Combines Zen 5 CPU, RDNA 3.5 GPU, and XDNA 2 NPU cores, first die shot revealed

AMD has officially launched its first Ryzen AI 300 “Strix” APU-powered laptop, which combines Zen 5 CPUs, RDNA 3.5 GPUs, and XDNA 2 NPU cores.

Introducing AMD’s First Copilit+ APU! Introducing the Ryzen AI 300 “Strix” Family for Laptops with up to 12 Zen 5 Cores, 16 RDNA 3.5 GPU Cores, and 55 TOPS XDNA 2 NPUs

The AI ​​PC ecosystem is evolving by the minute as everyone tries to cash in. AMD was the first chipmaker to introduce an “AI PC” SOC with its Phoenix APU family in 2023, followed by the release of a significantly improved Hawk Point series offering higher NPU TOPS.

With the introduction of the Microsoft Copilot+ PC platform, AMD is taking another big leap forward, introducing the Ryzen AI 300 “Strix” platform that combines several aspects to meet the demand for the next generation of innovative AI experiences in the consumer world. For AMD, the goal is very clear: improve AI performance, deliver rich experiences with the new GenAI and Copilot+, and do it all locally on the PC.

So today, AMD finally announced their new Ryzen AI 300 “Strix” APU family for laptops. The name may be a bit confusing, but this is a fresh approach to a new beginning for the AI ​​PC ecosystem. Here’s how AMD’s Ryzen naming scheme for the past two generations has looked like:

• Ryzen 7040 (Phoenix) -> Ryzen 8040 (Hawk) -> Ryzen AI 300 (Strix)

Starting with the 300 series is the third generation of AI PC APUs, with Phoenix being the first generation and Hawk being the second generation. It was also mentioned that with Ryzen AI, AMD has consolidated their portfolio and is no longer dependent on fixed TDPs for certain SKUs. So, they no longer need to use U/H/HS series branding. Instead, AMD is giving OEMs the flexibility to tune these APUs to their demands, with TDPs ranging from 15W up to 54W. The “HX” designation denotes the highest-end SKUs, while non-HX SKUs will be designed to save power.

As mentioned above, the AMD Ryzen AI 300 “Strix” APU combines three major core technologies:

• Zen 5 (CPU core)
• RDNA 3.5 (GPU core)
• XDNA2 (NPU core)

AMD Zen 5 is an all-new architecture (full details here) that will also be coming to several other processor families, including the Ryzen 9000 “Granite Ridge” desktop CPUs for high-end PCs and the 5th Gen EPYC “Turin” CPUs for data centers.

AMD Strix APUs are based on a single monolithic die based on the TSMC 4nm process node, and feature up to 12 cores, 24 threads, and employ both Zen 5 and Zen 5C cores. This approach is taken to achieve maximum efficiency with the Zen 5C cores and maximum performance with the Zen 5 cores, as the former tend to run at slightly lower clock speeds while maintaining the same ISA. The Strix APU die measures 12.06 x 18.71 mm, or nearly 225 mm2.

The AMD RDNA 3.5 iGPU for the Ryzen AI “Strix” APUs has been optimized to include certain architectural updates and support for more compute units, totalling 16 on the flagship parts.

Enhancements added in RDNA 3.5 compared to the RDNA 3 architecture include optimized performance per watt, optimized performance per bit, and improved power management for longer battery life. AMD RDNA 3.5 GPU features include:

• 2x Texture Sampler Rate: A subset of the most common texture sampling operations are now 2x the rate for common game texture operations.
• 2x Interpolation and Comparison Rate: Most of the rich vector ISAs for interpolation and comparison have doubled the rate of common Ops in shaders.
• Memory management improvements: Improved primitive batching to reduce memory accesses, improved compression techniques, workload reduction, and LPDDR5 access optimizations.

And the highlight of the show was the XDNA 2 NPU, which offers up to 55 TOPS on its own, which is big. This NPU is faster than the one featured on the Snapdragon X CPU platform, and faster than the Intel Lunar Lake platform, which AMD expects to offer around 45 TOPS.

AMD Ryzen AI 300 “Strix” APU SKUs and Specifications

Now that we’ve covered the core IP, let’s talk about the SKUs. As of now, AMD is introducing two SKUs within the Ryzen AI 300 “Strix” family: the Ryzen AI 9 HX 370 and the Ryzen AI 9 365.

Firstly, the Ryzen AI 9 HX 370 is a 12 core 24 thread chip with 4 Zen 5 and 8 Zen 5C configurations. The chip runs at a boost clock of up to 5.1 GHz, has 36 MB cache (24 MB L3 + 12 MB L2) and a Radeon 890M iGPU with 16 compute units or 1024 cores. So compared to the previous flagship Ryzen 9 8945HS, it has 50% more cores/threads, 33.3% more compute units and 3.12x more NPU performance, a nice generational gain.

AMD claims that the XDNA 2 NPU has enhanced AI tiles to double multitasking capabilities, more compute power, and is twice as efficient with improvements specific to GenAI workloads. This makes the XDNA 2 NPU up to 5x faster than the previous generation at LLM. The higher-end chip is the Ryzen AI 9 HX 375, which has 55 TOPS of NPU AI performance and appears to be limited to select laptops such as the HP OmniBook Ultra.

The second chip in the lineup is the AMD Ryzen AI 9 365. It is a 10-core, 20-thread chip with a boost clock of up to 5.0 GHz, 34 MB of cache, and a Radeon 880M iGPU with a total of 12 compute units or 768 cores. The chip features four Zen 5 cores and six Zen 5C cores. Both Strix APUs come with the same level of NPU capabilities with XDNA 2 hardware for up to 55 TOPS of AI performance.

AMD Ryzen AI “HX” APUs:

When it comes to laptops, the latest designs currently available are:

With the world’s fastest NPU delivering 55 AI TOPS, AMD is poised to revolutionize the AI ​​PC segment with its Ryzen AI 300 series chips, which are expected to see increased laptop shipments leading up to launch and greater availability by the 2024 holiday season.