AMD EPYC Venice beats Intel, NVIDIA in tests

AMD said the flagship Venice part also outperformed its own previous-generation EPYC 9965 by 78%, helped by the move to Zen 6 and a jump from 192 to 256 cores. It also said a 96-core, higher-frequency version of Venice beat NVIDIA’s Vera on single-thread performance by 20%, a reminder that the server race is not only about core count but also about per-core speed and software efficiency.

The comparison comes with caveats. AMD said the tests used different system configurations, different GCC compiler versions and, in NVIDIA’s case, benchmark data compiled from several sources. The company also did not say whether simultaneous multithreading was enabled, leaving room for some of the headline numbers to be scrutinized by rivals and customers.

Still, the announcement adds to the competitive pressure on Intel, which has been losing share in high-end server CPUs, and on NVIDIA, which is pushing deeper into the CPU layer of AI systems alongside its dominant GPUs. For investors, the key question is whether AMD can convert benchmark leadership into socket wins, pricing power and sustained share gains in enterprise and cloud accounts.

AMD shares are already trading well above their 50-day moving average, and the stock’s recent run has pushed technical readings into overbought territory, with the RSI above 80 on the latest data. NVIDIA and Intel remain central to the same AI infrastructure buildout, but benchmark leadership can still influence procurement decisions before the next wave of server refreshes.

The next catalyst is whether AMD translates the Venice numbers into launch timing, customer design wins and volume shipments. If the performance claims hold up in independent testing, AMD could strengthen its case for premium server pricing; if not, the benchmark gap may narrow once competitors release their own next-generation parts.