Back in February of 2013, AMD announced that it was working on a new graphics API called Mantle, and that this was going to be an alternative to Microsoft's dominant DirectX API, and also the OpenGL API. On paper it seems good, but in reality things don't seem to be as impressive.
Benchmarks tests carried out by AnandTech, which are based on beta code as opposed to the final release final, show that Mantle provides big performance gains in certain situations, with small performance gains otherwise.
The benchmark results show that the best performance gains from Mantle are to be had when a high-end GPU is combined, which can result in as much as a 30 percent performance increase over DirectX. However, when the same high-end GPU is combined with a high-end processor, those gain fall to under 10 percent.
So, Mantle seems to be able to offer a good workaround for a slow CPU, but when both the CPU and GPU are high-end, it's not much better than the DirectX.
And since not many people are going to put a $500 GPU in with a cheap CPU, the appeal of mantle is going to be limited.
What we don't have is data on how well Mantle performs when a mid-range or budget GPU is combined with a budget GPU. If gains here are good then there might be a case for Mantle – but success hinges on a lot of factors.
Right now, I think that Mantle is a nice science project for AMD, and good things could come of it, but it's not a game changer and certainly not in a position to start taking on the likes of DirectX.