Intel Xe2 GPUs Official: 50% Performance Uplift, New Ray Tracing Cores, Coming To Lunar Lake First & Battlemage Discrete Cards Later

Intel's Xe2 is official and will be coming to Lunar Lake CPUs and the next-gen Arc discrete graphics lineup codenamed "Battlemage".

Intel Xe2 GPU Architecture Achieves 50% Performance Improvement, New Ray Tracing Units & VVC Support Onboard, Arc Battlemage Coming Soon!

At ITT 2024, Intel squashed all rumors around the cancelation or delay of its GPU and Arc lineup. Tom Petersen gave one of the most charged presentations during the event which was centered around the next-generation Xe2 architecture. Starting with the details, Intel is making things simpler, and instead of using LP, LPG, HP, and HPG naming schemes, the company is simply calling its next-gen lineup Xe2. Internally, these chips will still feature these codenames but it won't be used for the client side anymore.

Related Story ASRock Displays 13 Z890 Motherboards Ready For Intel Arrow Lake CPUs: Aqua, Taichi, OC Formula, Nova, CAMM2 Designs

Some of the goals with Xe2 for Intel were to achieve higher utilization, improved work distribution, and less software overhead. It is a design from the ground up and has fixed several major issues that were noticed with Xe "Alchemist" GPUs. Right off the bat, Intel wowed the audiences with an IP performance efficiency chart that shows gains of up to 12.5x which are quite significant and we have this deep dive to showcase what is Xe2 and how Intel is achieving these gains.

Intel states that the Xe2 architecture, just like Xe, is highly scalable which will lead to its integration within low-power mobile SOCs such as Lunar Lake and up to higher-end Arc graphics cards with discrete options that come out later.

Intel Xe2 Architectural Deep-Dive

So beginning our deep-dive, the second generation Xe core or Xe2 comes with several compute resources that are repartitioned into native SIMD16 engines for increased efficiency.

The Xe2 core features:

• 8 512-bit Vector Engines
• 8 2048-bit XMX Engines
• 64b atomic ops support
• 192KB Shared L1$/SLM

The Vector Engine has also been updated