Microsoft has begun testing a dedicated allocation management feature designed to give users direct control over shared system RAM on modern PCs. Uncovered in recent insider preview builds, the feature allows system memory to be explicitly reserved for graphics processing units and neural processing units.
This development comes as hardware manufacturers increasingly move toward unified memory architectures that share high-speed RAM between the main central processor, integrated graphics, and dedicated artificial intelligence silicon. By letting users configure dedicated memory carve-outs, the operating system aims to eliminate performance bottlenecks during intensive tasks like local large language model inference and high-end gaming.
Unified Memory Allocation Setting in Build 29648
The new functionality was first discovered within Windows 11 Build 29648.1000, where internal code references point to an unreleased feature internally named IntelligentCarveout. System files associated with the update, including a dedicated dynamic link library named SettingsHandlers_UnifiedMemory.dll, indicate that Microsoft is building a dedicated interface inside the Windows Settings app.
Text strings embedded in the experimental preview build outline how the controls will operate for end users. One prompt explains that enabling the setting lets the operating system reserve additional unified memory for graphics and AI intensive games and applications. Microsoft explicitly notes within the interface that memory carved out under this feature will be isolated and reserved exclusively for hardware accelerators, making it unavailable to general background applications.
Balancing RAM Between NPU and GPU Workloads
Traditional x86 personal computers have long relied on a split memory architecture, where the CPU uses main system RAM while discrete graphics cards maintain their own dedicated video RAM. Integrated GPUs historically used dynamic shared memory, where Windows assigned dynamic ceilings based on workload demand. However, that dynamic pool was subject to constant OS renegotiation, which could lead to stuttering or memory allocation failures when running heavy computational models.
With the rapid rise of local AI deployment, software frameworks like LLM runners and real-time generation tools often require massive amounts of continuous video memory to load model weights into active RAM. The ability to assign a static, guaranteed pool of memory to hardware accelerators prevents system overhead from pulling RAM away from processing units during active execution.
Target Hardware and System Requirements
While the setting remains hidden behind internal feature flags in preview builds, early technical analysis indicates that the control panel will target specific next-generation hardware profiles. Copilot+ PCs equipped with dedicated NPUs and systems featuring high-bandwidth unified memory architectures are prime targets for the feature.
Industry observers note that the control system aligns closely with upcoming hardware platforms, such as multi-chiplet processors and mobile workstations designed with merged LPDDR5x memory buses. Systems built around unified architecture can contain up to 128GB of shared system memory, making explicit allocation toggles vital for developers who want to split capacity between operating system tasks and massive local AI models.
Expected Timeline for Windows 11 Release
Microsoft has not publicly announced an official release timeline or specific version rollout for the unified memory controls. Because the feature is currently buried within experimental Insider builds, the user interface remains subject to design changes, functional adjustments, or potential delays depending on testing feedback.
However, the inclusion of fully formed localization strings and dedicated background DLL files suggests that development is progressing beyond early conceptual phases. Industry analysts expect the controls to surface officially in an upcoming feature update or target release for AI-capable devices later this year.
As desktop operating systems adapt to the shifting demands of modern computing, granular hardware management is becoming essential. By providing direct controls for unified memory allocation, Microsoft is taking a necessary step toward ensuring Windows 11 can handle heavy local AI processing alongside demanding gaming tasks without performance degradation.
