A new Geekbench entry for AMD's upcoming 10-core "Medusa Point" APU, identified as 100-000001713-33_N, has surfaced on the Plum-MDS1 FP10 platform. This engineering sample, reportedly running at just over 2.0 GHz, demonstrates improved performance compared to earlier leaks. The chip is listed as a Ryzen 9 with 10 cores and 20 threads, featuring 10 MB of L2 cache. Its core layout is expected to be a 4 + 6 configuration based on the Zen 6 architecture. The benchmark scores achieved are 3,174 in single-core and 15,092 in multicore tests, representing approximately 29% and 22% increases over average Ryzen AI 9 365 results, a notable jump considering the low clock speeds.

These performance gains are significant, especially for an early engineering sample where clock speeds and tuning are still in development. Previous reports suggest the Medusa Point lineup could extend to a flagship Ryzen 9 model with up to 22 CPU cores, achieved by pairing this 10-core APU die with a 12-core desktop CCD. The integrated graphics are anticipated to feature eight RDNA 4m compute units. A new detail revealed by this Geekbench submission is support for FP16 AVX-VNNI, indicating enhanced FP16 instruction handling in Zen 6, which is particularly beneficial for AI and machine learning workloads.

While these benchmark numbers provide an early glimpse into Zen 6 capabilities, it's important to treat them with caution. Final clock speeds, firmware, and platform optimizations are still works in progress. AMD's Medusa Point processors are anticipated to launch next year, with a potential debut at CES 2027 in January. The advancements in core design and instruction support signal AMD's continued push in both general computing and specialized AI acceleration, positioning these chips for a competitive market.