The Scientists\' Opinions on Gaming Physics

[pschmid]

Ageia recently released its PhysX card to the public, and it didn\'t take long for Nvidia and ATI to in turn present their own take on how to best calculate physics. There are some, however, who say that none of these methods is especially good for the task. According to physics scientists, there are just no algorithms good enough to yield satisfying results.

Speak out in the Toms's Hardware reader survey!

 

[quantumsheep]

Ageia recently released its PhysX card to the public, and it didn\'t take long for Nvidia and ATI to in turn present their own take on how to best calculate physics. There are some, however, who say that none of these methods is especially good for the task. According to physics scientists, there are just no algorithms good enough to yield satisfying results.

Speak out in the Toms's Hardware reader survey!

I don't get that :S.

Surely Ageia has it's own physics scientists that say there are algorithms that are more than adequate for the task of processing physics.

One side has to be wrong!

 

[torque79]

did you read the article? you're over simplifying. they talked about how some physics functions are more simple than others, and therefore require less hardware horsepower to perform properly.

there is no "right" or "wrong" solution in a problem with such varying questions and requirements. what is "adequate" to a scientist? how about to a gamer? to a hardware manufacturer? how about a hardcore gamer with a $5k rig, vs an occasional gamer with a $500 one?

neither side is "wrong", THG is just providing the viewpoint of a few different scientists to give you more multifaceted information than the marketing "information" you get from Nvidia, ATI and Ageia.

 

[TheTallGuy]

I believe it, its much easier (and faster) to linearize differential equations than to solve them directly. It only makes sense that Ageia would take that approach because to make use of the full algorithms would be useless for gaming because of the slowdown.

 

[TechnologyCoordinator]

One thing people must realize is this. The physics card isn't offloading from the CPU as much as it is adding a completly new element. Perhap Ageia, ATI, and nVidia are all wrong. The best way may be throwing more cores. Imagine the power of a Quad-Core Core 2 processor!?

 

[quantumsheep]

I got owned.

 

[engrpiman]

Here is the solution. Make a Physics card that has a licensed version of the Maple (maplesoft) math engine. by doing this you would allow developers to bypass lengthy coding and just pass simple strings of math to the PPU to be calculated.

so if you want to solve a system of equations you send the string

solve( 'equation 1', 'equation 2', ... 'equation n')

and the maple engine along with the hardware calculated the solution . it would make vector math and linear algebra easy and would reduce the overall bloat of the game

the hardware exists , so does the software why not put them together to maple a ppu

 

[quantumsheep]

Here is the solution. Make a Physics card that has a licensed version of the Maple (maplesoft) math engine. by doing this you would allow developers to bypass lengthy coding and just pass simple strings of math to the PPU to be calculated.

so if you want to solve a system of equations you send the string

solve( 'equation 1', 'equation 2', ... 'equation n')

and the maple engine along with the hardware calculated the solution . it would make vector math and linear algebra easy and would reduce the overall bloat of the game

the hardware exists , so does the software why not put them together to maple a ppu

Cos corporations are stingy and don't want to "waste" money on licensing things!

 

[sviola]

Here is the solution. Make a Physics card that has a licensed version of the Maple (maplesoft) math engine. by doing this you would allow developers to bypass lengthy coding and just pass simple strings of math to the PPU to be calculated.

so if you want to solve a system of equations you send the string

solve( 'equation 1', 'equation 2', ... 'equation n')

and the maple engine along with the hardware calculated the solution . it would make vector math and linear algebra easy and would reduce the overall bloat of the game

the hardware exists , so does the software why not put them together to maple a ppu

You know that would be extremely slow, don't you?

 

[Zok]

"In the gaming industry, physics seems to be the new focal point."

I disagree. Maybe all the gaming hardware manufacturers are telling us that we need more realistic physics in our games so we can spend more money, but in the end, more realistic physics are not what's going to make a game great, because they don't add gamplay value. Sure, it's a neat sidenote to be able to blow stuff apart and have it be realistic when it does, but when you're playing the game and involved in whatever task is at hand, you really don't care.

It would probably be cool to have realistic physics in some games like Grand Theft Auto that have a lot of freeform gamplay value, because just running around and blowing stuff up is a fairly fun part of the game. But what about a game like World of Warcraft, which is probably more similar to the big games we'll be seeing in the next 5-10 years than anything else? That's very freeform as well, but nobody cares that you can't go around blowing stuff up. Especially since you would blow it up for everyone else too, which wouldn't work at all. And what about something like "Madden/NCAA Football", hugely popular games on the consoles? What good would better phyics do for them? nothing.

