AMD released a white paper detailing performance metrics for its 6th Gen EPYC 9006 Venice server processors. The company claims its 256-core EPYC 9996 chip achieves 2.24 times the platform-level performance of an 88-core NVIDIA Vera system in SPECrate 2026 Integer benchmarks.
The report also highlights comparisons against Intel Xeon 6980P processors. AMD reports performance gains ranging from 2.4x to 3.7x in workloads including Java, OpenSSL, MongoDB, Redis, and NGINX. In HPC tests such as GROMACS, the company claims improvements of up to 3.13x.
AMD notes that a 96-core version of the Venice chip is 1.2 times faster per core than the Vera system. The company clarified that the 2.24x performance figure compares a 512-thread flagship against a 176-thread Vera configuration. Testing used GCC 16.1 for AMD and GCC 15.2 for NVIDIA.
Production of the Venice series is currently underway. Major OEM platforms are scheduled to launch later this year, with cloud providers expected to begin deployments in the same timeframe. Official SPEC results for both Venice and Vera are not yet available for public release.
