Havok physics are handled in software and calculated on the CPU, rather than offloading calculations to the GPU like PhysX.... Havok is effectively stepping on their business by pulling off dynamic deformable objects and doing it at a software level.
What does this mean, exactly? Phys-X calculates physics interactions on a chip. Havok calculates physics interactions on a chip. That they do it on different chips doesn't make one any more of a "software-level" solution than the other, does it?
If I've got it wrong, please explain, but near as I can see it, they both do the same things, but you're acting like Havok being able to do it on the CPU is a major trump card over doing it on another GPU or PPU, and I don't see how it is.
Well, I'm not saying that being able to do physics on the CPU is a trump card, but what I do mean is that by Havok not being hardware locked (i.e. NVIDIA only), they are set up for a significantly more broad exposure and adoption.
Havok has been in place for awhile. Many games use the tech, and people know what it is. Any game can use Havok, and any hardware can run it. Havok has lagged behind PhysX in ability and features, but it has been WAY ahead of PhysX in terms of adoption. PhysX representation is terribly weak right now, and I really don't see it improving any time soon. So if we can have physics near or at the level of PhysX (like the videos for Havok Destruction display), and have the accessibility of the engine running at a software level and not relying on a specific set of GPUs, then everybody wins.
Assuming it can do it efficiently and look this good, I like the idea of physics running on the CPU. I have 8 freaking threads with over half of them sitting idle while gaming. Meanwhile, the GPU is getting hammered; why give it yet another task?
Isn't a GPU more already more developed for this sort of task, though? I thought I remember reading something about how GPUs would always be better performers than a CPU ever would be - which is why people buy like 9800GT's to run physics only alongside some other nVidia offering in their rigs.
Better, maybe, but until everyone gets over their obsession with winning the GPU physics war, only half of the market will get to enjoy it on any given title.
In general PhysX will win as older Nvidia cards are replaced more systems will be PhysX ready. Also if a game runs on Nvidia PhysX and a GPU is not present the PhysX will run off the CPU just like Havock.
Well i always loved havoc, it has the best "felling" of the bodies. Their tech is awesome and i'd love to see it on my own screen and play with it. BUT Physx runs on the GPU so it's much faster and it can do all this stuff really fast. Just watch some Dark Void demos, or run Water demo. It owns.
Also i now that HAvoc is suppsoed to work on ATI stream?
Also(2) with Open CL it won't matter what gfx card you have, soon both engines will run on our gpu's.
My love and adoration for real-time physics and cloth simulations in games has only gone up after seeing these demos. I'm not sure if I've witnessed them before, but I think this is the first time I've seen dynamic deformations of things (the steel drum being crushed).
Also, another issue I'm hoping to be relieved with the cloth sims is the following: it used to be that characters with props like swords or clothing like cloths would have them clip through the character model. For example, a character with a cape may have the ends of the cape pass through the character's legs while the cape is having a "blow in the wind" animation, or a character running with a giant sword slung over their back will have part of it move through their body on the run animation. This has always irked me... and I'm hoping that physics can help solve realism issues such as those I just described.
Destruction never looked so good. Advances like this would get me gaming again. Just wish console gaming would die off so faster advances on pc gaming would be made. I know, will never happen, but I can still wish.
For example, a character with a cape may have the ends of the cape pass through the character's legs while the cape is having a "blow in the wind" animation,
Cloth clipping drives me up a wall as well, but accurately presenting occluded cloth dynamics is a very intensive thing to do, especially when you're talking about games where animations can change dynamically in any instant. For example, a 3rd person fighting game where you can change attacks mid-move. The animations will swap and blend together, which can leave the cloth sim in a compromising situation.
Batman Arkham Asylum has some really impressive cape cloth. I haven't spent a ton of time with the game as I don't own it, but I'm fairly certain it does a very good job of keeping the cape from clipping through the character's body.
Also if a game runs on Nvidia PhysX and a GPU is not present the PhysX will run off the CPU just like Havock.
It's my understanding that this is not the case... I thought that without a PhysX compatible card, the PhysX effects were stripped out of the game altogether.
PhysX can run in software and hardware.
If the game strips it if you don't have a PhysX compatible card then it's the game's choice.
