Analyzing NVIDIA's SLI Physics
With NVIDIA and Havok publicly announcing their GPU-based physics project for the first time, the industry as a whole has taken a giant step forward in terms of realism and lifelike interaction in today's gaming environment. However, one of the most profound aspects of this new technology rests not in how tomorrow's games will look and feel. Rather, this new technology stands to dramatically change the options available to consumers with regards to new features and functionality and the necessary hardware to support them. With future NVIDIA GPU's fully supporting Havok FX at no additional cost to the consumer, the industry is now faced with a break from tradition.
Historically, developers and hardware vendors have worked closely to develop new technologies and usher in a new level of realism into the games of tomorrow. Certainly, no consumer will oppose more lifelike and realistic games so these efforts are seen as welcome advancements. However, many new advancements in software have some sort of correlation involving new hardware. In order to fully take advantage and support the latest and greatest new developments in software and coding, consumers typically are forced to go out and purchase the latest and greatest hardware. Granted, this is largely unavoidable as hardware vendors can only offer support for the current crop of software techniques and methods. However, having hardware that "supports" the latest software features is far from having games that utilize those software features. As a result, we often find flagship hardware supporting features that go far beyond the current crop of games which are available today.
The problem with having flagship hardware that supports the latest software features is that many times, that hardware will be extremely dated by the times games become available that utilize those advanced software features. In fact, by the time those advanced games do arrive there is often a new flagship piece of hardware that offers complete support for those features along with substantially faster overall performance. Therefore, the early adopters which paid a significantly higher costs for the previous flagship card are occasionally left with an empty wallet full of false promises.
With the development and pending launch of the Havok FX project, full physics processing for the latest and greatest games will be made readily available to all NVIDIA graphics card owners for no charge. Armed with the appropriate drivers, every graphics card from the mainstream level up to the current flagship GPU will be able to fully support the Havok FX architecture. This is a very interesting turn of events, as Ageia has made an entire business model out of creating specific hardware for this very purpose. When you consider the estimated additional cost of $100-$200 for a dedicated PPU physics card, NVIDIA's "free" GPU physics processing seems like a consumer's dream. Unfortunately, the choice here is far from simple with a few obvious pros and cons to each side.
Looking at NVIDIA's solution, the consumer is required to purchase two GPU's which must be run in SLI mode to "unlock" the physics processing functionality. On the surface, this requirement seems somewhat trivial. However, we must cover every aspect of each solution to truly understand their strengths and weaknesses. As such, for the sake of discussion, let's consider the consumer which has a fixed budget for a GPU purchase of roughly $299. Whereas some gamers would want to purchase a single GeForce 7900 GT and have the option of purchasing a second GeForce 7900 GT down the road, this is not a viable option for those looking to have their GPU provide physics processing. Rather, in this case the consumer would be forced to purchase two GeForce 7600 GT's and forego the possibility of a simple upgrade path through the purchase of a second GPU. However, the presence of two GeForce 7600 GT's would match or pass the performance of a single GeForce 7900 GT and also provide full physics processing in the latest and greatest games which supported the Havok FX architecture. However, this would be done at the expense of scalability as the possibility of having the performance of two GeForce 7900 GT's in SLI is obviously lost. Factoring AGEIA's likely performance advantage and NVIDIA's unknown headroom for physics in a given title, we can all agree we need more time and some benchmarking tools to even begin to declare the real value propositions of each solution, at this early stage.