Back in February, graphics chip company NVIDIA (NVDA 1.56%) launched the GeForce GTX Titan, a desktop graphics card that retailed for $1,000 and, through today, remains the fastest single-GPU card on the market. Competitor, Advanced Micro Devices (AMD 2.20%), offered nothing that could compete in the high-end, and the company lost market share as a result.
Now, however, Advanced Micro Devices, or AMD, is back with a vengeance. At the end of September, AMD announced a new set of GPUs, the Radeon R7 and R9 series, which aim to take back the high-end throne. Part of AMD's plan is the introduction of a new graphics API, Mantle, which allows developers intimate access to graphics hardware. The company claims that its new cards will trounce those offered by NVIDIA in terms of performance, provided that affected software is developed specifically for the Mantle API.
Is this the first step in AMD's resurgence in the GPU market? Or, will the Mantle API fade into irrelevance along with AMD?
A brief overview of graphics APIs
There are two main graphics APIs in use for PC gaming today: DirectX from Microsoft (MSFT 1.57%) and OpenGL. Most games for the Windows operating system use DirectX, and it's become the de facto standard of the graphics API world.
DirectX, however, introduces significant overhead. Because there are so many different graphics cards available, the API must be designed to work with all of them, which makes giving developers direct access to the hardware largely impossible. Game consoles don't have this problem, since every PS3 or Xbox 360 is identical.
AMD's Mantle API looks to change this. It allows games that were developed for Mantle and run on a new AMD card to directly access the GPU, eliminating much of the overhead involved with DirectX. AMD claims that Mantle allows for nine times as many draw calls per second compared to other APIs, meaning that a game running on Mantle could, in theory, draw a frame nine times faster than one running on DirectX.
There are no benchmarks that prove this yet, but we shouldn't have to wait long. AMD and Electronic Arts announced a deal that would put the Mantle API into the Frostbite game engine. The upcoming blockbuster, "Battlefield 4," runs on this engine, and EA is working on adding Mantle support to the game.
The cards aren't all that great
Although AMD's new cards may be fast using Mantle, the highest-end offering basically matches the NVIDIA Titan running on DirectX. It took AMD seven months to come up with something that matches NVIDIA's offerings, and this lag is not a good sign. NVIDIA is rumored to be releasing refreshed cards in November, and new cards based on the upcoming Maxwell architecture are set to be released next year. It seems that, in terms of real world performance running on the standard DirectX, NVIDIA is still far in the lead.
Can Mantle succeed?
The problem with Mantle is that it introduces more steps in the game development process. Games must specifically include Mantle support, thus taking more time and money to develop. It's rumored that AMD spent up to $8 million to get "Battlefield 4" to support Mantle, and it's difficult to imagine many developers jumping on board.
If Mantle does attract developers, there's nothing stopping NVIDIA from developing its own API. This would have the unfortunate effect of fragmenting the gaming market, causing some games to work far better on certain cards. My guess is that Microsoft isn't thrilled about this development.
PC games today are almost universally designed around DirectX, and since this API only works on Windows it takes significant work to port games to other platforms. This gives Windows an edge over Mac and Linux-based systems, and ensures that Windows is the priority for game developers.
Mantle, being a low-level API, could presumably run on any platform. A game developed for Mantle could run on a PC, a Mac, or an Android device, as long as the graphics hardware supported it. If Mantle succeeds it could destroy Microsoft's near-monopoly on the PC gaming market, giving many a good reason to eschew Windows altogether.
Mantle has a long road ahead convincing developers to use the API, especially given the prevalence of NVIDIA cards. If "Battlefield 4" running on Mantle blows everyone away, NVIDIA may have to scramble to develop an API of its own, or license Mantle if AMD allows it. But, if the performance gains aren't enough to convince developers to support it, Mantle could fade away into irrelevance, leaving AMD as second-best for the foreseeable future.
The potential comes with risk
If Mantle succeeds, it could be huge for AMD. However, the company is currently in terrible financial shape. Years of being second-string to Intel, and now years of being second-string to NVIDIA, have taken its toll. The company recorded a loss of over $1 billion last year as revenue declined by more than 17%. If Mantle can help bring the company back to profitability, there are likely huge gains ahead for shareholders. But it may be too little too late, as the hole which AMD has dug itself into is extremely deep.
NVIDIA is in much better shape financially. The company has a huge cash reserve of nearly $3 billion, and with annual earnings of nearly $600 million the company can certainly weather any short-term loss of market share to AMD. This cash also gives the company the resources needed to innovate. In 2012 NVIDIA spent over $1.1 billion on research and development. This compares to $1.3 billion for AMD, split between the CPU and the GPU business. What's more, as AMD burns through its cash spending will necessarily need to be cut, allowing NVIDIA's technological lead to widen.
The bottom line
AMD is making big moves with its Mantle API, and if developers adopt it, the company could gain back some of the market share it lost to NVIDIA. But, it's unclear whether developers will want to spend the time needed to support two different APIs. If performance gains fall short of expectations, DirectX will remain the de facto standard. Mantle is a bold move on AMD's part, a Hail Mary of sorts, but the road to success is littered with obstacles.
