We believe AMD's next-generation Strix Point APU laptops and gaming handholds have long been rumored to be of near-mythical quality. But thanks to the Computex2024tech show, it's finally here and looks like an absolute beast with more CPU cores, more GPU compute units and a much more powerful NPU.

The first laptop with a new APU will be launched on May 7, and guess what the rumored "Ryzen AI" brand is true. The new chip is known as the Ryzen AI300 series. Indeed, rumors that AMD last changed its brand from Ryzen AI100 to Ryzen AI300 are also true. Some elements of the marketing material have not been fully updated, and the "Ryzen AI100 series" remains stenciled on at least 1 of the images on the chip. oops...

But enough pre-launch trivia, the first thing that makes Strix Pont tick is a combination of both Zen5 and Zen5C cores. To learn more about AMD's new CPU design, read Nick's article on the new Ryzen9000 desktop chip that AMD is launching today. But AMD claims an average IPC profit of 16%, and this is a decent generation rise.

The clock speed is comparable to the previous Apu, but it is the number of cores that are not. The older 7000 series Phoenix and 8000 series Hawk Point APUs had up to 8 CPU cores, as seen in various laptops and gaming handholds. The Zen5C is AMD's compact CPU core, but it has exactly the same execution unit as the Zen5 and has less cache. As a result, a block of Zen5C cores takes up significantly less space than the equivalent number of fully cached Zen5 cores. For some software, such as media encoding, the Zen5C will be almost as fast as the Zen5. For games, the extra cache in Zen5 makes a noticeable difference. Of course, it means that careful thread scheduling will be required to ensure that important game threads are prioritized on these Zen5 cores. But this is not new. AMD has already done this with asymmetric multi-chip desktop cpus, such as the ryzen9 7950X3D with 3D V cache on one CPU die, rather than asymmetric multi-chip desktop cpus with 3D V cache on the other CPU die.

Anyway, there are 2 models at launch. It will be. The AMD Ryzen AI9HX370 features 4 Zen5 cores and 8 Zen5c cores, with a total of 24 threads, a total cache of 36MB, and a maximum boost clock of 5.1GHz. On the other hand, AMD Ryzen AI9 365 loses 2 Zen5c cores, drops to 34MB of cache and works a bit slower at 5GHz. And yes, it's right, 1 is the Ryzen AI9 "HX" chip and the other 1 is simply the Ryzen AI9. Tech branding, go to the figure.

The other major element, of course, is graphics. Strix Point raises the count of graphics CU (or compute unit) from 12Cu on the previous best APU to 16Cu on the top Ryzen AI9HX chip. The non-HX models feature the same 7840 Cu as the Ryzen 8840U and 12U

but these Cu are not the same RDNA3 technology as before. AMD calls it RDNA3.5. Unfortunately, AMD isn't terribly forthcoming about the details, but rumors so far suggest that RDNA3.5 is going well because of a pretty big improvement in ray tracing performance, which is true, because the only thing that's really wrong with RDNA3 is its relatively weak ray tracing performance.

The other major element was the AI Accelerated Neural Processing Unit (NPU), where AMD made the biggest gains. AMD claims 5 times better performance and 2 times more efficiency than the Phoenix APU's first NPU. AMD says Strix Point's new NPU is suitable for the top 50 in AI performance.

In particular, it is over the top 45 of the new Qualcomm Snapdragon X Elite chips and will be first on the market with Copilot+chops in just a few weeks.

Microsoft, of course, has set a minimum TOP of 40 to qualify for the official Copilot+AI PC standard. Therefore, the Ryzen AI300 series will be the first authentic x86 chip to meet that standard. In particular, AMD needs to clearly delve into the Snapdragon chip and its Arm CPU core to emulate traditional x86 apps that are not available in the native Arm format "Emulation

Speaking of performance, how does this new APU stack up?" There are only a few AMD metrics. In most cases, AMD does not compare the new chips to previous Apus and prefers to compare them with Intel's Meteor Lake replacement.  Compared to the INTEL Core Ultra9 185H, AMD believes the Ryzen AI9HX370 is 47% faster than the Intel chip in Cinebench rendering and 40% faster for video encoding. For games on integrated GPUs, AMD claims the gap is between 28% and 47%, but it's not clear which settings are being used or at what power level.

Another detail AMD is talking about right now is the silicon nodes used in these new Apus.1 It is almost certain that it will not be one of the latest N3 nodes, but TSMC1, which is also called 5N, or 5NM, or something very close to it. It was the latter, you can be sure that AMD is screaming about it.

Still, this is a very strong indication overall. If anything, we're most excited about it as a basis for portable gaming chips. It's not clear what the impact of battery life is, but in addition to the extra four graphics CUs, the ray tracing upgrade expected bodes very well.

In that note, it's striking that AMD doesn't mention much about battery life. If you are using the same or similar silicon node technology, the efficiency gains will be architecturally from the Zen5 core. We'll take a look, but if AMD isn't making a lot of noise about improving battery life, it makes sense to keep expectations down. 

In any case, AMD's laptop Apu was already very competitive. These new ones seem likely to increase their benefits. I can't wait to get a laptop with Strix Point, or even a better handheld.