Another Geekbench entry for AMD's Medusa Point APU has been spotted by a well-known hardware leaker, HXL. This time, the submission lists the identifier 100-000001713-33_N on the Plum-MDS1 FP10 platform and is a step up from the previous leak in terms of performance, despite the chip reportedly running at just over 2.0 GHz. What we have there is listed as a Ryzen 9 part with 10 cores and 20 threads, 10 MB of L2 cache and 32 MB of L3; however, the L3 is likely misreported. The core layout is believed to follow a 4 + 6 configuration based on Zen 6. Scores indicate 3,174 single-core and 15,092 multicore, roughly 29% and 22% higher than average Ryzen AI 9 365 results, which is interesting given the low clock speeds in play. Engineering samples at this stage typically run below final frequencies, so the actual performance gap at launch could look quite different.
According to previous leaks, AMD's upcoming Medusa Point lineup could scale up to 22 CPU cores on the flagship Ryzen 9 model by combining a 10-core APU die with a 12-core desktop CCD. Integrated graphics are expected to have eight RDNA 3.5+ (RDNA 4m) compute units, with the lineup remaining unchanged despite the higher CPU core count. Another interesting new detail in this Geekbench submission is that we can see listed the FP16 AVX-VNNI support for a Zen 6 chip, pointing to expanded FP16 instruction handling in AMD's next-generation architecture. This is mostly relevant for AI and machine learning tasks. However, as with any early ES benchmark, take the numbers with a grain of salt. Pretty much everything from clock speeds, firmware, and platform tuning is a work in progress at this point. AMD Medusa Point is expected to arrive sometime next year with a likely initial launch at CES 2027 in January.