What is a GPU Tile for Intel® Core® Ultra Processors?

Unlike 13th generation and earlier processors that are monolithic dies, i.e. all the components (CPU, GPU, IO, etc) are implemented in a single tile, the Intel Core Ultra processors family uses a disaggregated architecture to implement the different components (named tiles or chiplets). This means every single component can take advantage of the existing manufacturing processes, be fabricated separately, and then come together onto one substrate to build the device. Also this allows to turn on and off each chiplet individually optimizing for power and have different processes per chiplet optimized for the performance needs of that silicon. In the end, this provides the ability to create market-specific variations.

Specifically for graphics, the processors have their own Graphics Tile for the graphics compute tasks, and other functions like Media and Display Engines and Display interfaces are implemented in other tiles of the processor. Instead of just having one spot that turns on for all graphics and media, it is spread around the chip. This allows, for example, having the GPU Tile turned off if you are just playing media (since no rendering is necessary), which will just use the Media decode block (in the media engine). This translates to power efficiency.