AMD's white paper provides granular details on the Venice versus Vera comparison, though it mixes data from different sources and compiler versions. The company tested a down-cored 9996 Venice chip, dropping from 256 cores to 96 for the per-core comparison, and used GCC 16.1 for its own results while Nvidia used GCC 15.2. In a separate memory bandwidth test, AMD's down-cored 9996 was about 18% faster than Vera in the Stream benchmark, with per-core performance about 8% ahead. The benchmarks also show Venice leading in cloud workloads like database, Java, and cryptography, and outperforming Intel's Granite Rapids-AP in HPC tasks.

AMD's Venice chips are positioned as a direct competitor to Nvidia's Vera, which is part of the company's Rubin platform. Nvidia has previously highlighted the Vera Rubin NVL72 system's performance in MLPerf Inference benchmarks, claiming up to 3.7x higher throughput than the previous generation. The competition between AMD and Nvidia in the server market is intensifying as both companies push for higher performance and efficiency in data centre workloads.