I am currently thinking about which type of graphics card I am going to buy in a few months (maybe in october). If I had to choose right now, I would pick a nvidia GeForce GTX 780 because of the reliability. It seems that the Radeon R9 290 has some issues with temperature and is quite loud.
But then there is Mantle, the new API from AMD. I didn't read much about it, but I heard some people say that it would be nice with Battlefield 4 and brings a lot more frames per second. I am not really sure, if I can believe this, because today I read an article in a german computermagazine called CT (like in Computer Technology), which is one of the few respected magazines with good reputation, in which they stated that most of the time Mantle doesn't bring you any benefit. Unless you have a poor CPU and 3D-scenes that rely on CPU-power.
And then there is this table from another article of the same issue of CT:
With the Radeon R9 290X there is just a minimal gain of one single frame... With Radeon R9 270 the performance is even worse! :? What might be the reason?
So what did you experience with Mantle and what do you know about it? Will it get better in the future or is it just nice for multiplatform gaming? (Linux)
AMD Mantle is extremely useful, if employed properly, which isn't the case with modern AAA titles such as BF. One of the biggest reasons behind it is that there's no DX alternative to it so it doesn't worth the effort just to please AMD users, so they don't try too hard. The other big reason is little programmers education level on Mantle usage.
NVidia cards are actually having quite a bit worse hardware than AMD cards, it's just extremely optimized 3d driver that makes them competitive. On computational tasks, AMD beats NVidia flat out, and Mantle removes CPU data bus bottleneck even though increasing a bit GPU load. When we would come to raytracing rendering (which I beleive will happen eventually) 3d performance wouldn't matter, only plain number grinding capabilities, the FLOPS, would measure card's horsepower.
I did like the idea of Mantle, but I don't think a proprietary graphics API will get that much traction. Just look at the adoption of GPU PhysX, which rarely gets any love outside of games Nvidia has sponsored.
One thing that is cool about Mantle is greater control over multi-GPU rigs. For example, doing stereo 3D rendering with one card per eye. So there are definitely some wins there. But I don't know if that alone is enough to make it relevant.
Especially with the recent optimizations Nvidia has done w/ DirectX, and the improvements in DirextX 12, it seems like there is no point in some other closed API with dubious gains.