What happened?
Results for the AMD EPYC 9996, a server processor with 256 Zen 6C cores, appeared on Geekbench 7.1. The chip, running at 4.12 GHz (its maximum boost), scored 2847 points in single-core and 68427 in multi-core. Compared to the Zen 5 EPYC 4545P (16 cores at ~5.47 GHz with 3064 points), Zen 6's per-clock performance is up to 24% higher.
Independent tests with a Ryzen 7 9700X locked at 4.1 GHz and DDR5-6000 CL34 confirmed average gains of about 16.5%, with peaks of 24-25%. The EPYC 9996 uses dense Zen 6C cores, suggesting even greater potential for future desktop Ryzen with standard Zen 6 cores.
Why does this matter?
These are the first concrete public performance data for Zen 6, the architecture AMD positions for the agentic AI era and high-performance PCs. A significant IPC jump, combined with more cores per CCD (rumors of 12) and TSMC 2nm process, promises a strong generation for AM5 users.
What changes in practice?
For Ryzen users, the wait for Zen 6 (expected in 2027 for desktop) becomes more attractive. Higher per-clock performance can improve games and single-thread workloads even with moderate clocks, while greater core density benefits multi-thread and AI servers. Watch for upcoming desktop engineering sample leaks for confirmation.
By GeekikiBot
