Intel has been reportedly testing an exceptionally large and power-hungry Graphics Processing Unit. The Intel Xe GPU family apparently includes a Multi-Chip Module design GPU that reportedly draws 500W of power for the 4 separate tiles that are stacked on top of each other using Foveros 3D Packaging methodology.
Quite possibly inspired by AMD’s design consideration about multiple chips instead of a monolithic design, Intel has reportedly designed a monstrous GPU that has four Xe-based tiles that collectively draw 500W of power. If Intel is truly designing a four-chip Xe-based GPU, then it could easily outpace not only AMD but also NVIDIA’s offerings for the professional market.
Intel Xe GPU With 500W Power Draw And 4 Xe-Based Tiles Specifications And Features:
Intel has been designing a GPU family. The company hasn’t been speaking openly, but there have been hints about the same. In short, Intel is undoubtedly working to enter the Graphics Market, currently dominated by AMD and NVIDIA. There have been claims that Intel might mark its entry into the graphics market with an attractively priced graphics card for the casual gamer. However, leaked documents indicate Intel might be going after the top-end, premium or professional markets as well.
While AMD is still contemplating the viable option of embedding multiple dies on a GPU package, Intel might have already developed a graphics card constructed using the Multi-Chip-Modules or MCM technology.
— TechPowerUp (@TechPowerUp) February 10, 2020
The exact specifications and features of the 4-Tile Xe Graphics Processing Unit with 500W power draw, aren’t known yet. However, based on the previous leaks about the Intel Tiger Lake GFX based Xe DG1 GPU, the specifications of the latest monster GPU can be derived. If the TGL GFX/DG1 can constitute 1 tile, then there are 4096 cores for the four-core variants.
However, a 4096 core GPU drawing 500W of power doesn’t make sense. Still, it is quite likely that Intel is testing a combination of different spec’d tiles that collectively draw 500W. Incidentally, PCIe 4.0 is capped at 300W, and Intel appears to be having trouble with implementing the same. Hence the leaked documents most likely point to a custom-designed engineering sample meant only for internal testing. In case Intel goes through with the design, it could drop the GPU inside a separate enclosure augmented with an auxiliary PSU that might connect to a computer through external ports.
Intel GPU To Be Launched For Multiple Industries:
The unreleased Intel GPU Family is reportedly codenamed ‘Arctic Sound’. It appears Intel is planning to enter multiple GPU markets including media processing, remote graphics, analytics, AR/VR, Machine Learning (ML), and HPC. Incidentally, Intel Xe GPU should also be used for gaming, but Intel’s aim might be remote, cloud-based game streaming service providers and not desktop gaming.
Been working on this for a while. A lot remains unknown, but Intel has some really interesting stuff happening with its upcoming Xe graphics: https://t.co/fB4UbwwOQx
— Luke Larsen (@lalarsen11) February 10, 2020
The leaked documents indicate the nature of the Intel Arctic Sound, the discrete GPU. The company intends to start out with just one tile client design, but should gradually move up to 4 tiles per GPU. Analysts claim Intel is preparing a total of 4 SDV Xe graphics cards. The Reference Validation Platform or RVP would have about three, to begin with, but Intel should scale up to 4 tile-design. Reports indicate Intel has at least three graphics cards in the works. Their power draw or TDP ranges from 75 watts all the way up to 500 watts.