AMD is gearing up to launch its new Ryzen AI 300 series “Strix Point” APUs, but the excitement surrounding the even more powerful “Strix Halo” APU is already reaching fever pitch. Benchmarks from Geekbench 5 show a purported engineering sample (ES) of the Strix Halo APU boasting an impressive array of features, including 16 high-performance Zen 5 cores and up to 40 Compute Units based on the upgraded RDNA 3.5 GPU architecture. This powerful combination is set to deliver gaming performance never seen before in an APU, surpassing even the capabilities of the upcoming Strix Point APUs, which will feature 16 or 12 Compute Units based on the same RDNA 3.5 GPU.
The Strix Halo APU packs a punch with its 8 cores and 16 threads of Zen 5 processing power, capable of reaching clock speeds up to 5.36GHz. While the Geekbench 5 log shows the ES APU running at speeds between 3.4GHz and 4.8GHz, falling short of the rated 5.3GHz+, the retail Strix Halo APUs are expected to hit these speeds. This performance leap is further emphasized when compared to the upcoming Ryzen AI 300 series “Strix Point” APUs, which will max out at 5.1GHz. The Strix Halo’s 5.36GHz rating suggests that retail units could reach even higher speeds, potentially hitting 5.4GHz to 5.5GHz. This exceptional performance is further amplified by the inclusion of 40 CUs of RDNA 3.5 GPU power.
The Strix Halo APU’s internal architecture is equally impressive. It houses 32MB of L3 cache and 8MB of L2 cache (4MB L3 + 1MB L2 per Zen 5 core). AMD has adopted a chiplet configuration for the Strix Halo APU, allowing for a maximum of 16 cores and 32 threads, and up to 64MB of L3 cache. Furthermore, the integrated RDNA 3.5 GPU boasts its own 32MB of MALL cache and an impressive 40 Compute Units (significantly more than Strix Point APUs). The APU also supports a massive 128GB of RAM on the AMD FL11 platform, setting the stage for impressive performance once it drops later this year and continues to dominate the high-end APU market in laptops through 2025.
The Strix Halo ES APU has already made a strong impression in benchmark tests, achieving up to 2177 points in the single-core Geekbench 5 run and up to 13,993 points in the multi-core test. These results are particularly impressive for an early engineering sample, surpassing many of the leading laptop processors currently available from both AMD and Intel. The Strix Halo APU’s chiplet-based design, featuring one or two Zen 5 CCDs and a large SoC die, houses an oversized integrated RDNA 3+ GPU and 256-bit LPDDR5X memory controllers, features not found on the cIOD.
With its integrated RDNA 3+ GPU, boasting a powerful 40 Compute Units, translating to 2560 Stream Processors, 80 AI accelerators, 40 Ray accelerators, 160 TMUs, and an undisclosed amount of ROPs, the Strix Halo APU is poised to offer exceptional graphics performance. The GPU operates at clock speeds reaching up to 3.0GHz, further solidifying its position as a formidable competitor to Apple’s M3 Pro and M3 Max SoCs. AMD’s ambition is to achieve comparable CPU and GPU performance levels while maintaining a competitive PCB and power consumption profile.
