In addition to the Ryzen 9000 desktop processors and specialized AI computers, AMD also presented a line of mobile chips Ryzen AI 300 (Strix Point) at Computex 2024. Even the name of the new series already emphasizes the special capabilities of processors. However, almost all components of these models deserve additional attention. It seems that AMD developers have decided to combine their most technologically advanced solutions in one processor.

Ryzen AI 300 architectures

This time we are dealing with single-crystal chips (TSMC 4 nm) equipped with computing cores with the latest Zen 5 architecture. Strix Point also uses powerful integrated graphics based on RDNA 3.5 and an NPU unit with XDNA 2 architecture for AI acceleration.

Ryzen AI 300 series

Initially, the Ryzen AI 300 chip lineup includes only two models – the Ryzen AI 9 HX 370 and the Ryzen AI 9 365. The flagship of the series has a 12-core configuration with support for SMT multithreading technology. It uses 4 “large cores” with Zen 5 architecture and 8 “compact” cores on Zen 5c. It is worth reminding that in the case of AMD chips, such a hybrid layout really means a difference in the size of the computers, while their functionality is identical.

More compact cores allow for a smaller chip area and slightly higher power efficiency while operating the CPU with relatively low power consumption (10-15 watts). The exact frequency formula is not yet known, except that the “large” cores can accelerate to 5.1 GHz and have a base of 2.0 GHz. For sure, Zen 5c cores will operate at slightly lower frequencies.

AMD introduces Ryzen AI 300 mobile processors: Zen 5 architecture, RDNA 3.5 graphics and 50 TOPS NPU

The cache configuration is 1 MB of L2 for each core plus 24 MB of total L3. So in the case of Ryzen AI 9 HX 370, we have a total capacity of 36 MB. The standard TDP level of the processor is 28 W, but laptop manufacturers have the opportunity to configure the power package in the range of 15-54 W. The top-of-the-line chip also features integrated Radeon 890M graphics with 16 compute clusters, with the iGPU operating at frequencies up to 2900 MHz.

The Ryzen AI 9 365 model has a 10-core configuration. The chip is also equipped with four Zen 5 cores, while the other six have Zen 5c architecture. The frequency formula is 5.0/2.0 GHz. The total cache capacity is 34 MB. Ryzen AI 9 365 is already equipped with Radeon 880M graphics with 12 clusters (2900 MHz), so the iGPU capabilities will be somewhat more modest than the flagship.

Ryzen AI 300 front

Both Ryzen AI 300 models have a hardware NPU for AI acceleration. In AMD’s interpretation, the module is called Ryzen AI, and this time it is a unit with the new XDNA 2 architecture and performance up to 50 TOPs. And this is a frankly impressive figure for a chip of this level.

Ryzen AI 300 LLM performance

It is worth reminding that Ryzen 8040 (Hawk Point) has an NPU with 16 TOPs, while the AI accelerator of Intel Core Ultra 100 (Meteor Lake) mobile chips provides 11-12 TOPs. A visual relative comparison is presented in the chart from AMD’s presentation.

Ryzen AI 300 NPU performance

Such a radical NPU boost allows Ryzen AI 300 not only to break into the AI PC concept, but also to set the tone for this category of systems. Of course, if the claimed performance is confirmed in practical tasks. In this case, the new AMD chips will be able to compete with Qualcomm Snapdragon X Elite (45 TOPS), as well as future Intel Lunar Lake processors (40-45 TOPS) and Apple M4 (38 TOPS).

An NPU unit with 40+ TOPS performance is also a system requirement for compliance with Microsoft’s Copilot+ PC concept. Therefore, the corresponding label will additionally decorate systems powered by Ryzen AI 300.

As for the overall performance of the Ryzen AI 300, it has beaten all current competitors. Here are the charts from AMD’s presentation. The developers assure that in a direct comparison with Ryzen AI 9 HX 370, Intel Core Ultra 185H, Apple M3, and Snapdragon X Elite have no chance. Especially when it comes to conditions with multi-threaded data processing or tasks that use integrated graphics. In the case of Qualcomm chips, additional emphasis is placed on the absence of the need for emulation when performing certain tasks.

Ryzen AI 300 Gaming performance

Well, the Core Ultra 185H was “defeated” in games where the advantage of Ryzen AI 9 HX 370 was 28-47%. Of course, we are talking about using integrated graphics here.

AMD introduces Ryzen AI 300 mobile processors: Zen 5 architecture, RDNA 3.5 graphics and 50 TOPS NPU

As a result, we have very interesting mobile chips that allow us to optimize the system according to certain priorities. That is, Ryzen AI 300 can potentially be used in both cost-effective thin laptops with integrated graphics, good battery life, and a fast NPU, as well as in fairly powerful laptops with discrete graphics cards.

Mobile system manufacturers are already offering their own versions of such platforms at Computex . AMD assures that by July of this year, we can expect more than 100 designs from partners. We are waiting for the first models for a more detailed practical acquaintance.