Last year, AMD (AMD 0.87%) announced that it would be releasing a new graphics API called Mantle as an alternative to Microsoft's DirectX. Most PC games are built using DirectX, which is Windows-only, but AMD promised that Mantle would introduce substantial performance gains by removing overhead inherent in Microsoft's API. The first game to support Mantle is Electronic Arts' Battlefield 4, and after some delays, the first benchmarks are finally available. Does Mantle live up to the hype? More important, what does it mean for investors in AMD and its competitors, like Intel (INTC 2.20%) and NVIDIA (NVDA 2.17%)?
A performance rundown
Without getting too technical, a modern PC game works in two steps. First, the CPU updates the game state, dealing with user input, determining where objects in the game world need to be, etc. Then, the GPU takes that information and draws it on the screen. But the CPU needs to tell the GPU what to draw, and this process involves a significant amount of overhead. One of Mantle's biggest selling points is making this process more efficient, freeing the CPU from much of this burden.
Mantle also includes other optimizations on the GPU side, allowing for performance gains there, as well. The initial benchmarks from AnandTech, which are based on beta software and are not final, show that Mantle provides big performance gains in certain situations, with small performance gains otherwise.
In a nutshell, when a high-end GPU like AMD's Radeon R290X is paired with a powerful processor, like Intel's high-end offerings, the performance gains of Mantle compared to DirectX are minimal, between 7%-10%. In this case, the CPU is fast enough that the bulk of Mantle's optimizations don't matter much in the grand scheme of things. When a slower processor is used with the same high-end GPU, the performance gains are much larger, around 30% or so. In this case, the CPU is the bottleneck, and Mantle's optimizations are more meaningful.
What this means for AMD, Intel, and NVIDIA
The largest performance gains for Mantle came from the most unlikely scenario -- a high-end GPU combined with a low-end CPU. Pairing a $500-plus GPU with a cheap CPU doesn't make much since, since the CPU would prevent the full potential of the GPU from being realized, and using a higher-end CPU eliminates much of Mantle's performance gain.
The conclusion here is that Mantle won't have too much of an impact on the high-end GPU market. The 7%-10% boost that Mantle gives AMD's highest-end GPUs is nice, but it's not enough of an improvement to give NVIDIA a reason to worry. And since support for Mantle has to be built into each game in order to take advantage of these improvements, a single-digit performance boost may not be enough to convince developers to support the API.
NVIDIA is expected to release GPUs built on a new graphics architecture Maxwell sometime this month, and significant performance improvements over the previous architecture are likely. NVIDIA has been one step ahead of AMD in terms of performance for quite some time, with its nearly year-old GTX Titan claiming the top-performing GPU crown for the better part of a year before AMD released new GPUs. Maxwell should give NVIDIA a big lead in terms of performance, and Mantle doesn't seem to be able to close that gap.
Mantle may help AMD sell more CPUs, though, assuming developers support the API. If the trends from these benchmarks hold for lower-end GPUs, an AMD CPU paired with a mid-range or low-end GPU could potentially achieve a significant performance boost using Mantle. From a performance-per-dollar perspective, this setup may have a big advantage over a similarly priced Intel-NVIDIA gaming PC. The high-end CPU market should be unaffected, though, since Intel's i7 processors are fast enough, based on the benchmarks, that most of Mantle's optimizations don't increase performance by very much at all.
The bottom line
Although these benchmarks are very early, Mantle doesn't look like a game changer for AMD. The improvements over DirectX are impressive, but Mantle seems to disproportionally benefit lower-end PCs. For serious gaming PCs with high-end parts, the small improvements brought by Mantle likely aren't enough to pull customers away from NVIDIA. Cnvincing developers to support the API will prove difficult, at best. There's still room for these numbers to improve as AMD's Mantle drivers are refined, but it's unlikely that Mantle will ultimately live up to its promise.
