The most intriguing part I have gleaned from Advanced Micro Devices' (NYSE:AMD) GPU2014 presentation is by far the collaboration between AMD, Dice, and Electronic Arts (NASDAQ:EA) on the "Mantle" project.
What is Mantle?
Dice's Johan Andersson describes "Mantle" as a thin driver. Typically when games are written, they are done so using what are known as APIs. API stands for application programming interface.
When these APIs are used, there is quite a bit of overhead when getting the software and GPU to work together to render frames. Bit-Tech has an interesting article on explaining the performance differences associated with rendering on consoles versus PC.
The simplest way to think about the difference is that the console is more efficient at rendering than a PC. While it is true that PCs are orders of magnitude more powerful than consoles, the overhead associated with rendering frames on a PC negates part of the PC advantage. The argument that PCs are more powerful than consoles is an incomplete argument -- PCs are more powerful, but less efficient, than consoles.
What Does This Mean For AMD?
The impact for AMD is completely dependent on the adoption rate. The motivation for adoption is the increased efficiency in frame rendering on PC. AMD's Mr. Raja Koduri states this change is the equivalent of making 2018 hardware available today.
Luckily for AMD, the first instance of "Mantle" will be used in Battlefield 4, which will be available as a patch that will be released in December. Battlefield 4 is a highly anticipated first-person shooter, and the Battlefield franchise is one of the most successful of all time. If "Mantle" is a success and delivers a truly immersive experience, literally millions of users will simultaneously take note. If this is the case, various game studios will be motivated to take a similar approach in order to remain competitive. The graphic above mentions that "Mantle" enables 9x more draw calls per second than other APIs by reducing CPU dependency. If this approach is successful, the risk studios take by not implementing a similar strategy is that their games will lose visual fidelity when compared against the competition.
Above I mention that most games are developed utilizing standard APIs. Another important point is that many games are developed first by creating an "engine, and then using this engine to create the game.
In the example of Battlefield 4, the Frostbite 3 engine is used. From the presentation, it sounds like "Mantle" is specifically designed for the Frostbite 3 engine. However, this same type of approach could be applied by various studios, allowing them to have more direct control over the hardware and program "close to the metal." "Close to the metal" is a term that relates to a programmer's level of control over hardware.
The "engine" discussion is important in that a majority of today's popular games are developed utilizing the same engines. The four major engines that come to mind are Unreal, CryTek, FrostBite, and Source. If studios choose to implement a similar approach to "Mantle," the process is simplified by the fact that many of the popular AAA titles that would benefit from the increased performance are built using game engines similar to the ones listed above. The work that goes into "Mantle" will most likely not be specific to a given game title, but instead to a game engine, meaning that the work can be reused across many game titles.
If you follow technology, you are fully aware that AMD typically loses to Intel (NASDAQ:INTC) in benchmarks that rely heavily on single threads. Part of the reason behind this is the way the traditional APIs (DirectX or OpenGL) control the flow of information. DirectX and OpenGL rely more heavily on single threads.
According to the presentation by Johan Andersson, "Mantle" will allow coders to more evenly weight the threads, which will help boost AMD's CPU performance as well.
A major cornerstone of AMD's turn-around is their focused Unified Gaming Strategy.
By being the hardware target in each gaming console, there is motivation to optimize these "Mantle"-type APIs to AMD hardware in order to reach the largest audience. A second order effect in this case is also a boost in AMD's CPU performance when compared against Intel by more equally weighting program threads. This is all made possible by AMD hardware being at the center of each console.
After the stock price dropped during the Q2 earnings call, I wrote an article stating that I was more focused on the fact that consoles would motivate programmers to code with AMD hardware in mind. To me, this was more important than the analyst focus on margins.
AMD expects a 22% increase of revenues over Q2, primarily driven by the new console shipments. Given a total Q2 revenue of $1.16B, a 22% increase puts AMD's target revenue for Q3 at ~$1.4B. If the majority of this revenue increase is derived from consoles, this represents a total revenue of ~$250M from consoles in Q3. By comparison, in Q2 Computing Solutions revenue came in at $841M, with Graphics and Visual Solutions revenue at $320M. For me, the possible boost in AMD's larger and higher margin businesses associated from the synergy of AMD being the de facto hardware standard for consoles was far more important than the margin focus. Today's announcement further strengthens the argument I made in July.
One last benefit I would like to mention to this approach is that it increases the value of the APU. If these specialized APIs can drastically increase performance, this means APUs should be able to deliver stronger performance when used specifically for gaming.
A possible complication would be the inclusion of non-AMD hardware in Valve's pending SteamBox design. There also seems to be no barrier to entry for developers to take a similar approach with Nvidia (NASDAQ:NVDA) hardware. However, given AMD being at the heart of both the PS4 and Xbox One and the market this represents, I feel AMD is in the best position to take advantage of this trend.
Jon Peddie Research puts the PC gaming hardware market at ~$18B and growing (unlike the PC market, which appears to be stagnating). Compared against AMD's most recent quarterly revenue of $1.16B, growing market share through the use of these specialized APIs represents a potential large growth driver for AMD going forward. The first indication we will have if this approach has hopes at being successful will be if Battlefield 4 has sizeable performance increases running on AMD hardware after this patch is applied in December.
More details will be available on "Mantle" at AMD's APU 2013 event.
Special thanks to Joel Hruska for pointing me to the Bit-Tech link!
Additional disclosure: I am long AMD in both options and shares. I actively trade my AMD and INTC positions, and may add or liquidate shares at any time.