It also allows developers to streamline games for optimum performance, by queuing and distributing threads on the CPU and GPU, thereby harnessing more computing power from both. This process could be improved upon with a bit of help from HSA, which means upcoming AMD APUs could gain a bit of gaming muscle with no added silicon.
Then there’s parallelism – Mantle can use multiple CPU cores more efficiently than DirectX and OpenGL, reducing CPU workloads. It gets even better with multi-GPU support, as Mantle can basically “see” multiple GCN-based GPUs as a single GPU, improving load balancing and eliminating microstuttering in the process. It’s also interesting from an APU perspective, as it could potentially lead to even greater performance gains for users with AMD APUs and discrete graphics, but we’ll just have to wait and see.
It remains to see how much of performance boost we are looking at and the adoption of it.