AMD Medusa Point: Zen 6 Benchmarks Show Doubled Scores & Strix Point Rivalry

AMD’s next-generation mobile processors, codenamed Medusa Point, are showing promising early performance in Geekbench 6 testing. Recent results indicate the 10-core, 20-thread Zen 6 APU is achieving multi-core scores comparable to the Ryzen AI 9 365 “Strix Point” APU, despite operating at significantly lower clock speeds – around 2.1 GHz versus Strix Point’s 2.4 GHz and higher. This suggests substantial improvements in instructions per clock (IPC) with the Zen 6 architecture, potentially leading to more efficient performance and improved battery life in future laptops.

A Second Look at Medusa Point: Divergent Benchmarks

The latest Geekbench 6 run, using AMD Engineering Sample 100-000001713-21_N, yielded scores of 2,300 in single-core and 13,002 in multi-core. This represents a significant jump from a previous Medusa Point sighting, which scored 1,210 and 7,323 respectively under similar conditions. PauseHardware notes this wide variance raises questions about the consistency of pre-release silicon and the true potential of Zen 6’s IPC gains. The chip features a 4+6 core configuration, combining standard and efficiency-focused Zen 6 cores.

The performance difference between the two Medusa Point runs highlights the challenges of interpreting early benchmark data. Factors like firmware versions, memory configurations, and power limits can all significantly impact results. However, the higher scores in the recent test are encouraging, particularly when compared to the Ryzen AI 9 365.

Zen 6 and the Efficiency Equation

The fact that Medusa Point can match Strix Point’s multi-core performance at a lower clock speed is a key indicator of improved efficiency. This suggests AMD has made significant strides in the underlying microarchitecture of Zen 6, optimizing both the front-end (instruction fetching and decoding) and the back-end (execution and completion). TechPowerUp points to better utilization of the chip’s vector units, particularly in workloads leveraging AVX-VNNI and FP16 instructions.

AVX-VNNI (Advanced Vector Extensions 512 Neural Network Instructions) is a set of instructions designed to accelerate deep learning and AI workloads. The presence of FP16 (half-precision floating-point) support in the Medusa Point sample, as reported by Geekbench, is a notable addition. This capability allows the APU to perform calculations with reduced precision, which can significantly speed up certain tasks, especially those related to machine learning inference. VideoCardz details this as a broadening of low-precision compute support on Zen 6 mobile parts.

Implications for Mobile Performance

Increased efficiency translates directly into benefits for mobile devices. Lower power consumption means longer battery life, and the ability to maintain high performance within tighter thermal constraints. This is particularly significant for thin and light laptops, where cooling solutions are often limited. If Zen 6 delivers on its promise of improved performance-per-watt, we could see a new generation of mobile processors that offer both power and portability.

Benchmarking Context: Medusa Point vs. Strix Point

The Ryzen AI 9 365 (Strix Point) is AMD’s current-generation mobile APU, featuring Zen 4 CPU cores and RDNA 3 graphics. It’s designed for mainstream laptops and offers a balance of performance and efficiency. The fact that Medusa Point can compete with Strix Point at a lower clock speed suggests that Zen 6 represents a significant architectural improvement. However, it’s important to remember that these are early benchmarks, and real-world performance will depend on a variety of factors, including software optimization and system configuration.

The Medusa Point APU tested includes 32MB of L3 cache and 10MB of L2 cache. Cache size plays a crucial role in performance, as it allows the processor to quickly access frequently used data. Larger caches can reduce the need to access slower system memory, resulting in faster overall performance.

The Road Ahead: From Engineering Samples to Launch

AMD is currently targeting a launch for Medusa Point around the Consumer Electronics Show (CES) in 2027. Between now and then, there will be extensive testing and refinement of the architecture. Clock speeds, firmware, and power limits are all subject to change. The initial Geekbench results provide a glimpse of the potential of Zen 6, but they are not a definitive indication of final performance.

As Wccftech points out, the discrepancies between the two Medusa Point runs underscore the preliminary nature of these tests. Multiple data points will be needed before drawing firm conclusions about Zen 6’s IPC.

The development process will involve rigorous validation and optimization to ensure stability and reliability. AMD will too work with its partners to integrate Medusa Point into a wide range of laptops and other mobile devices. The ultimate success of Zen 6 will depend on its ability to deliver a compelling combination of performance, efficiency, and features.