Is Ageia's PhysX Failing?

[The Continental]

Tom's Hardware (http://www.tomshardware.com/index.html) takes a very in depth look (http://www.tomshardware.com/2006/07/19/is_ageias_physx_failing/) at the Ageia PhysX (http://www.ageia.com/) cards and the titles they support to see if the cards do indeed make the difference they claim. The results are interesting; so much so that they brought them to Ageia for comment.

We sent the findings to Ageia, and the company came back with some interesting comments. Apparently, it did not anticipate that the hardcore early adopters would try to figure out a way to enable gameplay without the PhysX PPU. Artificial Studios only had about four months to work with Ageia before releasing the first demos. There were certain things it did not incorporate in the software-only effects physics (non-hardware accelerated) in the game. The first is the cloth and cloth tearing simulations. In the previous versions of the demo these calculations were disabled when the game was launched with PhysX disabled.

This is a pretty good read regarding the current state and future of hardware physics processing, including ATI and Nvidia's implementations.

[DeadPixel]

I just got a new DELL XPS 400 recently and they were pushing this card with the sale. After looking at the demos I must admit I was tempted, but decided to hold off.

I think companies are trying to push the detail of graphics too much which drains your system resources from other tasks like physics. Although I could see this card taking a place on the market, I doubt it will be anytime soon. It seems that most companies are still pushing the looks factor of their games more than anything.

Just look at the how many new titles are announced promising a million features and which features are usually slashed? That's right the gameplay features go first before they start taking out all those graphic enchancements. Maybe it's because most companies develop the graphic engine first to show off the technology and by the time those gameplay elements make it in, the deadlines are up, QA does not have enough time and the publishers are pushing for the release.

[XenonCJ]

It's like I've always said.... Physics will be either handled by the graphics card or multi-core CPUs... Next thing they will try to sell you is an AI card....

[The Continental]

It's like I've always said.... Physics will be either handled by the graphics card or multi-core CPUs...

You've always said that? You probably should've patented that idea considering it was largely unheard of for physics processing to be offloaded to the graphics card prior to the recent announcement from Nvidia/Havok.

[SexualChoc]

After reading that it was a glorified graphics card, without the graphics, I lost interest. The Cellfactor demo looked damn fine, but it's no reason to buy a new piece of hardware.

[The Continental]

You can play the CellFactor demo without the card installed. It actually runs really well, hence the need for benchmarks to see what the card is actually doing.

[ZephidsEmbrace]

Very interesting read. Looks like I won't be worrying about buying a physics card for awhile to come...

[kraemer]

All they have to do is make a game with persistent world physics. Arent all of us sick of having a rocket launcher in a game where walls, windows and every "non-enemy object" is rocket and bulletproof?

All of this would change with ONE GAME. After that game, crummy static world objects just wouldnt cut it any more. The big question is, does an Ageia card have enough horsepower to support a game with all deformable geometry?

[51|RandoM]

1. $300 card
2. Games that get *WORSE* framerates using the card.
3. No compelling game physics currently exclusive to the card.
4. Very few titles with support.
5. Bad time marketwise, imho, to try get people to pay for another expensive enhancement. Good way to make consoles that much more attractive.

Physx isn't failing, it is stillborn.

[SlamDunk]

Next thing they will try to sell you is an AI card....
AIPU http://www.aiseek.com
NPU http://www.killernic.com

[Rafer]

1. $300 card
2. Games that get *WORSE* framerates using the card.
3. No compelling game physics currently exclusive to the card.
4. Very few titles with support.
5. Bad time marketwise, imho, to try get people to pay for another expensive enhancement. Good way to make consoles that much more attractive.

Physx isn't failing, it is stillborn.

You could say the same thing about the first 3d accelerators 10-11 years ago, I recall some 3d cards were slower than rendering in software.

[jBusy]

All they have to do is make a game with persistent world physics. Arent all of us sick of having a rocket launcher in a game where walls, windows and every "non-enemy object" is rocket and bulletproof?

This is not a physics issue, it's a rendering one. Spacial partitioning (http://en.wikipedia.org/wiki/Space_partitioning) is used in almost all 3d engines and it's limitation is that you cannot change the environment. There are ways around it, but it is usually very expensive.

Really, the only physics provided by an Aegia card are things like explosions and collapsing buildings. Eye candy. $300 is a lot for that kind of eye candy.

[Luagsch]

You could say the same thing about the first 3d accelerators 10-11 years ago, I recall some 3d cards were slower than rendering in software.
yes it was... and then came voodoo1 and the rest is history...

[falcon]

This is not a physics issue, it's a rendering one. Spacial partitioning (http://en.wikipedia.org/wiki/Space_partitioning) is used in almost all 3d engines and it's limitation is that you cannot change the environment. There are ways around it, but it is usually very expensive.

