Intel Graphics Multi-Adapter: Integrated and Discrete GPUs Together

Gaming PCs often pair a high-performance CPU with a discrete GPU. Those CPUs often include integrated graphics, usually left idle. This video demonstrates using the power of the integrated GPU to offload work from the discrete GPU.

Methodology / Approach

The sample's foundation is the Microsoft D3D12 n-body particles sample, which demonstrates using async compute to execute simulation and rendering in parallel on a single GPU.

This sample runs the simulation (compute shader) on the integrated GPU, so the discrete GPU has more room to work on graphics. It adds a copy queue on the discrete GPU to move data from system to discrete-local memory. All 3 stages run in parallel: simulation, copy, and rendering.

With this technique, the potential speedup is a function of how much work can be offloaded from the discrete GPU.

Technologies Used

D3D12, Direct3D12, Multi-Adapter, Cross-Adapter, Async Compute, Copy Queue, Simulation, Particles