Also, PhysX is free for developers (obviously not for commercial stuff) which is sweet for people who wants to do open-source stuff with it.
And x86 sucks, so software physics will also suck (in performance) compared to hardware physics.
CUDA is way faster, especially when it comes to math, than x86.
Dev's are not really lazy.. It is more how the tech will run with other tech within the games that can put a hold on implementation of stuff. You can tell which studios have the money to put behind games for proper development and proper tools and which don't. Overall its a huge deal for a studio to step into new tech as most times they have so much money wrapped up into the tech they have its hard to convince the money people to allow them to drop that tech. It takes months to build out new scenes and models to show off to get approval, let alone revamping an entire game to include the tech.
That's why the sandwich (or cake) technique is important, you can easily switch out an underlying layer to something else that you like more.
Making new (rought) models and scenes (or maps as I prefer to call them) takes a few hours, at the end of the day you can have a small prototype to show off (if you've got the tools for the job).
After that, it takes some months to refine it into something that reach a level of quality that the market can accept.
I can by no way say that I'm a pro at any one tool, but I use what I know of each one and combines them the best way I can.
Comments
What does this mean, exactly? Phys-X calculates physics interactions on a chip. Havok calculates physics interactions on a chip. That they do it on different chips doesn't make one any more of a "software-level" solution than the other, does it?
If I've got it wrong, please explain, but near as I can see it, they both do the same things, but you're acting like Havok being able to do it on the CPU is a major trump card over doing it on another GPU or PPU, and I don't see how it is.
Havok has been in place for awhile. Many games use the tech, and people know what it is. Any game can use Havok, and any hardware can run it. Havok has lagged behind PhysX in ability and features, but it has been WAY ahead of PhysX in terms of adoption. PhysX representation is terribly weak right now, and I really don't see it improving any time soon. So if we can have physics near or at the level of PhysX (like the videos for Havok Destruction display), and have the accessibility of the engine running at a software level and not relying on a specific set of GPUs, then everybody wins.
Hope that makes more sense.
Also i now that HAvoc is suppsoed to work on ATI stream?
Also(2) with Open CL it won't matter what gfx card you have, soon both engines will run on our gpu's.
Also, another issue I'm hoping to be relieved with the cloth sims is the following: it used to be that characters with props like swords or clothing like cloths would have them clip through the character model. For example, a character with a cape may have the ends of the cape pass through the character's legs while the cape is having a "blow in the wind" animation, or a character running with a giant sword slung over their back will have part of it move through their body on the run animation. This has always irked me... and I'm hoping that physics can help solve realism issues such as those I just described.
Oh, and help make explosions look cooler.
Cloth clipping drives me up a wall as well, but accurately presenting occluded cloth dynamics is a very intensive thing to do, especially when you're talking about games where animations can change dynamically in any instant. For example, a 3rd person fighting game where you can change attacks mid-move. The animations will swap and blend together, which can leave the cloth sim in a compromising situation.
Batman Arkham Asylum has some really impressive cape cloth. I haven't spent a ton of time with the game as I don't own it, but I'm fairly certain it does a very good job of keeping the cape from clipping through the character's body.
It's my understanding that this is not the case... I thought that without a PhysX compatible card, the PhysX effects were stripped out of the game altogether.
If the game strips it if you don't have a PhysX compatible card then it's the game's choice.
Also, PhysX is free for developers (obviously not for commercial stuff) which is sweet for people who wants to do open-source stuff with it.
And x86 sucks, so software physics will also suck (in performance) compared to hardware physics.
CUDA is way faster, especially when it comes to math, than x86.
It might be good for them to finally implementing stuff that has been around for ages.
I saw fully destructive objects and realtime clothing last century...
It's just that the devs are lazy, have a look at this page (last updated 2002): http://freespace.virgin.net/hugo.elias/models/m_cloth.htm
Making new (rought) models and scenes (or maps as I prefer to call them) takes a few hours, at the end of the day you can have a small prototype to show off (if you've got the tools for the job).
After that, it takes some months to refine it into something that reach a level of quality that the market can accept.
I can by no way say that I'm a pro at any one tool, but I use what I know of each one and combines them the best way I can.