We’re seeing more and more leaked benchmarks of AMD’s upcoming Strix Point processors. Rebranded as the Ryzen AI 300 lineup, these chips will battle Intel’s Arrow Lake-H and Lunar Lake-MX offerings. Today’s quota includes first-party benchmarks from GPD (courtesy of HXL) showcasing the multi-core and graphics prowess of AMD’s Zen 5 and RDNA 3.5 microarchitecture, respectively. The Ryzen AI 9 HX 370 is powered by 12 cores (4x Zen 5 + 8x Zen 5c) clocked at a peak frequency of 5.1 GHz, and backed by 36 MB of L3 cache.
The SoC has an adjustable TDP of 15-54W, but we’re looking at a 54W sample. Unlike Intel’s upcoming Lunar Lake family, all 12 cores support SMT, bringing the thread count up to 24. It also includes an XDNA 2-based NPU capable of delivering ~50 TOPs of inferencing performance, and 80 TOPs overall.
In Cinebench 2024, the AMD Ryzen AI 9 HX 370 scores 124 and 1525 points in the single and multi-core benchmarks, respectively. That makes it 34% and 13% faster than Intel’s Core Ultra 9 185H in content creation workloads. Concerning AI performance (inferencing), we’re looking at a 2.22x advantage for the Strix Point flagship, likely with a lower power draw.
The Ryzen AI 9 HX 370 features the Radeon 890M integrated graphics. It includes 16 Compute Units or 1024 stream processors (cores) clocked at up to 2.9 GHz. The leaked slide indicates a 3DMark Time Spy score of 4221 points for the Strix Point flagship.
Not only does the Radeon 890M beat the GeForce RTX 2050 (45W), but it comes quite close to the RTX 3050 (50W) while beating its predecessor, the 780M by 51%. Synthetic benchmarks like TimeSpy and Firestrike are highly optimized and often aren’t an exact representation of real-world gaming performance. They perform in the ballpark, so it’s not a useless comparison.
