Leak suggests Intel could launch desktop and laptop CPUs with on-die AI accelerators by 2023

Intel is certainly concerned about Apple’s achievements with the M1 processors, and, according to a previous leak from Moore’s Law Is Dead, it seems like Team Blue already has a plan to disrupt Apple’s edge. This plan includes the Royal Core architecture developed by Jim Keller so that Intel’s future X86 CPUs can be more efficient than Apple's ARM-based custom SOCs, but a full implementation of this new architecture will only come with the 2024 Nova Lake family. However, CPU families coming before Nova Lake, like Lunar Lake and Meteor Lake (supposed to launch in 2023), should already bring AI performance improvements by integrating VPU machine learning accelerators on the processor die, similar to Apple’s Neural Engine on the M1.  

AI acceleration may play a very big role in a few years at least for smartphones and mobile devices, yet it is not really clear what particular utilizations this tech could have for the average PC user. Tom from Moore’s Law Is Dead is suggesting that a vision processing unit could dynamically upscale lower resolution videos. For example, it may identify a low resolution video on a reddit thread and improve it similar to how Nvidia’s DLSS makes certain details clearer compared to the source resolution. Additionally, the VPU can help improve image scaling and content-aware tools in popular image editors like Adobe Photoshop.

While the VPU focuses only on video content, a general purpose AI accelerator could allow desktop-centric small developers to build apps that only mega companies such as Google, Facebook and Microsoft could afford before. Some interesting use cases come directly from Apple: real time conversation translation, improved text-to-speech capabilities for digital assistants like Siri, simplified map directions, plus instant identification of certain elements in nature via a large database that is trained with millions of pictures of plants and animals. Windows 11 could easily turn an online article into audio form that can be listened to while users perform other tasks.

Intel’s competitors are probably looking to add similar AI support. AMD could use the AI accelerators from Xilinx and Nvidia surely plans to make use of everything AI-related coming from ARM, if the acquisition deal is finalized. Right now, it may seem that AI accelerators do not really do that much, but we should not underestimate the impact that this technology can have in just a few years.

Buy the 13-inch Apple MacBook Pro with M1 SoC on Amazon