We couldn't blame anyone for being bored with the state of the PC hardware market right now. The most recent GPU hardware launch was in January, and while we've seen new CPUs drop from AMD, they didn't exactly set many enthusiasts' hearts alight. What we need is a paradigm shift, a truly new product that hasn't really been attempted before. Fortunately, AMD has just such a thing on the horizon.

We've written quite a bit about the upcoming Strix Halo already, but thanks to a new leak from the ever-informative "Golden Pig Upgrade Pack", we now have an idea of how AMD could brand these processors. According to the leaker, there will be three models of Strix Halo SoC, and they will be called "Ryzen AI Max".

Apparently, the top configuration will have sixteen Zen 5 CPU cores and a 40-CU RDNA 3.5 GPU. For comparison's sake, that's kind of like a Ryzen 9 9950X and a Radeon RX 7700 sharing the same package. That chip will supposedly be titled the "Ryzen AI Max+ 395", while the other two parts will miss the "plus" and come with a cut down to 12 CPU cores for the 390, while the Ryzen AI Max 385 will lose another four CPU cores and 32 of its GPU compute units.

The CPU cores are easy to understand, but to put the GPU in perspective for you, we're talking about something with 2.67 times the raw GPU horsepower of the Ryzen Z1 Extreme processors that live inside many of the popular gaming handhelds, like the ASUS ROG Ally X and the Lenovo Legion Go -- and that's the Ryzen AI Max 385. The top-end part has 3.33 times the GPU of Phoenix -- to say nothing of potential optimizations between RDNA 3 and 3.5.

While we're keen to compare a Ryzen AI Max+ 395 against the PlayStation 5 Pro, the most obvious comparison for the Strix Halo "Ryzen AI Max" processors is actually to Apple's M-series chips. Like those parts, they have a wide range of CPU core configurations, and also like Apple's chips, they come with big GPUs and the requisite wide memory bus to feed said graphics components. GPUs are notoriously hungry for memory bandwidth, which is the real reason integrated GPUs haven't been very large in the past.