And sure, Warcraft could use a little better collision detection and there are definately other nuances in the physics that could be fixed, just like any other game. But these things certainly don't require a whole new PPU. The only thing I see a PPU being good for is taking developer's attention off the gameplay, which is really the deciding factor in whether a game is fun or not, and spending the time instead on some silly physics that people spend 30 seconds looking at and going "thats cool" and then going on to play the game.

Just my 2 cents.

 

[TheTallGuy]

And what about something like "Madden/NCAA Football", hugely popular games on the consoles? What good would better phyics do for them? nothing.

How could improved physics not help a game with 22 bodies colliding simultaneously and an oblong ball being batted around?

The game gets more "realistic" in every iteration because they just program in more things like gangtackling, but its not real. Have you ever seen what happens when on a fumble? The ball bounces around like its landing on firecrackers. I think real physics in sports game would be spectacular. Along with virtually everyother game.

 

[SciFiMan]

Just plain wrong, and gameplay IS physics (the story line is different). Physics in games is much more important than pretty graphics- it's probably the most important thing, and the hardest thing to do well. That's why the developers need new hardware help. Most people think it's just stuff looking like it's blowing up but that's wrong. Most of that is just pre-scripted events because they don't have access to real physics calculation horsepower. The building will "blow up" exactly the same way every time no matter where the missle hits it or how strong the missle is. Physics includes the whole "world" in the game. Gravity is applied to objects like jeeps, bullets, people, so they are drawn towards world-center. So when your character jumps, you come back down and don't behave like superman or a jet. Like the article said, it's simple to make a flat wall look textured like a stone wall. But once something happens to that wall you need physics. And the more the better.

Right now cloth doesn't behave ingame like cloth very well, nor does water. Empty soda cans on the ground don't even move, let alone crush when stepped on. Physics would let me pick up an individual can and toss it in a room to distract the enemy so I can flank them. Or move/stack boxes so I can climb on a roof. Some games have varying degrees of these things. Bottles and shelves move in FEAR but you can't do anything with them. Let me pick up a flare in one room, then 3 rooms later find a gas can, and let me make a IED if I need too.

Last time I threw a football and it hit the ground it randomly bounced all over. Right now in a computer football game the football behaves as if it was round, last time I saw. Of course they have to make it unrealistic to a degree so the gamer can easily get the ball, especially on consoles. In every game except maybe checkers, physics is very important but can't be done well yet. Even the original Pong game had a basic physics.

 

[AeroEngy]

And what about something like "Madden/NCAA Football", hugely popular games on the consoles? What good would better phyics do for them? nothing.

TallGuy beat me to it.. But you have to be kidding right. The better the physics the better the football. Guys cracking into each other, the ball bouncing off of helmets and getting popped out of a guys arm if the collision is just right. There are all kinds of ways that realistic physics could improve Madden.

 

[SciFiMan]

pschmid, WHY do you keep starting physics/graphics discussions in the Memory section?!

 

[sviola]

"

But what about a game like World of Warcraft, which is probably more similar to the big games we'll be seeing in the next 5-10 years than anything else? That's very freeform as well, but nobody cares that you can't go around blowing stuff up. Especially since you would blow it up for everyone else too, which wouldn't work at all.

I disagree with you. I think real physics (altough would be tough to simulate real physics on spells lol) would add a lot of depth to MMORPGs. You could have new gameplay, never done before (I won't give examples as I'm working on the development of one of these to make use of physics and we intend on bringing a whole new level of interaction and gameplay to our game).

Also, in sports it would be really awesome to have real physics...imagine a formula 1 game with physics? What about a soccer game (no more fifa 94 cheat goal places lol)? And other games like flight simulators? The options are so many...

And this is just the first step to a more immersive entertainment system, and if include here innovations like the Nintendo Wii controler and virtual reality the next 10 years look very promising. (Got a little carried away here, but what the heck, it sure would be fun to play a game like battlefield 2 with virtual reality).

 

[sviola]

And WHY do people keep starting physics/graphics discussions in the Memory section?!

Blame the reviewer!!

 

[BGP_Spook]

The tricky part in this article that maybe was missed was that Tom's asked scientests if the PPU(or the like) would make real physics.

