As Nvidia readies up to launch its RTX Spark platform with unified memory this fall, future Windows 11 builds could allow users to decide where unified memory is prioritised - which could prove great for ensuring strong gaming performance.
As originally spotted by Windows Latest, new references in internal Nvidia documentation suggests Windows 11 could soon let users of PCs and laptops with unified memory setups, such as Nvidia's upcoming RTX Spark platform and AMD's Strix Halo architecture, have finer control over how that shared memory pool is dished out.
A unified memory setup marks a radical departure from conventional PC architecture, where standalone GPUs carry their own dedicated video RAM which is separate from the system memory. By combining these into a single memory pool, the GPU can typically draw from a much larger amount of memory than it would have access to as a standalone graphics card.
The same references also turned up on the backend of a Windows 11 build, suggesting this potential feature won't be exclusive to Nvidia's RTX Spark architecture, but be available on all systems with unified memory, with users able to choose how the OS allocates this memory pool. For example, a set amount of unified memory could be reserved for GPU-intensive tasks such as gaming or AI acceleration. Once reserved, that portion becomes inaccessible to other tasks running within the operating system.
The timing of the discovery is notable, as Nvidia gears up to launch its Arm-based RTX Spark platform. Set to arrive later this year on a selection of laptops and mini PCs, it combines a Blackwell-based GPU with an Nvidia Grace CPU, packing up to 128GB of combined memory depending on specification.
This potential feature is only an early sighting, and there's no confirmation from Microsoft on whether this will in fact become a feature on future versions of the OS. With more devices with unified memory hitting the market soon though, it's one to watch.
سەرچاوەی فەرمی بە زمانی ئینگلیزی: https://www.pcgamesn.com/windows-11/unified-memory-allocation-amd-nvidia