Richard believes that AMD have made the right choice in supporting Bullet physics as their solution. He says "it's clearly going to benefit from being open sourced, and it benefits from the fact that it's being added to by a platform company - we don't care whether the GPU or the CPU wins. What we do is deliver value." He also believes that the open-sourced nature of Bullet prevents it from focusing on benefits for one manufacturer's technology. "We couldn't get away with putting shoddy code in there and calling it ‘really good'". When asked about how AMD see Havok, Richard stated "Havok isn't in the guilty situation that PhysX is".
However, he also pointed out that "almost all games [that use PhysX] run on the CPU, and only those [that are accelerated by GPU] will work with NVIDIA technology". Clearly a sore point, Richard explains that the free nature of Bullet is bound to prove its success. "It's not our plan to pay people to put physics on the GPU, we want them to have a free choice. Because it's open source I imagine that [we'll] take our gaming program and wrap it around games. It'll move towards the open standards; PhysX is unlikely to be embraced by the whole industry."
Ashu and Nadeem have made it clear that there's no money being exchanged in return for PhysX implementation in games, but they defend their proprietary solution fiercely. "If we weren't hands-on [developing PhysX] we wouldn't be able to innovate at the rate we do - our competitors are jealous of that because they're not working with games developers. If you look at the features and technologies they're introducing, they're a few generations behind."
The huge support for PhysX is clear, but it's also rivalled by Havok. Ashu explains that "Havok is a proprietary physics package sold for a decent chunk of change. They do not give it away for free; they only give away the binaries. There is no standard for rigid body behaviour, and there is no game engine standard." When asked about the danger of Bullet encroaching on PhysX's market share, he remarked "when is this awesome solution [Bullet physics] expected to actually be seen?"
Richard refused to let that comment sit in the open without an answer, and explained the future of Bullet. "I can't preannounce any titles that use [Bullet], I know at least three developers - one of which is committed to launching this year. It's the third most common physics engine which is used in games." AMD's support of Bullet is certainly impressive, and on top of development support, AMD have also arranged for Pixelux's Digital Molecular Matter engine to be incorporated into the PC version of Bullet.
"It's a very complicated piece of maths that allows the game to smash things up without scripted rules - proper glass and metal simulations. It's a rather beautiful implementation, license free on PC, and runs only on DirectCompute". Again placing emphasis on developer choice, Richard highlights that "one way or another we've arranged that it is license free - there is no coercion". We gave Richard the chance to fire back a question at NVIDIA, and he asked "Is there a single GPU accelerated game that uses PhysX that is not a part of NVIDIA's marketing program, that implies effective payment from NVIDIA?"
Ashu responded with "Instead of wasting time doing PR stuff, they should do stuff that actually helps games come out. When AMD had DX10.1, did they not pay developers to put out DX10.1 content? Have they not been doing similar programs with developers for DX11?" While we can't confirm that AMD had been supporting DX11 developers, Nadeem mentioned something that was quite unexpected: "If we found a title that used GPU physics that we did not know about, we'd be fired!"
Whether or not this is hyperbole seems unclear, but Ashu finally answered Richard's question directly: "PhysX isn't an application; the physics subsystem can support a multithreaded [CPU] environment. It's up to the game developer to decide how many threads are used." However, when prompted further, Ashu admitted that "the majority find they have a single thread dedicated to physics".
Issue: 137 | June, 2012