The reason why this is important is that you can make algorithms that produce results that are very close to real(as close as you want or need them to be) but are only valid for some range of a variable(like time or distance or mass....). The real advantage to these types of algorithms is they are very quick and easy to calculate(for a computer anyway).

The real physics that scientests need are not done that way(generally) though.

The difference would be like comparing algorithms that arrive at 2.49999999999(almost instantly) vs. 2.5 (exactly but took 50-100 times as long).

The first number is a very good approximation of the second.
It would work well in a game for describing something and you would likely not notice the difference, provided that the range of the algorithm was not exceeded.
In scientific work that kind of error in calculation is unacceptable if it can be avoided.

 

[mpjesse]

Just plain wrong, and gameplay IS physics (the story line is different). Physics in games is much more important than pretty graphics- it's probably the most important thing, and the hardest thing to do well. That's why the developers need new hardware help. Most people think it's just stuff looking like it's blowing up but that's wrong. Most of that is just pre-scripted events because they don't have access to real physics calculation horsepower. The building will "blow up" exactly the same way every time no matter where the missle hits it or how strong the missle is. Physics includes the whole "world" in the game. Gravity is applied to objects like jeeps, bullets, people, so they are drawn towards world-center. So when your character jumps, you come back down and don't behave like superman or a jet. Like the article said, it's simple to make a flat wall look textured like a stone wall. But once something happens to that wall you need physics. And the more the better.

Right now cloth doesn't behave ingame like cloth very well, nor does water. Empty soda cans on the ground don't even move, let alone crush when stepped on. Physics would let me pick up an individual can and toss it in a room to distract the enemy so I can flank them. Or move/stack boxes so I can climb on a roof. Some games have varying degrees of these things. Bottles and shelves move in FEAR but you can't do anything with them. Let me pick up a flare in one room, then 3 rooms later find a gas can, and let me make a IED if I need too.

Last time I threw a football and it hit the ground it randomly bounced all over. Right now in a computer football game the football behaves as if it was round, last time I saw. Of course they have to make it unrealistic to a degree so the gamer can easily get the ball, especially on consoles. In every game except maybe checkers, physics is very important but can't be done well yet. Even the original Pong game had a basic physics.

I agree 100%. I think a lot of people are underestimating the importance of realistic physics. When you start getting into completely interactive environments, physics is probably the most important factor. I think half the reason HL2 is so great is because of its physics engine.

The best analogy they made in the article was the 'building collapsing under its own weight" analogy. Naturally, no game can do this (yet). There are some games that simulate it (with scripts and only 1 outcome). But in all games, you just can't simply walk up to say... a water tower and blow the legs out from underneath it.

How cool would that be? You've got some NPC's sniping you from a tree or tower and to kill them you shoot an RPG at the base. Then the tree or tower collapses realistically. That's the kind of stuff that makes a game cool. And it can only be done with some sort of accelerated physics.

 

[sviola]

The tricky part in this article that maybe was missed was that Tom's asked scientests if the PPU(or the like) would make real physics.

The reason why this is important is that you can make algorithms that produce results that are very close to real(as close as you want or need them to be) but are only valid for some range of a variable(like time or distance or mass....). The real advantage to these types of algorithms is they are very quick and easy to calculate(for a computer anyway).

The real physics that scientests need are not done that way(generally) though.

The difference would be like comparing algorithms that arrive at 2.49999999999(almost instantly) vs. 2.5 (exactly but took 50-100 times as long).

The first number is a very good approximation of the second.
It would work well in a game for describing something and you would likely not notice the difference, provided that the range of the algorithm was not exceeded.
In scientific work that kind of error in calculation is unacceptable if it can be avoided.

Well, in most scientifical research, an error of the magnitude of your example would be negligible, but I get your point.

I think that there is a mistake (not sure if thhis is the right word in this case) in the arcticle: many times it is said that there's few information on how the PhysX works, but nonetheless, one of the interviewees claims he wouldn't use it because "physics algorithms are locked into the hardware" it seems to me quite controversial. If few is know about the architecture of the ppu, how can he state such thing and dismiss the solution as whole (that's the kind of thing that makes physicist be shuned by engineers). Other thing to add here, is that the PhysX can be upgraded by software as stated in Ageia's website, so maybe some customization could be done for the complainer .

For not real physics being able to be done in using PhysX:
http://www.microsoft.com/presspass/press/2006/jun06/06-20MSRoboticsStudioPR.mspx

 

[enewmen]

I don't care about speed yet.
I just want to see accurate physics! You know, get simple object collisions right first?? I still see hard objects poking out of other hard objects. This does not take 1000 gflops.

