AMD Unveils Ryzen AI 300 Series Processors at Computex 2024

At the Computex 2024 event, AMD launched its latest generation of Ryzen laptop processors, specifically designed to enhance generative AI workloads. The newly introduced Ryzen AI 300 series signifies a rebranding of AMD’s high-end Ryzen 9 chips, continuing to use the HX suffix to highlight top-tier performance rather than power consumption.

This series positions AMD against competitors like Qualcomm’s ARM-based Snapdragon X Elite and Intel’s x86 Lunar Lake chips.

Technical Details of the Ryzen AI 300 Series

The new Ryzen AI processors utilize AMD’s advanced architectures for neural processing, integrated graphics, and general processing tasks. These include the XDNA2 for the neural processing unit (NPU), RDNA 3.5 for the integrated GPU (iGPU) featuring up to 16 compute units, and Zen 5 for the CPU.

AMD has initially launched the Ryzen AI 9 HX 370 and Ryzen AI 9 365 models in this series. Both models are equipped with an NPU capable of 50 tera operations per second (TOPS), with the HX variant being the higher-end model in terms of performance.

The Ryzen AI 9 HX 370 model features a 12-core/24-thread configuration, can reach a maximum clock speed of 5.1GHz, includes a 36MB cache, and is paired with Radeon 890M graphics. On the other hand, the Ryzen AI 9 365 model comes with a 10-core/20-thread setup, a slightly lower maximum clock speed of 5.0GHz, a 34MB cache, and Radeon 880M graphics.

Competitive chipsets, and how Ryzen AI 300 series compared to its predecessors?

Notably, the Ryzen AI 300 series distinguishes itself with the highest TOPS among chips configured with NPUs. For comparison, Qualcomm’s Snapdragon X series NPU achieves 45 TOPS, Apple’s M4 reaches 38 TOPS, and Intel claims that their upcoming Lunar Lake series will exceed 40 TOPS.

The previous AMD Ryzen 8040 series could only manage 16 TOPS, underscoring the significant performance leap represented by the Ryzen AI 300 series.

However, while TOPS is a critical performance metric, it is not the only measure of a system’s capability to handle AI tasks. Factors such as architectural efficiency, software optimization, and the specific characteristics of the workload also play essential roles.

According to AMD, the XDNA2 NPU architecture offers five times more compute capacity and double the power efficiency compared to its predecessor. This improvement is attributed to the architecture's unique FP16 "block" design, which handles both 8-bit performance (INT8) and 16-bit accuracy (FP16) generative AI workloads without needing quantization. Quantization, a technique used by competitors like Qualcomm, can enhance power efficiency but often at the expense of AI model accuracy. AMD's new architecture aims to deliver rapid and precise AI processing without such trade-offs.

Market Availability

The Ryzen AI 300 series is set to enter the market, with shipments beginning in July 2024. These processors will be integrated into select Copilot Plus PCs, which were recently announced at Microsoft’s Surface event. Among the initial devices to feature the Ryzen AI 300 series are the Asus Vivobook S15 and HP OmniBook.