AMD details EPYC Venice benchmark claims against Nvidia Vera

AMD expanded its performance claims for EPYC Venice with a new white paper covering server, cloud, HPC, memory bandwidth, and agentic AI workloads. The company still says a 96-core, high-frequency Venice chip is around 20% faster than Nvidia’s 88-core Vera in SPEC CPU 2026’s Integer Rate test, while broader SPEC CPU 2026 intrate results put the 256-core 9996 at 2.37x faster than the Intel Xeon 6980P and 2.24x faster than Vera.

The strongest comparison may be AMD’s own generational uplift. AMD’s data shows the 9996 is around 78% faster than last-gen’s 192-core EPYC 9965, reinforcing Venice as a major server CPU update even before competitive claims are weighed.

Several benchmark caveats matter. AMD tested a down-cored EPYC 9996, dropping from 256 cores to 96 cores, and the chip appears to use a 600W power budget, while AMD’s 96-core, high-frequency Venice SKU tops out at 500W. The Vera comparison also mixes compiler data, with AMD using GCC 16.1 against Nvidia results produced with GCC 15.2.

AMD also cited Stream memory bandwidth results where Venice is ahead by about 18%, with per-core performance about 8% ahead. Cloud, HPC, and agentic AI tests further extend AMD’s positioning, though server performance remains highly dependent on software, build choices, and deployment needs.