AMD Ryzen AI 300 shows dominant performance over Intel Lunar Lake and Qualcomm Snapdragon X
Our first tests with AMD’s AI-focused mobile processors
🥊 AMD reveals benchmarks pitting its Ryzen AI 9 HX 370 against the Intel Core Ultra 9 185H and Snapdragon X Elite X1E-84-100
📈 The charts show Ryzen AI 9 HX 370 comes out on top for computing, creativity, and gaming vs Intel and Qualcomm
🎮 I played Fallout 4 and Ghost of Tsushima at 1080p 60fps on an Asus ZenBook S 16 with a AMD Ryzen AI 9 HX 370
🎨 AMD’s Block FP16 datatype allowed me to generate OnnxStack in mere seconds one after another
For the first time in history, there’s a three-horse race between Intel, Qualcomm, and AMD vying to be the best processors for Windows 11 Copilot PCs. At a closed-door event in LA, AMD revealed benchmarks of how its top AMD Ryzen AI 300 CPU compares to Intel and Qualcomm’s offerings.
AMD’s benchmarks pit the 12-core and 24-thread Ryzen AI 9 HX 370 against the Intel Core Ultra 9 185H and Snapdragon X Elite X1E-84-100. From the charts AMD has supplied, we can see AMD holds a 13% lead in PCMark 10 and a 1.21x lead in Proycon office productivity. In more creative applications, the Ryzen AI 9 HX 370 dominates with a 3.8x faster 3D render speed in Blender while pushing 1.6x faster video encoding in Handbrake.
Meanwhile for gaming AMD claims it can deliver well over 1.25x higher frame rates in games like Assassin’s Creed Mirage, CyberPunk, FarCry 6, and GTA V. It’s also worth noting that AMD alleges that Qualcomm’s Snapdragon X Elite chip wasn’t able to run a third of the gaming benchmarks it tested.
Beyond the charts, I played Fallout 4 and Ghost of Tsushima on an ultra-thin Asus ZenBook S 16 running on an AMD Ryzen AI 9 HX 370 and only saw a gorgeously smooth 1080p and 60 frames per second experience.
The AMD Ryzen AI 300 series of CPUs of course focus on AI as well, boasting 50 TOPs of power, the most out of the NPU on its own compared to Intel Lunar Lake’s 48 TOPs and Qualcomm Snapdragon X Elite’s 45 TOPs. AMD also argues that it can leverage its Block FP16 datatype to combine the full accuracy of FP16 with the speed of INT8 without requiring quantizing, tuning, or retraining of existing models.
Up until now, Block FP16 sounded like a pipe dream, but I was able to test its real-life performance in person and generate images with OnnxStack in mere seconds one after another.
Of course, that’s just the tip of the iceberg. We’ll soon be able to test Ryzen AI 300-powered laptops by the end of the month and really put AMD’s new processor through its paces.
Kevin Lee is The Shortcut’s Creative Director. Follow him on Twitter @baggingspam.
Isn't the Intel Core Ultra 9 185H Meteor Lake and not Lunar Lake?