AMD Demonstrates Microsoft's WorkGraphs at GDC 2024 with Radeon RX 7900 XTX
TapTechNews reported on June 21, AMD demonstrated Microsoft's WorkGraphs using the Radeon RX 7900 XTX graphics card at the GDC 2024 conference, indicating that its performance is 40% higher than the traditional ExecuteIndirect method.
WorkGraphs is an important feature that can improve the usage efficiency of GPU driver rendering, thereby reducing the workload of the CPU.
In simple terms, currently, part of the work of rendering images is done by the CPU and part by the GPU. But sometimes, the CPU spends too much time deciding what the GPU should do next. WorkGraphs is a system that helps the CPU and GPU work together more efficiently.
This technology allows the CPU to plan ahead and organize tasks for the GPU, enabling the GPU to generate work tasks for itself in a timely manner during rendering. In this way, the CPU can plan the tasks to be done in advance, and the GPU doesn't have to wait for the submission of another task.
The test configuration is Ryzen 7 5800 X CPU, 32 GB DDR4 memory, and Radeon RX 7900 XTX. Some statistics from AMD's collaboration with Kolberg University are as follows:
6600 draw calls per frame (after coalescing)
13 million triangles per frame
200,000 work items
37 nodes and 9 mesh nodes
< 200 MiB of backup storage memory for WorkGraphs
AMD's demonstration shows the potential of WorkGraphs in accelerating rendering workloads. WorkGraphs and mesh nodes are important new features of Microsoft's DirectX API, and all major GPU manufacturers are currently striving to support these features in their drivers.
Related readings:
Microsoft DirectX 12 WorkGraphs Released: Letting the GPU Work Independently and Breaking Through the CPU Bottleneck
Microsoft Plans to Add WorkGraphs Function in Direct3D 12, Able to Release the Bandwidth Limit between the GPU and the CPU