Moving forward we saw the same thing bite ATi/AMD when they used the VEC5 architecture in their GPUs. This was an architecture with smaller, vector, rendering in mind. If the game or video application was not optimized for this workload, AMD GPUs suffered. NVidia, on the other hand, went another direction. They opted for more general purpose shader processors so they worked better across the board (at the time). AMD learned their lesson and change the basic design of their GPUs, but they still held onto the idea of optimize code for their platform and after much development work they launched Mantle.
Mantle is a low level API that allows a game (or other application) to make full use of the power in an AMD GPU. The scope of what Mantle does is beyond the intent of this article so we will not get into too much detail here except to say that depending on the GPU in question Mantle can improve performance by as much as 50% in a given game. NVidia had this type of API before AMD launched Mantle although much of it was in separate APIs. Now they have combined it into a single API to get the same advantage on their GPUs as AMD’s Mantle is giving.
Both companies have been hawking their respective platforms in an attempt to sway the gaming industry one way or the other. AMD has much more to lose in the fight, but we are not seeing NVidia slacking off either. Also in the game is Microsoft’s DirectX and OpenGL, both are low level API, but are not hardware specific both AMD and NVidia also code with these APIs in mind.
Now there is a third party who has recently entered the GPU market in a big way: Intel. With the inclusion of a GPU in every “non-E” CPU, Intel has become one of the largest GPU sellers in the world. This does not mean that these GPUs are in use, but that with almost every CPU sold you are getting an Intel GPU so sales are high. Utilization, on the other hand, appears to be mediocre at best. AMD and NVidia are still looked to when you want to really game with Intel graphics being seen as good enough for basic gaming and business needs.
From most tests performed on Intel GPUs it is clear they need some help with performance and it looks like they might have turned to AMD to get some. According to Intel they would like to take a look at Mantle to see how it helps to improve performance. This is most likely a low level look to get an idea what Mantle does rather than an attempt to adopt the same technology. AMD also confirms that this request from Intel is real, but that Intel will have to wait until Mantle is publicly released.
Now that the news is out, Intel has made a statement that this is merely exploratory and that they are still banking on DX and OpenGL to ensure that they perform as expected. They also hope that DX12 will give them the performance boost they need without requiring their own hardware specific API to get the job done.
Intel could also be checking out Mantle for their push into mobile technologies as they try to gain ground on ARM based products. Having a low level API to improve graphics while keeping power consumption down would be a very big benefit to Intel at this stage of the game. Again this is probably not about adopting Mantle, but in seeing how it works and what can be done by Intel to develop their own API exclusive to Intel GPUs.
It looks like the next battle ground might not be in the GPU, but in low level software APIs designed to squeeze every bit of performance from the GPUs in use. Now let’s hope that game developers will use these tools to give us better games in the future.
Tell us what you think in our Forum