At CES, AMD unveiled an expansion of its Strix Halo lineup by introducing two new Ryzen AI MAX+ processors: the MAX+ 392 and MAX+ 388, categorized under 12-core and 8-core sizes, respectively. Notably, these processors feature a significant upgrade in graphics configuration, both supporting full Radeon 8060S integrated graphics, which comprises 40 RDNA 3.5 compute units. This enhancement, previously exclusive to the top-tier MAX+ 395, is now accessible in the more energy-efficient models.

Strix Halo distinctly separates itself from conventional mobile APUs by focusing on "SoC-level performance density." It integrates CPU, GPU, and NPU within a shared memory pool of up to 128GB of LPDDR5X and employs a coherent interconnect to minimize latency and power drain during data transfers. This innovative approach is gradually being adopted by OEMs across various non-traditional devices like mini-consoles and handheld gadgets, elevating practical performance without merely increasing frequencies or core count in a lower TDP framework.

Traditionally, GPU capacity dictated the performance hierarchy within the Strix Halo series. The MAX+ 392 and MAX+ 388 redefine this spectrum by incorporating the full 8060S iGPU, previously limited to the high-end MAX+ 395, thereby broadening access to superior integrated graphics.

Examining CPU specifications reveals that the MAX+ 392 boasts a 12-core, 24-thread Zen 5 architecture, while the MAX+ 388 features an 8-core, 16-thread design. Both models can reach up to 5.0GHz acceleration and offer 50 TOPS NPU performance, emphasizing differences primarily in core count and cache size rather than frequency. Additionally, the memory subsystem has been revamped, with these processors now supporting 8533 MT/s LPDDR5X, up from the previous 8000 MT/s, translating into enhanced graphics and mixed-load performance.

The full enablement of 40 CUs extends beyond gaming, with the Radeon 8060S delivering nearly 60 TFLOPS at FP16. This provides the MAX+ lineup with a competitive edge in local AI inference, content creation, and select rendering tasks. Comparative benchmarks illustrate notable throughput advantages for the MAX+ 395 over NVIDIA’s DGX Spark across various LLM models, while the MAX+ 392 and 388 outperform Intel’s Core Ultra 9 285H and Ultra 7 255H, respectively, in applications like Blender, V-Ray, 3DMark, and several games.

The MAX+ 388 exemplifies this potential; paired with the full 8060S GPU, it exceeds the capabilities of traditional 8-core processors in gaming and graphics-heavy tasks, a perfect match for handhelds and compact consoles. Its configuration ensures maximum GPU resource utilization, maintained within a flexible power consumption range of 45-120W.

With the introduction of the MAX+ 392 and 388, AMD fills a previous gap in their lineup, making advanced iGPUs more accessible without necessitating the highest core counts or prices, enabling OEMs to create more balanced product offerings.

Accompanying Strix Halo is the mainstream mobile-focused Ryzen AI 400 "Gorgon Point." While this generation maintains the use of Zen 5, RDNA 3.5, and XDNA 2 architectures, enhancements focus on frequency, memory support, and SKU combinations, including CPU speeds up to 5.2GHz and 60 TOPS NPU performance, along with improved memory capabilities and ROCm support.

The simultaneous release of Strix Halo and Gorgon Point showcases AMD’s dual strategy in mobile SoCs: striving for SoC-level performance density for newer form factors with substantial iGPUs and a shared memory vision, while refining frequency, energy efficiency, and AI prowess in mainstream devices. The emergence of the MAX+ 392 and 388 solidifies the former strategy as market-viable and consumer-friendly.