Everything these days is AI this and AI that. So it’s on surprise that AMD is bringing out hardware not only for data centers that need deep learning or generative AI horsepower, but also for individual users who just want to have some of that AI magic built into their local machine.

Just like with the new ARM-based Qualcomm Snapdragon Elite CPUs, these AMD chips come with an NPU or Neural Processing Unit built into the chip package. This is a processor that’s geared towards accelerating neural network tasks, the exact sort of processing you need to run image generation or Large Language Models. An NPU allows you to run more of these generative AI programs locally on your computer. Which means no subscription feed, no privacy issues, and no need to be connected to the internet for it to work.

You Need a Heap of RAM

While these chips back a lot of AI processing power, it’s not all that useful if you don’t have enough RAM to take advantage of it. As of this writing, the WIndows CoPilot+ features that need an NPU to work are still unavailable for x86 systems, but over in ARM country the minimum requirement is 16GB of RAM. If you want to run other software that can take advantage of the NPU you may need much more RAM than that.



Image Credit: AMD

Remember that these systems have unified memory, so all processors of every type share the same pool. So expect 24GB or 32GB as a reasonable target for many AI apps. Since you can’t usually upgrade the RAM on ultrabooks later, you should choose carefully at the time of purchase.