DirectX Shader Model 6.10 Brings Ray Tracing 2.0 and Major Performance Gains

Microsoft's latest DirectX enhancements are bringing the next wave of PC graphics. It's more than just a visual enhancement; it also supports neural rendering, ray tracing, path tracing, and overall game performance. 

DirectX has come a long way, leaving behind the confines of DirectX 11. As DirectX 12 continues to grow with frequent extensions and upgrades, we're on the cusp of another evolution.

Shader Model 6.10 and Agility SDK 1.720 are in Public Preview

Microsoft has today announced the public preview of Shader Model 6.10 and Agility SDK 1.720. Although game developers have not yet implemented these features in their games, they will be used in future games.

It comes on the heels of Shader Model 6.9, released in February, which introduced several new technologies, including long vectors, 16-bit floats, opacity micromaps, and shader execution reordering. 

If you've heard of these before, it's because hardware vendors have already pushed them in recent generations. We've seen this previously, particularly with ray tracing and mesh shaders, which eventually made their way into DirectX.

Better Performance Scaling Across All Hardware

Another goal of Shader Model 6.10 is improved performance scaling. Regardless of whether you have an RTX 5090 or an RTX 3070, it's important to make the most of your hardware.

It includes native matrix operations tailored for AI and neural rendering. It also introduces new batch-asynchronous command list APIs to reduce CPU load and improve GPU scheduling. These changes are about getting more performance out of the hardware, rather than just needing new hardware.

Linear Algebra Comes to Graphics

One of the largest features of Shader Model 6.10 is linear algebra matrix operations (or LinAlg, as it is called internally). They have a broad range of applications, including for machine learning and neural rendering.

They help developers to incorporate machine learning operations into shaders. This allows better use of today's GPUs. So if you have a GPU with AI acceleration units, developers can more effectively distribute workloads across traditional shader cores and AI acceleration units. We can better utilize resources, and you can enjoy better performance without upgrading your hardware.

More Efficient Use of GPU Resources

Today's GPUs have various processing units for different tasks. These new APIs allow better allocation of the rendering workload. Rather than just relying on the shader cores, processing can be offloaded to dedicated cores when needed. 

An analogy might be better task allocation on the GPU. This will increase efficiency, particularly for AI rendering workloads. Optimization in games remains an issue, so tools at the API level to support it are quite welcome.

Variable Group Shared Memory Removes Old Limits

Another feature of Shader Model 6.10 is Variable Group Shared Memory. In the past, group shared memory was capped at 32KB (28KB for mesh shaders). This was a software limitation. The new memory feature allows developers to target the real hardware capability.

Support varies by GPU architecture. All RTX GPUs support the feature, but memory capacity varies. Existing AMD GPUs support the default memory limit, with the full range of memory supported in future driver releases.

As this is a public preview, we are likely to see more improvements before it is released to retail. Microsoft is also adding new ray tracing intrinsics. These intrinsics enable shaders and ray queries to obtain the positions of a triangle's vertices after an intersection.

This, in turn, enhances the organization of ray tracing workloads. We'll see improved performance in complex scenes, particularly when developers start taking advantage of these features in new games.

Development Tools Continue to Improve

Microsoft aims to provide a bigger picture here: DirectX should become easier, faster, and more efficient to develop for on both PC and Xbox. Improved profiling makes it easier to find performance bottlenecks. Better scheduling of work enables more efficient CPU and GPU usage.

Notably, many PC games still suffer from shader stutter. Luckily, it's a big priority for many. Hardware vendors, Microsoft, and engine developers are all involved. It's been slow, but the continuing cooperation bodes well for improvements. That's good news for PC gamers in the long run.

Software is already readying for new hardware. AIDA64 v8.30 now fully supports Intel's Nova Lake and AMD's Medusa 6 APUs set to release in 2027. It's reportedly in the works, with engineering samples already circulating. As these products are released, we will hopefully have more detailed performance data.

NVIDIA has also launched a new RTX 5070 laptop with 12GB of GDDR7 memory. Previously, the RTX 5070 laptop had 8GB. Thanks to denser memory, NVIDIA can now provide more VRAM without other major changes to the product specifications. This can offer laptop users more performance in some games and AI-based applications.

A new generation of graphics is emerging.

Microsoft's latest DirectX releases are a key milestone in the evolution of gaming. Neural rendering, advanced ray tracing, improved GPU scheduling, and many other optimizations are becoming commonplace.

We are moving towards a time when software upgrades can be as important as hardware upgrades. As games start to use Shader Model 6.10, you'll see improvements in performance, visual quality, and hardware efficiency.