In the world of PC gaming, Microsoft's (MSFT +0.66%) Direct3D, part of DirectX, has long been the dominant graphics API. Almost all high-profile PC games use DirectX, which is tied exclusively to Windows, making the OS the standard for PC gaming.
Advanced Micro Devices (AMD +18.37%) attempted to change this last year with the introduction of its own proprietary graphics API, Mantle. Mantle vastly reduced the amount of CPU overhead compared to DirectX, leading to big gains in performance in situations where the CPU was the bottleneck.
I argued last year that Mantle had absolutely no chance of success, and that it would provide no benefit whatsoever to AMD, because the next version of Direct X would introduce the same efficiency gains. With the release of Windows 10 and DirectX 12 coming later this year, the first benchmarks comparing DirectX 12 to Mantle have been released. It's not pretty for AMD.
DirectX 12 is the real deal
The folks over at Anandtech recently tested an early version of DirectX 12, and the results show two things.
Source: Anandtech
First, the gains in performance compared to DirectX 11 are substantial, and in cases where the CPU is the bottleneck, big gains are very possible. Second, DirectX 12 produces similar results on AMD cards compared to Mantle. Mantle is still a bit faster in most cases, but given DirectX 12 is still months away from release, this difference may be temporary.
Both DirectX 12 and Mantle achieve these performance gains by eliminating much of the overhead involved with sending commands to the GPU. DirectX 11 hits a severe bottleneck if the number of commands grows too large, and it looks like DirectX 12 will fix this problem.
Why DirectX 12 is important for Microsoft
One reason, although certainly not the only reason, that Windows has remained the dominant PC operating system is gaming. The fact that almost all PC games are built on DirectX, instead of the alternative cross-platform OpenGL, gives Windows an enormous advantage. According to the Steam hardware and software survey, about 95.5% of Steam users are running some form of Windows, with the scraps split between Mac and Linux.
Maintaining Windows' dominance in the PC gaming market is critical for Microsoft, and that requires DirectX to remain the de facto standard graphics API. Mantle was always unlikely to usurp DirectX, given that it only works on a subset of AMD products, and neither Intel nor NVIDIA was ever going to support it. But allowing DirectX to fall too far behind could have opened the door for the next version of OpenGL, which will also introduce gains in efficiency. This now appears unlikely given the early DirectX 12 performance numbers.
Why AMD's Mantle will lead nowhere
Mantle was always a long shot, but at the very least it likely spurred Microsoft to focus on efficiency. Mantle has gained the support of a decent number of games and game engines, but it's difficult to fathom where Mantle fits once DirectX 12 is released. There may still be some performance benefits for AMD's GPUs, but with NVIDIA the market leader in discrete GPUs and Intel the overwhelming market leader in integrated GPUs, supporting Mantle likely won't make much sense for developers.
AMD's PC business has been falling apart recently, and it was honestly a waste of resources to develop an alternative graphics API. It's now clear from the DirectX 12 performance figures that Mantle's advantage is going to largely disappear later this year, and it's also clear from NVIDIA's market share gains that Mantle has done exactly nothing to boost AMD's GPU business.
AMD deserves credit for starting the conversation on eliminating graphics API overhead. But that conversation will be finished without it, and AMD will get no credit where it matters: the bottom line.
