AMD’s new Ryzen AI CPU lineup got off to a good start at Computex, so it’s time to take it up a notch with Ryzen AI Max and Max+ at CES 2025 .
Follow our CES 2025 live blog for all the biggest tech and gadget news straight from Las Vegas. And be sure to follow Tom's Guide on TikTok for the coolest videos live from the show.
Yes, I know these chipset names are sounding more and more like energy drinks, but what this means is up to 250% improved graphics performance.
That puts the Ryzen AI Max+ Pro in RTX 4060 territory, and not to brag…but do you remember when I made the prediction that AMD could lead the charge to the endgame of the dedicated GPU? Well, in my mind, this process has only just begun.
What’s new with Halo?
Two core upgrades have been made here to supercharge AMD’s chipsets in two key ways — many more graphics cores for GPU performance and a vastly faster shared memory interface with quicker bandwidth.
What this enables is some truly monstrous improvements over Intel’s Core Ultra 9 288V. For example, for creative workloads, the Ryzen AI Max+ 395 has an average 2.6x faster rendering time in the likes of Blender and Cinebench. As for games, you’re seeing an average 1.4x faster GPU performance in 3DMark tests.
Most surprisingly is that in 3D rendering performance (what is usually Apple’s bread and butter), AMD’s top of the line chip takes out both M4 and M4 Pro in key 3D rendering benchmarks.
And finally, that NPU + GPU for additional AI performance boost means the chip has 2.2x faster AI performance than a ruddy RTX 4090 — all while consuming up to 87% lower power.
Outlook
One Copilot+ PC comparison that is absent from these tests is with the Copilot OG — Snapdragon X Elite. But while I’m definitely curious to see what this face-off is like, I’m confident that AMD is certainly heading in the right direction.
Not only is AMD taking the x86 potential forward against Windows on Arm, but the company is making me question that the intersection point between dedicated and integrated graphics will come sooner rather than later.