azure previewed HDv2, HXv2, and ND MI455X v7 instances, and OCI announced Compute E7 instances powered by the new processors.