 

[ninjahedge]

Here's thething, physics is important in just about everything that is meant for entertainment.

From animation to gameplay to movies. Good examples:

Animation:

Although Disney has some of the highest quality animated cells out there (Pocahantas, Lilo and Stitch, Beauty and the Beast) its physics is DEPLORABLE! Some of their series such as Gargoyles had some excelent still frames and character animations, but the guys looked like they werehanging on strings when they flew.

OTOH, you go for something liek Batman or Samurai Champloo, and you get some incredible physics, albeit stylized, than make you feel when someone gets hit or a table gets knocked over.

Movies:

One of the WORST thngs I see in movies is ignorance of action-reaction. Movies like the hulk stank because of little things like giant metal balls that bounced off thin steel railings and the fact that the hulk did not leave craters whenever he jumped (he did when he landed, but what about the force needed to get him in the air?). A lot of people do not notice these things, but you get more of a realistic feel to a punch when it is done right.

Movies like Bulletproof Monk and, god forbid, True Lies ignored physics so much it was horrible. You never got the feeling that you were watching through a window to a fantastic story, but that you were watching a movie.

Games:

Here's the biggun.

Things I liked in FEAR that made it more enjoyable were little things like lights that could sway on cords when you hit them. Ragdoll physics for teh bodies, and objects that could be manipulated.

But as much as I like that, there are still som many gaps.

Examples:

-Graphic plants that do nt really move when yuo go through them, and offer no cover when you are behind them (see Oblivion).
-Chainlink fences that can stop a bullet, or rocket from going through them.
-Wooden doors that are somehow impossible to break.
-Explosions that only effect some items in the area (why don't windows shatter all the way down the block? And what about the concusion wave?)
-Vehicles that have little spin, or impact when driving on games and hitting things. The worst are vehicles in FPS's.
-The inability to USE things in the environment, like chairs, boxes, lamps. Or to build actual things like tunnels and ramps that would be subject to realistic damage from attacks, not just pre-programed scripts and damage levels.
-Sports games where a stiff arm can send a defender down, gut a well placed hit can send a guy spinning rather than sticking magically to the defenders arm and going down in a heap.



Me? Things I would like to see? Environmental interaction. I would liek to be able to blow up buildings by taking out the supports and not having to hammer it until its HP's are gone.

I want boxes i can smash open without magically making it all splinter and dissapear, but also possibly damaging what is inside.

I want life.


Now can this new chip do it? Probably not. Finite elements and time sequence loading takes a while to do even on simple models, to do it 60 times a second would be difficult to say the least. But if they are starting,I am all for it.

 

[IVAces]

If you wanted to get REALLY serious about graphics rendering this could be the way to go. I mean NO price is too much to get every last FPS out of your box; Right?

Its a PCI 16X Cell super computer that gives you an extra 180GFlops of CPU.

http://www.mc.com/cell/pr/news_details.cfm?press_id=2006_07_31_0900_105320_121687pr.cfm

I'd like to see an article on this that shows how it works on current games and applications.

 

[gomerpile]

Ageia needs credit for what they have done, anyone can say after, I could have did that but they did'nt. ATI and Nvidia have hired their own to figure out the physics of Ageia. Good work Ageia for giving the world something to think about.

 

[TheTallGuy]

If you wanted to get REALLY serious about graphics rendering this could be the way to go. I mean NO price is too much to get every last FPS out of your box; Right?

Its a PCI 16X Cell super computer that gives you an extra 180GFlops of CPU.

http://www.mc.com/cell/pr/news_details.cfm?press_id=2006_07_31_0900_105320_121687pr.cfm

I'd like to see an article on this that shows how it works on current games and applications.

I was an intern at Mercury for two summers, that card is legit but a b**** to cool. We had a blower on it when I was there and that thing was loud. Not sure what they ended up doing.

 

[ninjahedge]

Ageia needs credit for what they have done, anyone can say after, I could have did that but they did'nt. ATI and Nvidia have hired their own to figure out the physics of Ageia. Good work Ageia for giving the world something to think about.

It is just like the old Math Co-Processors that came out with the 486 chipset. The only difference being, the limited scope of application making it illogical and impractical to include on the CPU.

I would like to see more like it, and it is good that they are doing this, but I woudl not go so far as to call it innovative...