The road to GPU physics acceleration on video games has been a long and not exactly rewarding journey. Even though Nvidia has been pushing its PhysX engine a great deal lately, only a handful of titles currently support it, while a majority of developers sit on the side wary of adopting proprietary technology. AMD on the other hand partnered with Intel-owned Havok to improve the way its graphics chips handle physics, but we've yet to see any significant development from that camp.

Read the whole story

Take that nVidia! for being cheapskate for your PhysX

Heh, love those anonymous ignorance posts... So what if nVidia paid big money to acquire PhysX, and has incorporated that proprietary library into their GPU to eliminate you needing to buy an expensive separate PhysX processor card (like you had to when it was Aegia)... YOU are obviously WAY more important than them making money to recoup their investments and stay in business, so they should just give it away to anyone who wants it, right? Seriously, get a clue.

That said, way to go AMD! I like the move, it tends to even the playing field (in the long run), and a widespread standard that game developers can target, rather than having to individually program for each of the specific physics engines, will make development cycles easier. It would definitely encourage more widespread adoption of in-game physics effects. I'm sure it will annoy nVidia and Intel (if they are getting royalties now from Havoc), but that's just tough noogies!

I myself think the whole 'physics' thing is still pretty dubious, but having an open initiative instead of all these proprietary technologies would be much better for both consumers and developers. Here's hoping there'll be similar openness in the future for technologies like CUDA.

I was more than a little skeptical about physics processors, particularly when the Aegia unit was unveiled - why spend $300+ more on a board that gives you features but nobody uses it? Then I saw a few interesting demos on the possibilities of physics in games, and my interest was aroused. But the clencher was when some friends of mine (game developers) showed me a little bit of what they called the "shrapnel proof of concept" routine they had played around and developed using the nVidia PhysX engine. Basically, in a shooter now, if you are in a position and a grenade goes off within a set distance, you receive damage, and the damage level is usually factored by the proximity to the blast center. My friends then showed me the PhysX version, where a grenade went off nearby, shredded a few crates, and the shrapnel flew through the air, whistling past their character, with only a small chunk hitting it, giving very slight damage. Same grenade again, this time without the boxes in the way, and grenade chunks ripped through the character, more damage. Same grenade again, this time behind a barrel, and the barrel is blasted right at the character, smacking into him and sending him flying, doing even more damage... It was very cool, seeing the interactions possible through the physics calculations, which my friends pointed out would be nearly impossible to accomplish effectively if the CPU was in charge of that, along with everything else it already does. It definitely sold me on the concept,

All the physics engines are OK and i don't really care about physX b/c it is the past have any of you seen euphoria/havok with dmm. Look at the game backbreaker it looks sick just because of the fluent physics. (Please don't flame at me if I don't know what I'm talking about.)

This is a very good news. We need an open engine, that can be used by everyone; this is the future, not physx.
I hope to see some concrete results soon