ATI v AGEIA: who is stronger, who is faster?
It is primarily about to revolutionize the physical engines of computer games about physical accelerators we can read in more and more places as we have to wait less and less to actually appear. But what does ATI think of all this?
We've had almost tangible information about the first solution based on the Ageia PhysX processing unit for a couple of weeks now: ASUS cards based on this core are almost complete, you can see quite a few pictures in our previous news.
Because the graphic quality of advanced computer games is already at an astonishingly high level, more and more people are dealing with other details when judging appearances. Thus, the physical engine of a program, i.e., the set of codes that determine how each element of the playing field interacts in different situations, is increasingly a central issue. However, these are incredibly computationally intensive tasks that cannot yet be performed in sufficient quality, primarily due to the limitations of machine performance. It’s no coincidence that the name AGEIA, which promises a solution with a target processor specifically for these operations, is becoming more and more popular, with more and more people waiting to see what’s going on. However, ATI's opinion on all this is very interesting, as they see things surprisingly differently.
In their view, their graphics controllers are already able to effectively model physical phenomena, and it is only a matter of time before they make this feature available on their widely shipped cards.
One of ATI’s leading designers explained that even the world’s first DirectX 9 graphics accelerator, the Radeon 9700 Pro, is capable of processing vertex blocks, a operation very similar to that of the AGEIA solution. Data arrays are processed by the pixel shading unit of the graphics core. According to him, the only obstacle for today's controllers to take over the physical computing tasks as well is the fact that the 16 parallel pipelines are not enough for this purpose. But as we know, in the future, the Pixel Shader and Vertex Shader will be merged according to Microsoft specifications, meaning that one and the same part will be responsible for both types of operations. In addition, a good example is that the Xbox 360 graphics core, the R500, has 48 units capable of processing arithmetic and logic operations, so parallelization is becoming more widespread.
According to ATI, in addition to the apparently more cost-effective implementation, graphics cards by their very nature have a major advantage: after processing, they have much more bandwidth to transmit data than a physical destination controller can ever get.
A much more important practical issue is that if a widespread physical engine, such as the one underlying the more than 40 reputable games SNOW will begin supporting this form of physical computing, the company will not need to develop another programming interface to allow developers to harness the power of GPUs for this purpose.
It is a difficult question to judge the evolution of developments in the light of this new information, as on the one hand we have no idea how a stand-alone physical accelerator can compare to using a graphics controller for a similar purpose, but we don’t even know what Ageia PhysX’s performance should do. will suffice for example Unreal Tournament 2007 in case of. Assuming that the two lines are capable of about the same computing power, it’s still hard to decide which is better for us: a more colorful, innovative, but more complex market, or a more wallet-friendly but undoubtedly more cloudy solution, that is, if the good old ATI and NVIDIA will take over this role as well.
