DirectX 12 Update Allows CPU and GPU to Access VRAM Simultaneously
Microsoft announced new DirectX12 GPU optimization features combined with Resizable-BAR. GPU upload heap (opens in new tab), This allows the CPU to directly access GPU memory concurrently. This eliminates the need to copy data from the CPU to the GPU, resulting in better performance and less system RAM usage for DX12 titles. This new feature is now available in the Agility SDK.
I don’t know the real impact of this feature, but the performance benefits can be huge. The memory size of graphics cards and his VRAM consumption of video games are getting bigger every year. As a result, the CPU has to move more and more data between itself and the GPU.
Using this feature can significantly reduce your game’s RAM and CPU usage, as it only reduces data transfers. This is because the CPU no longer needs to keep copies of data in both system RAM and GPU VRAM to interact with it. Another advantage is that GPU video memory is so fast these days that there is no latency penalty for leaving data only on the GPU. In fact, high-end GPUs with fast video memory may improve latency in CPU access times.
The only requirement gamers need is support for resizable bars or smart access memory on both CPU and GPU. Resizable-bar is the foundation of the GPU upload heap. This feature allows Windows to manage GPU VRAM directly.
For developers, this feature is already supported by Nvidia, Intel, and AMD drivers. For example, Nvidia’s latest Game Ready and Studio drivers (version 531.41 and newer) and Intel A-series/Xe GPUs (driver 31.0.101.4255 and newer) already include it. For AMD GPUs, the developer should consult his AMD Alliance his manager to obtain supported drivers.
This feature was just announced by Microsoft, so don’t expect it to be implemented in games anytime soon. Instead, developers need time to see if it offers enough performance benefits to make it worthwhile.