Wednesday, March 18th 2026

AMD "Medusa Point" APU Early Benchmarks Match "Strix Point" at Half the Clock Speed

AMD is preparing to launch its "Medusa Point" APU in early 2027. However, more benchmarks are emerging to showcase what the actual SoC can do as AMD and its OEM partners test the chip. In the latest Geekbench v6 run, AMD's 10-core, 20-thread "Zen 6" chip appeared with the AMD Engineering Sample number 100-000001713-21_N, achieving a 2,300 single-core and 13,002 multicore score while officially running at only a 2.4 GHz base frequency. In real-world operation, it ran within the range of 2.0-2.1 GHz during the benchmark. The most surprising factor is that this "Medusa Point" test system can match a 10-core, 20-thread AMD Ryzen AI 9 365 "Strix Point" APU that operates at more than double the frequency. When comparing the two, "Medusa Point" scores slightly lower in single-core performance, while the multicore score is surprisingly higher.

This phenomenon could be attributed to the fact that the "Zen 6" cores in the "Medusa Point" APU are much better performing and more optimized for the workloads that Geekbench tests. The IPC improvement target from "Zen 5" in "Strix Point" to the newest "Zen 6" could be a high single-digit to low double-digit gain on average. It is likely that the combination of new instructions and IPC improvements is what is pushing "Medusa Point" so high. Since the new APU also appeared in firmware running AVX-VNNI in FP16 precision, we might be seeing these workloads getting accelerated thanks to the lower precision of the floating-point operations. For now, the situation remains a mystery, at least until more benchmarks are available in the coming months. AMD is expected to launch this new APU around CES 2027, so we still have a lot of time before official and third-party benchmarks are released.
Sources: Geekbench, via Wccftech
Add your own comment

6 Comments on AMD "Medusa Point" APU Early Benchmarks Match "Strix Point" at Half the Clock Speed

#1
JohH
Clock speed is most probably misreported.
Also that's an exceptionally low score for a Ryzen 9 365.
Posted on Reply
#2
R0H1T
Yeah that's (nearly) impossible, besides GB is notoriously bad when it comes to clock speeds :shadedshu:
Posted on Reply
#3
maskbit
You have to understand that initial silicon (A0) is probably ~12 months before release and during bring up you don't run these silicon parts of Fmax. You actually very slowly bring it up bits and parts. This basically means they done weeks/months of bring up to get into OS and are able to run complex workloads. All of these potential leaked number means nothing for what will be the final product. So they have a good start and they will continue bringing it up, making any required workarounds for discovered issues/erratas, and figure out if they need A1 or B0 to resolve the current round of issues.
Posted on Reply
#4
A&P211
It will launch at CES but wont see any release for 6months after CES.
Posted on Reply
#5
dismuter
If they want to catch up with Apple's M chips in single core performance, it needs to well into the double digit IPC percentage improvements. More like around 50%.
Posted on Reply
#6
Imsochobo
JohHClock speed is most probably misreported.
Also that's an exceptionally low score for a Ryzen 9 365.
default fedora linux behavior.
here is with a simple fix, amd_pstate active cmdline option

browser.geekbench.com/v6/cpu/compare/17155028?baseline=17136190

Kinda funny, cause I think the geekbench result used for comparison might be mine... they match up timewise lol
Posted on Reply
Nov 12th, 2026 04:42 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts