GPUs are extremely parallel’d, making them not suitable for many CPU-geared tasks, on LEGACY OSs. Since Haiku is built from the ground up to be multithreaded, it should start utilizing GPU threads past 95% CPU usage and GPU rendering should be offloaded to CPU if GPU passes 95%. That should really give a new meaning to fully utilizing hardware.
Imagine how fast video conversions would be at 100% CPU and 100% GPU usage on the same task! Or how fast games could run on even legacy hardware!
It would give a whole new meaning of speed combined with the awesome speed Haiku already has.