Microsoft is testing a new setting in Windows 11 that lets users control the amount of shared memory allocated to graphics and artificial intelligence. The feature appeared in beta build 29648 on August 17, but the full release date is still unknown.
This innovation is primarily designed for modern systems, where processors, graphics, and AI accelerators use a common pool of RAM rather than separate VRAM. This approach is becoming increasingly important: local AI models require more and more memory, and built-in GPUs cannot have the same amount of video memory as discrete graphics cards.
Judging from early implementations, Microsoft will provide users with a slider for manual settings and several out-of-the-box options. By dedicating more shared memory to graphics and accelerators, games and native AI applications will receive more resources. When configured in reverse, more memory is left for Windows itself and common programs.
As of now, Microsoft has not officially revealed the final form of the feature nor announced its release date. However, the idea itself seems quite logical: when CPU, GPU and AI accelerator share a resource, the ability to manually set priorities may be more useful than another “magical” optimization decided by the system for the user.