The Gaming Power Triangle
While the name for this concept might seem a bit grandiose, the ideas behind it are very logical and put some good ideological weight behind the PhysX chip.
In this diagram AGEIA promotes that by adding a PPU into the mix with the CPU and GPU, they are essentially increasing the power that both other components effectively have and also pushes them beyond their limits.
Since the PPU would take a large majority of the physics work off the CPU, the processor would be less bogged down and thus could address more cycles to feeding the GPU with data to crunch and keeping the game state in order. This should mean an automatic frame rate increase for any games that were significantly CPU bottlenecked before the introduction of a PPU.
Also though, there are some aspects of the PhysX chip that increase work loads on both the CPU and GPU. Imagine that you have wall of castle raining down on your filthy, peasant head in a 3D, physics enhanced Trogdor game. Now, instead of having just 30 pieces falling as you might with a CPU-based physics engine, you now have 3000 pieces falling. All the physics calculations are being done on the PPU but now the CPU has to deal with an additional 2970 objects that AI and must avoid or deal with. In the same right, the GPU now has to worry about rendering many different objects with possibly even different textures on portions of the wall that have broken away in different styles.
This will in turn (hopefully) push AMD and Intel as well as NVIDIA and ATI to find new ways to increase their computational power in order to continue to increase the realism of gaming titles.
PPU Life Cycle
One interesting aspect that came up during my discussions was the idea of a life cycle or product cycle for a PPU. In the world of GPUs, we are pretty used to seeing a new flagship product every 6-8 months from both NVIDIA and ATI. Because graphics is quite a bit more scalable than physics, this hasn’t really been an issue. But for physics programming that may require a separate physics card, the idea of having new cards out every 6 months can be gut wrenching. Would you like to be REQUIRED to buy a new PPU every 6 months for the latest titles to run on your system? Surely not.
AGEIA was adamant that this would not be the case for the foreseeable future. In fact, they have already seen the opposite occurring; upgraded firmware and drivers for the PPU has opened up new features and options to programmers on the same hardware. Cloth simulation is the first example of this as it wasn’t ready initially when the SDK went out to software developers but has since been perfected and added into physics engine, without a need for a hardware upgrade. AGEIA told me they feel that this will continue to happen through 2006 and into 2007 as the power of the PPU is being less utilized until the CPU and GPU can keep up (see the previous section).
Of course, that can’t happen indefinitely and when I asked specifically about new hardware with faster physics processing, they didn’t have much to say other than it would come, but not for “at least another year.” My impression was that it would be even longer than that but of course no one would commit to anything. It is pretty obvious though that a market like this requires a certain stability that the GPU and CPU realms don’t need.