I typically read obscene amounts of topics on this board without posting. From the AEG to CRTs, from NV40 to R600, from shimmering to jaggy shadows, I lurk and enjoy the proceedings.
But this latest issue of GPU accelerated physics I am unable to stay quiet about. The majority attitude, it seems, is that Havok and nVidia has arrived to save us from the impending doom of big mean AGIEA forcing "yet another $300 component" down everyone's throats. Numerous clarifications need to be made.
Primarily, people seem to have the notion that this is Havok + nVidia vs. ATI + AGIEA. This is absolutely not the case. ATI was the first to voice promise of GPU-assisted physics. (Who did what first is insignificant and will not be discussed in this topic, BTW) Havok might as well not be in the situation at all, it is ATI vs. nVidia vs. AGIEA. Keep in mind that Havok even admits their hardware accelerated design SHOULD work on AGIEA's card if designed properly. It comes down to the fact that these are the companies who want you to buy something to do "the physics thing".
The unavoidable fact of the matter is, a dedicated piece of hardware such as a physics card should be VASTLY superior to a make-shift adaption of hardware designed for a comepletely different purpose. GPUs are superior to CPUs at this task apparently, but from what we have seen a PPU ought to blow both completely away. GPUs makers can't have that though, not when they see a golden opportunity to encourage new product adoption at higher than ever before rates. In fact, this idea of scalable, managed resources fits perfectly with ATI and nVidia's ultimate goal of everyone buying multiple GPUs.
The GPU makers will insist that GPUs are a better investment, as multiple GPUs can also be applied to increased frame-generating horsepower or superior AA. This is marketing at its worst, a desperate attempt to cash in on the physics hype by making everyone overlook the fact that additional GPUs are a much more costly and much less powerful solution. This isn't to say that the technology behind it is a waste, however. Quite the opposite, the concept of GPU accelearated physics is quite possibly what will make the other unfeasible PPU successful. See, the PPU concept want the industry to jump straight from 100% software to high-end of specailized, dedicated acceleration hardware. GPU accelerated physics allows for the badly needed middle ground between these two, the bridge that will force developers to start coding hardware-assisted physics so that the PPU can flourish. More specifically, as dual-core chips and dual GPU cards increase in numbers, we will have multiple tiers: Shared CPU -> Deticated CPU -> Shared GPU -> Deticated GPU -> Deticated PPU. This will establish a development structure that allows for scalible physics, a development structure that the PPU needs to exist.
Before I close, I would like to address anyone who still opposes a dedicated piece of hardware for physics: no possible outcome will generate a market where physics cards are required to play games. Furthermore, why do you oppose hardware functions being independent? Go buy a laptop where everything is one piece and be happy. Upgrading is for nerds, after all, right?
-An annoyed lurker
But this latest issue of GPU accelerated physics I am unable to stay quiet about. The majority attitude, it seems, is that Havok and nVidia has arrived to save us from the impending doom of big mean AGIEA forcing "yet another $300 component" down everyone's throats. Numerous clarifications need to be made.
Primarily, people seem to have the notion that this is Havok + nVidia vs. ATI + AGIEA. This is absolutely not the case. ATI was the first to voice promise of GPU-assisted physics. (Who did what first is insignificant and will not be discussed in this topic, BTW) Havok might as well not be in the situation at all, it is ATI vs. nVidia vs. AGIEA. Keep in mind that Havok even admits their hardware accelerated design SHOULD work on AGIEA's card if designed properly. It comes down to the fact that these are the companies who want you to buy something to do "the physics thing".
The unavoidable fact of the matter is, a dedicated piece of hardware such as a physics card should be VASTLY superior to a make-shift adaption of hardware designed for a comepletely different purpose. GPUs are superior to CPUs at this task apparently, but from what we have seen a PPU ought to blow both completely away. GPUs makers can't have that though, not when they see a golden opportunity to encourage new product adoption at higher than ever before rates. In fact, this idea of scalable, managed resources fits perfectly with ATI and nVidia's ultimate goal of everyone buying multiple GPUs.
The GPU makers will insist that GPUs are a better investment, as multiple GPUs can also be applied to increased frame-generating horsepower or superior AA. This is marketing at its worst, a desperate attempt to cash in on the physics hype by making everyone overlook the fact that additional GPUs are a much more costly and much less powerful solution. This isn't to say that the technology behind it is a waste, however. Quite the opposite, the concept of GPU accelearated physics is quite possibly what will make the other unfeasible PPU successful. See, the PPU concept want the industry to jump straight from 100% software to high-end of specailized, dedicated acceleration hardware. GPU accelerated physics allows for the badly needed middle ground between these two, the bridge that will force developers to start coding hardware-assisted physics so that the PPU can flourish. More specifically, as dual-core chips and dual GPU cards increase in numbers, we will have multiple tiers: Shared CPU -> Deticated CPU -> Shared GPU -> Deticated GPU -> Deticated PPU. This will establish a development structure that allows for scalible physics, a development structure that the PPU needs to exist.
Before I close, I would like to address anyone who still opposes a dedicated piece of hardware for physics: no possible outcome will generate a market where physics cards are required to play games. Furthermore, why do you oppose hardware functions being independent? Go buy a laptop where everything is one piece and be happy. Upgrading is for nerds, after all, right?
-An annoyed lurker
Facebook
Twitter