Return to the high-performance GPU market after 22 years Intel will detail Xe architecture on GDC

After the first roll-out GPU was introduced in 1998, Intel will launch a new high-performance GPU this year, returning to the stand-alone market, with AMD and NVIDIA. Intel officially announced its first game card, the DG1, at the CES show, but without details, Intel will announce the secrets of the Xe architecture GPU at the GDC conference in March.

This year’s GDC Game Developers Conference will be held from March 16 to 20, Intel used to be a regular visitor to the GDC Conference, but previously talked about CPU-related or 3D technology content, can now use the GDC Conference to disclose their GPU architecture.

According to Intel, Intel Senior Development Engineer Antoine Cohade will give a speech at the GDC to fully analyze the self-developed Xe architecture, reveal what Intel is building the architecture and performance of the Xe GPU, and the new features and innovations of the Xe GPU architecture. Unique insights into 3D rendering and computing, and clear optimization guidance for developers.

Return to the high-performance GPU market after 22 years Intel will detail Xe architecture on GDC

Return to the high-performance GPU market after 22 years Intel will detail Xe architecture on GDC

According to Intel, the goal of the Xe architecture is to adapt to a wide range of applications, from the highest-end HPC to low-power notebookcan can be met, so the Xe architecture is divided into three levels – Xe HPC, Xe HP and Xe LP, games, mobile PCs and hypermobile products will use the Xe LP architecture.

Return to the high-performance GPU market after 22 years Intel will detail Xe architecture on GDC

Intel previously announced the 7nm process of Ponte Vecchio (Vicio Bridge), which is also an Xe architecture, the EU unit can be greatly expanded to 1000 units, and each unit is newly designed, FP64 double-precision floating-point computing capacity is 40 times the current capacity.

In the Xe HPC architecture, the EU unit connects to HBM’s high bandwidth memory via the XEMF (Xe Memory Fabric) bus, while integrating a large, consistent cache “Rambo” that is accessible to both the CPU and GPU, connecting multiple GPUs together to provide extreme memory bandwidth and FP64 floating-point performance, and supports memory/cache ECC error correction, and the Strong SRAS.

Return to the high-performance GPU market after 22 years Intel will detail Xe architecture on GDC

In the game graphics card, can now be confirmed is the first graphics card code DG1, the specific performance has not been disclosed, but TDP power consumption should not exceed 75W, the overall performance is known between the GTX 950 and GTX 1050, positioning is not high, Because the current DG1 graphics card is actually a mobile GPU for notebook platforms, so don’t expect too much.

For the Xe architecture OF the GPU, we are interested in the desktop version, whether to support light pursuit is also a focus, in the light pursuit technology Intel layout very early, but can not do hardware acceleration light chase on the GPU, specifically to see the PERFORMANCE level of the GPU, low-end graphics card use is also meaningless, High-end graphics depends on how well Intel’s 10nm process can be.

Return to the high-performance GPU market after 22 years Intel will detail Xe architecture on GDC