Microsoft has released Shader Model 6.10 and AgilitySDK 1.720 in preview, giving DirectX 12 developers a new set of tools for modern graphics, machine learning, and heavier GPU workloads. The update also arrives with DXC 1.10.2605.2, the latest DirectX Shader Compiler preview needed for these new shader features.
This release is not aimed at everyday Windows users. It is mainly for game engine teams, graphics programmers, GPU vendors, and developers working on advanced rendering. But the long-term impact could still matter to players because these features may help future games use AI-assisted rendering, reduce wasted GPU work, and handle complex visual effects more efficiently.
Shader Model 6.10 gives developers better tools for neural rendering, ray tracing, memory use, and GPU scheduling
The biggest addition is LinAlg, short for Linear Algebra. It brings new matrix APIs to HLSL, Microsoft’s shader language. Matrix math is important for AI, image processing, and neural rendering, so adding better matrix support directly into shaders gives developers a cleaner way to use modern GPU hardware for those workloads. Microsoft’s DirectX specs say the goal is to create a more consistent baseline for matrix operations across hardware from different GPU vendors.
This matters because the future of graphics is not only about drawing more pixels. More games are using AI-style techniques for upscaling, denoising, lighting, animation, and image reconstruction. Better matrix support in shaders could make those systems easier to build and easier to run across different GPUs.
Microsoft also added Group Wave Index. This gives shaders a safer way to understand how waves are arranged inside a thread group. In simple terms, it helps developers organize GPU work without relying on fragile tricks that may behave differently across hardware. That should make some compute and rendering code cleaner and more portable.
Another useful change is Variable Group Shared Memory. Earlier limits capped group shared memory at 32KB, or 28KB for mesh shaders. Shader Model 6.10 lets developers query the real hardware limit and declare how much memory a shader needs. Microsoft says this can help with large tile culling, software rasterization bins, and big matrix workloads.
| New feature | Why developers may care |
|---|---|
| LinAlg Matrix APIs | Helps with neural rendering, AI workloads, and image processing |
| Group Wave Index | Makes wave-level GPU work easier to organize across different hardware |
| Variable Group Shared Memory | Lets shaders use more of the memory modern GPUs can provide |
| Raytracing intrinsics | Gives shaders more direct information during ray tracing work |
| Batched Async Command List APIs | Lets independent copy, clear, and resolve work overlap instead of running strictly in order |
The update also includes new ray tracing intrinsics, such as TriangleObjectPositions and ClusterID. These give ray tracing shaders more information about what was hit during a ray query or hit shader. ClusterID is more forward-looking because Microsoft says clustered geometry support is not ready yet, but it is part of where DirectX ray tracing is heading.
On the D3D12 side, Batched Asynchronous Command List APIs may help reduce unnecessary GPU waiting. Older commands like copies and clears often ran in strict order, even when they were working on separate memory areas. The new approach lets independent work overlap, while developers add synchronization only when real data hazards exist.
Hardware support is still mixed because this is a preview. Microsoft says AMD, Intel, and NVIDIA all have driver paths for the release, but support differs by feature and GPU generation. For example, NVIDIA supports LinAlg Matrix on all RTX hardware, while AMD supports it on Radeon RX 9000 GPUs. Intel plans LinAlg support in a future release.
For now, this is a developer-facing preview, not a feature players will notice immediately. Still, it shows where DirectX 12 is going. Microsoft is preparing the graphics stack for a world where games use more AI math, more ray tracing, more parallel GPU work, and more advanced memory use. Future games may benefit from this work quietly in the background, even if players never see the words Shader Model 6.10 in a settings menu.
Discussion (0)
Be the first to comment.