Really, the only physics provided by an Aegia card are things like explosions and collapsing buildings. Eye candy. $300 is a lot for that kind of eye candy.

You left out the water and cloth rendering. I only bring it up because the cloth was one area where the game chugged w/out the card

[kraemer]

This is not a physics issue, it's a rendering one.
Wow, maybe the time has come for Voxels.

[Pretzel]

Wow, maybe the time has come for Voxels.

I had a better reply to this, but my web browser just crashed, taking my comments with it. :(

Besides the technical challenge, you also have to create a lot more content to do destructable geometry. You have to have textures for the inside of all walls, doors, and windows. Buildings that were once just decorative now have to be fully fleshed out. Areas that the player had no need to go into now would have to be populated. Walls that used to cut off visibility for performance reasons would now be potential frame-rate killers.

Take a look at Red Faction. At first they were saying "It's the holy grail of gaming! You can destroy anything!" and by the time they released it, only certain walls and areas could be destroyed.

So, in the end, the big thing destructable geometry will do is add a lot more work for the artists and programmers, and in the end there would still be places you couldn't get to--they'd just look different. Instead of walls, you have cliffs, oceans, or big textures with the works "go back" on them. Or you'll have GRAW type zones where when you enter them they say "You're leaving the field of battle" and end your game when you go too far.

[jBusy]

You left out the water and cloth rendering. I only bring it up because the cloth was one area where the game chugged w/out the card

eye candy.

[kraemer]

Not when you make models out of Voxels. You can just "chip away" chunks of the model until there is nothing left. Who says every model needs a texture? Why does a white wall need a texture? Its because the lighting sucks and the textures compensate for it. The big thing missing here is Voxel acceleration, but I fearlessly predict that a voxel type methodology for 3d rendering is the future. This polygon crap was originally invented as a way to render 3d quickly. Its not neccesarily the best, just what we're used to.

[Sensei-X]

yes it was... and then came voodoo1 and the rest is history...

Yes, but to be fair 3dfx is dead and buried, having pioneered a market they never benefitted from, such is the fate of Ageia.

[Pretzel]

Not when you make models out of Voxels. You can just "chip away" chunks of the model until there is nothing left. Who says every model needs a texture? Why does a white wall need a texture? Its because the lighting sucks and the textures compensate for it. The big thing missing here is Voxel acceleration, but I fearlessly predict that a voxel type methodology for 3d rendering is the future. This polygon crap was originally invented as a way to render 3d quickly. Its not neccesarily the best, just what we're used to.

Voxels require a HUGE amount of memory. Especially if you want them to look as good as polygons. Collision detection and physics in a voxel world would be difficult. Content creation with voxels is also a risk. Dynamic lights and shadows on voxels would be difficult and slow. Animation using voxels would be crazy memory wise, so you'd still have to use polys for characters. It also doesn't address any of the other issues I mentioned.

The days of a fully voxel based renderer are long gone.

[XenonCJ]

AIPU http://www.aiseek.com (http://www.aiseek.com/)
NPU http://www.killernic.com
Hehe now that's that funny.... I knew about the NIC ones.... 1st I've heard of an AI one....

I actually think an AI one makes more sense than a physics one. I just want a brain in a jar connected to my computer via bluetooth....

[Tempest261]

I wouldn't write Aegia off yet. Doesn't anyone remember NVIDIA's NV1 (AKA Edge 3D)? You can look it up, and then look at where they are now...

[Mason]

So, in the end, the big thing destructable geometry will do is add a lot more work for the artists and programmers, and in the end there would still be places you couldn't get to--they'd just look different. Instead of walls, you have cliffs, oceans, or big textures with the works "go back" on them.
I have tried to explain that point pretty often, to little avail. Where do people think they have to go, beyond a game's "intrusive" boundaries? It puzzles me how someone can very ardently desire a thing that doesn't make sense.

Games are clever illusions, and future games will have better illusions, but true world destruction or levels without artificial boundaries will always be impossible. That said, an occasional door blown off its hinges can do a lot to keep a gamer beguiled.

[Shifteh]

Arent all of us sick of having a rocket launcher in a game where walls, windows and every "non-enemy object" is rocket and bulletproof?

All of this would change with ONE GAME.

Uh, I'm not sick of those walls. Well, to be honest, it depends on the game. If I'm playing a Quake-style multiplayer game, then no, I dont want walls that get annihilated.

For RTS', sure. I refuse to pay extra money for that though.

[CannibalKid]

I'd rather have a 3rd CPU core than another processer dedicated to a specific function which that 3rd core can do pretty well as it is...