Geekbench listing suggests up to 24% higher per-clock performance for AMD Zen 6, while AMD's HIPRT 3.2 lists Medusa GPU IDs

A Geekbench database entry has prompted a claim that AMD's Zen 6 core delivers up to 24% higher per-clock performance than Zen 5. The figure did not come from AMD: Wccftech derived it, writing "If we do the math's, the per-clock performance of the AMD Zen 6 chip is up to 24% higher than the Zen 5 chip," and the article cites no AMD statement. AMD (AMD) has not confirmed it in any source reviewed.

The entry is a 256-core, 512-thread EPYC 9996 that scored 2847 in single-core tests at 4.12 GHz, its maximum boost frequency. Wccftech compared it with a 16-core Zen 5 EPYC 4545P that scores 3064 at a boost clock of up to 5.5 GHz. The Zen 6 chip's raw score is lower, so the 24% depends on scaling scores by clock speed, and the article does not show its calculation. A separate X post cited by Wccftech compared the chip with a Ryzen 7 9700X and found about 16.5% higher at roughly 4.1 GHz. Overclock3D headlined the result as a Zen 6c part; "6c" appears to denote the compact Zen 6c core, not a six-core chip.

Separately, AMD published HIPRT 3.2, an update to its ray-tracing library labelled a pre-release. According to Geeknetic, its notes add support for gfx1170, gfx1171 and gfx115E, grouped under Medusa, a name tied to a future APU family. These are graphics targets in software, not a product announcement.

Leaked database entries can differ from retail silicon. Watch for AMD's own performance claims, the final Medusa specifications and launch timing, and independent testing of production chips.