When Newton figured out his apple from his elbow in relation to gravity he caused game developers some problems to solve 350 years later. Recreating real-world physics is something that has only thus far been mimicked rather than recreated in videogames like Half-Life 2. Going the whole hog to recreate perfectly accurate physics, with all its millions of variables and potential outcomes, requires an awful lot of processing power and clever coding.
Until recently, the problem has been largely ignored, with the industry developing graphical eye candy before worrying about real-world physics. However this is gradually changing, with companies like Ageia producing dedicated physics processing units (PPUs) and developers integrating physics of varying levels of complexity into their games.
There’s no doubt that physics (or PhysX, as Ageia likes to call it… I’m told it sounds trendy) is the next big thing in videogames. We’ve lived without it thus far, but the theory is that once you’ve played a game with total real world physics, you’ll never want to go back. Playing a game where you can bring down a wall (or heck, a building) on top of an opponents head without it being a set-piece, Call of Duty style action sequence, is addictive stuff.
All is not well in physics land however. Everyone is in agreement that physics is the way forward. What the tech world at large is not in agreement about is how you should crunch all those numbers. Some, like Ageia, say that you need a dedicated physics card, like you have a graphics card, to get the best results. Others like Nvidia and ATI say that their GPUs can handle the load just fine, and that if people really want to go for the best physics they should get an SLI or Crossfire system rather than a PPU.
On the face of it, your graphics card has enough processing bandwidth to do both the physics and the graphics math at the same time. Your graphics card can run through the physics math and then render the scene as normal, and if you’re using a multi-GPU solution, like Crossfire or SLI, one of the cards can do the physics while the other renders the scene.
When I recently spoke to ATI, they told me that they’re also looking at unlocking their onboard graphic chips, which currently sit idle on the motherboard when graphics cards are present, to do the physics work. Everything is gravy, apparently.
Not so, says Ageia’s co-founder and current CEO Manju Hegde. For one, he says that the idea of using onboard graphics chips to do the physics work “doesn’t make sense. To expect integrated physics is really amazing,” he told me. “We aim physics at hardcore gamers, evangelists – who are driven by performance in every category. They would be let down by onboard physics.”
Hegde bases his entire argument for Ageia PhysX on the premise of quality. He believes that Ageia, being a company for whom physics is their only focus, has an inherent quality advantage on the physics being produced by ATI and Nvidia. In terms of physics affecting the FPS output of a graphics card, Hegde disputes ATI’s claim that having physics done off of the GPU will only result in a (relatively) marginal, 5% – 10% decrease in FPS output. He says that the good old smoke and mirrors game, so popular among the graphics crowd for more than just the good in-game effects, is probably in action here; and hypostasized that ATI was probably basing these figures off of old SDK models.
Whatever the case with regards the sheer shunting power, the feature set which Ageia is offering on a dedicated PPU puts the more simplistic models that ATI and Nvidia can currently run with to shame. Whereas ATI and Nvidia are adding basic functionality via their driver updates, Ageia says that it will be adding more and more features as we go forward. But the big whopper of a question for you and I is this: Should we shell out for a physics card when GPUs can do the work?
Sure, the Ageia cards may have the better feature set, and I don’t think any of us will argue that by being a bread and butter physics company they probably do have the technical edge; but unless there are games taking full advantage of physics, what’s the point in getting a physics card now?
Well, for one, even the ATI boys like Richard Huddy admitted that if you want to build the best and the brightest machine that can expunge every last frame per second from its bowels, getting a physics card to shunt the load is a good idea. Of course, they told me that on the basis of a minimal drain on FPS output while the GPU is also processing the physics, and I wouldn’t trust those results as far as I could toss their entire PR department without going over the fine print and seeing the tests done in our own labs.
But more fundamental than how many frames you can rescue by using a PPU is the question of games, which are taking full advantage of all the features that a dedicated physics card can offer. Unreal Tournament 2007 is one example that Ageia’s Hegde gave us. “When physics become a fundamental part of the story/game (such as in UT2007) a GPU won’t be able to handle that without a redesign of the architecture.” In other words, you’re not buying a physics card for the processing power (though it still counts) – you’re buying one for the features.
Developer uptake on all of this is very important of course, and if everyone sticks to the more basic stuff the GPUs can do and goes no further with their game development then we won’t need to buy PPUs. However this is a rather unlikely happening, and more likely we’ll be seeing many more games coming over the next year or so which will be able to take full advantage of what the likes of the PhysX card can offer, and the core evangelisers Hegde is targeting will go for the best they can get. Then, goes the logic, like a snowball turning into an avalanche we’ll all get ourselves a physics card. Who knows, maybe we’ll even get a DirectX type standard to make everyone’s lives easier in the process.
This isn’t a sure thing just yet, but Hegde tells us that he’s expecting 25 or so titles to be released before the end of the year that will take full advantage of Ageia’s physics card, and he expects well over 100 titles to be available by the end of 2007. Games like Unreal Tournament 2007, which attract the high-end system builders as it is, will help to increase the install base dramatically over the next year.
At the moment, a multi-GPU solution looks better to me than getting a graphics card and a physics processor, but I don’t intend to build or buy a new system until well into 2007 so it’s not a major worry I have right now. For those of you agonizing over a decision of whether or not to get a physics card, I’d look towards the software. At the end of 2007 I can see a PPU, from Ageia or otherwise, being a no-brainer for most system builders. But that’s a year and a half away. The water is much murkier right now, and I’d suggest even waiting until the next games rush begins in September/October to see what new titles are announced and, more importantly, what features they’ll be taking advantage of.