• Server GPUs have achieved initial results, but market share increase still needs time to fermentation

Intel has been developing GPUs for some time, and its product strategy is similar to that of NVIDIA, which is to develop a GPU architecture, and then develop different GPU products according to the applications of laptops, desktops, and servers, etc. Intel's GPU architecture is Xe-Core, and its main feature is the introduction of 256-bit Vector engine and 1024-bit Matrix engine, based on which other functions, such as optical tracking, will be introduced separately to further satisfy the needs of image-based applications. Based on this foundation, if there is a need for other functions, such as light and shadow tracking, the light and shadow tracking architecture will be introduced to further satisfy the image-based application scenarios. Therefore, from a broad perspective, Intel's GPU products are generally able to cope with AI and machine learning application scenarios, but with different levels of computing power, the AI computing performance they can support also varies. In other words, Intel's AI high-computing power GPUs are still focused on the server domain, much like server processors.