Firaxis use a Split-screen Approach for Low-latency Multi-GPU Rendering Through Mantle

Rather than doing alternate frame rendering, Firaxis are using the explicit control of multiple GPUs exposed by Mantle to have each GPU draw a different region of the same frame. Presumably they also drive the GPUs from different CPU threads to get higher CPU utilisation in rendering.

Given the hints of a connection between Mantle and OpenGL Next previously revealed by the Khronos Group, the interesting part for OpenGL Next is whether this explicit control of GPUs will also be enabled in the first version of that API. Another intriguing possibility is that of a heterogeneous set of GPUs from different vendors, all painting the same framebuffer. AMD, NVIDIA, and Intel GPUs all working together through the same API. Wouldn’t that be something?

Read the Firaxis blog post by John Kloetzli here:

TAGS > , ,

Post a comment