Intel Unveils Xe3 Graphics Architecture for 50% Performance Boost
Intel's examination of the next-generation Xe3 graphics architecture lines up considerable performance and efficiency improvements compared to its predecessors. The new architecture is designed to provide impressive uplift in graphical capabilities in future Intel CPUs.
Key Features of the Xe3 Architecture
The Xe3 architecture will come into two main kinds a 4-core version for energy-efficient systems, and 12-core more powerful version for non-discrete GPU systems.
- 4-Core Xe3 It consists of 32 XMX blocks, 4 MB of L2 cache, and 4 ray tracing (RT) blocks.
- 12-Core Xe3 It features 96 XMX blocks, 16 MB of L2 cache, and 12 RT blocks.
According to Intel, the Xe3 architecture has a performance boost of over 50% higher than the Xe2 architecture on Lunar Lake chips and it is more than 40% power-efficient compared to Arrow Lake-H graphics. Improvements for the architecture include the redesigned Xe Vector Engine, which could process 20% more number of threads; the new URG controller in the Render Slice for faster anisotropic filtering and depth check; and, overall, 33% more L1/SLM cache per core when compared to Xe2.
In some specific tasks such as depth writing, Intel internal benchmarks would show performance improvements of up to 7.4 times. Panther Lake would be the first line of processors integrated with Xe3 graphics.
Intel Introduces Multi-Frame Generation with XeSS 3
Along with the new architecture, one of the latest releases from Intel is XeSS 3, the next version of their AI upscaling technology, which includes an impressive addition called multi-frame generation that would dramatically augment the frame rates players experience in their games.
The new thing, XeSS-MFG (Multi-Frame Generation), has the capacity of producing as many as four in-between frames from every single rendered frame, providing an up to four-times increase in the actual frame rate at the end, similar to the function of frame generator from NVIDIA's DLSS 4. Most of course, have to be supported by Arc GPUs with XMX matrix compute units, including the Arc A series and the integrated graphics in Core Ultra 200 (Xe2) processors, with support in future for older Xe1 GPUs.
Some other notable features of XeSS 3 are as follows:
- Frame Generation Override A universal choice to allow frame generation in all compatible games.
- Memory Override Provides the ability to allocate system memory to the GPU and NPU.
- Shader Precompilation Reduces the initial load duration of a game by pulling and precompiled shaders out of the cloud.
- HubSpot
- Intelligent Bias Control v3 An updated iGPU scheduler for Panther Lake chips.
Intel has not yet provided a release date for XeSS, however, as the API remains unchanged in this version, all existing games will still be able to work with XeSS 2.
