Intel will announce and release its next-gen Arrow Lake processors later this year, representing the first top-to-bottom client CPU architecture featuring totally new CPU and GPU architectures in many years from the company.
Intel's current-gen Core Ultra 100 series "Meteor Lake" processors feature brand new GPU cores in Xe-LPG, but Arrow Lake CPUs will come with either Xe-LPG or Xe-LPG+ graphics, something we're learning today in some new leaks.
Intel engineer Haridhar Kalvala, explained: "Some SKUs of Arrow Lake use slightly newer Xe-LPG+ graphics IP (version 12.74). Add some additional PCI IDs and extend the code to support the newer IP version. The general code flow should continue to match existing MTL and Xe-LPG code paths".
- Read more: Intel's next-gen Arrow Lake-S CPU teased: no Hyper-Threading support
- Read more: Intel's next-gen Arrow Lake is first gaming CPU with AI
- Read more: Intel to debut Thunderbolt 5 with Arrow Lake CPUs
It looks like Intel's new Arrow Lake-H mobile processors will use the updated Xe-LPG+ graphics, while the Arorw Lake-S desktop processors will use the same integrated GPU as Meteor Lake, with Xe-LPG graphics.
What's the difference between Xe-LPG and Xe-LPG, you're wondering? The + there with Xe-LPG+ means it should support DPAS (Dot Product Accumulate Systolic) instructions. These instructions are already used inside of the Xe-HPG architecture but were disabled in the Xe-LPG variant.
They support FP16, BF16, IN4, and INT4 multiplication with 16 or 32 bits accumulate, meaning that through the XMX core, the GPU can perform more operations per clock. Small changes but something worth noting.