ATi Radeon can use PhysX!

[Cyber-Mav]

you think that destructable environments dont have anything to do with physics. then your a bigger fool than i thought you were. if there was no physics interaction in a game then nothing exciting would happen. your knowledge of physics seems to be limited to nothing but bullcrap. no phyics in a game means no laws of physics being applied whatsoever. so you loose stuff like mavity to begin with.
also how are you going to blow up a building without out physics? how you going to throw a grenade? where will it go?

in games like even battlefield 2 you dont have destructable environments but physics is still being applied to everything. different vehicles have weight, rockets have propulsion force. impact of explosions.
i can throw a grenade in bf2 to bouce off a wall to get closer to the target im aiming for simply because i understand the laws of physics, the weight of the grenade, the distance to target and the type of surface it will hit and the amound of rebound force that will be generated.

without physics that grenade i throw will be stuck in mid air.

i can remember destructable environments in games all the way back to red faction 2. yes it had destructable environments but it was limited in what it could do but again guess what, its using its own physics engine to understand that a chunk blown out of a wall shoudl land on the floor and no in the sky.

you say that "How walls blow up has NOTHING TO DO WITH THE PHYSICS ENGINE" you just could not be any more wrong.

how you gonna calculate angles of impact and amount of breakage and debris and the direction of the debris without some sort of basic physics engine that takes into account the size and weight of the debris and the acceleration of movement and mavity into account.

and then there is phyics in half-life 2 which without a physics engine that mavity gun is about as much use as your posts are on this forum.

and you ask why everything cant be broken, its simple, not enough computation power is available yet for that sort of level of physics interaction. the cpu won;t be able to handle it, hence why hardware solutions are welcomed to try.

 

[Strife212]

Hardware Physics is a good idea;

PhysX seems dire though. I'm running it on a GTX 280 and it lags and also looks so poor its unreal. (On Unreal Tournament 3 this is)

 

[Lightnix]

Long story short, 3D meshes are still mathematical entities, they can be broken up with algorithms which is in a very simple sense what a physics engine does when breaking something is the appropriate course of action.

 

[james.miller]

That would NEVER work and will never happen. its ridiculous to even suggest that, really, i mean one of the most ridiculous things ever. SLI already has a massive problem, as does crossfire, with load balancing, you think it will just add in another load balancing issue. Also Gpu's basically don't have unused cycles, they might have unused shaders, and unused this and that, but it works as units. YOu can't just sub in a few commands here and there for another application. Gpu's basically use all power available, there are basically no spare cycles, you could not run ppu stuff in spare cycles and you WOULD have to turn off sli and use only one graphics cards for, graphics and the other for PPU, as I said, completely correctly, this would only ever decrease performance.

that would mean a gpu is going full tilt all the time in 3d mode, with power draw peaking at its highest constantly. this quite obviously isnt the case.

 

[BigglesPiP]

Nice, so when I get a new graphic card, I can keep the X1900XT for physics duties.

 

[Monkeynut]

you could not run ppu stuff in spare cycles and you WOULD have to turn off sli and use only one graphics cards for, graphics and the other for PPU, as I said, completely correctly, this would only ever decrease performance.

No, you wouldn't have to turn off SLi. SLi has NOTHING to do with CUDA, which is what GPU PhysX runs on.

"Only ever decrease performance?" The game could be CPU limited. in that case having the GPU loaded with both graphics and PPU work loosens the bottleneck on the CPU.

You can't just sub in a few commands here and there for another application.

Nvidia do all the load balancing driver-side. You tell the PhysX API what to do as if you were using a PhysX PPU card and make your 3D game, and the drivers balance both on the card. So the card can do both. At the same time.

 

[Pottsey]

I am going quote normally so you have no excuse for not responding.

“Also Pottsey, yet again twisting my words, i VERY CLEARLY and obviously to anyone in the universe was talking about the specific levels in the games that support Physx that are for Physx only”

That doesn’t change much. My points still stand and you’re wrong. You’re not limited to 2 maps for the most part, the maps don’t take x50 longer to make, turning PPU mode on doesn’t slow down the FPS. Estimated CPU physics cannot do everything and I listed things it cannot do.

“Gpu's basically use all power available, there are basically no spare cycles, you could not run ppu stuff in spare cycles and you WOULD have to turn off sli and use only one graphics cards for, graphics and the other for PPU, as I said, completely correctly, this would only ever decrease performance.”

Then why has every single benchmark so far showed increased performance? Yes I used the wrong word when I was rushing the post, cycle isnt really the right word. But my point stands the free resource’s on the GPU do physics. The benchmarks agree.
Have you got one bit of evidence saying in SLI you have to use 1 GPU for physics only? It looks like your lying and making stuff up again.
It’s funny how none of the benchmarks show decreased performance yet you say “I said, completely correctly, this would only ever decrease performance”
Show some benchmarks. Instead of making stuff up.

If there is no spare power how do you explain http://en.expreview.com/img/2008/06/20/cpu-gpu.png The FPS doubles, which proves Gpu's dont basically use all power available for gfx and have spare power for physics.

“Everything else i said was accurate too, Pottsey, basically everything you said is bull, twisting everyone elses words and like you clearly did to my post, reading exactly what you want into my words, twisting and misquoting them to make your own point which is, again, incorrect.”

That defence doesn’t work as anyone with basic reading skills can see I didn’t twist your words. Can you show me one bit where I twisted your words? I didn’t read into your post what I wanted, I read what you typed which was wrong. What you said wasn’t accurate it was lies made up by up with no evedince or links backing you up.

Everything you said was accurate, thats a poor joke right? What about “ATI or anyone else to buy a licence still means money in their pocket,” which is a complete wrong as its free for ATI to use and it already works on ATIs cards.

You said ”Every single useful thing you can do with real physics, you can do with estimated cpu friendly physics.” Look at Crysis and how it turns to a slide show with the CPU doing lots of trees falling down at once. So instead they made it so large explosions don’t knock tress down. Look at liquids and cloth which fail to work on the CPU with estimated physics at playable speeds. Can you show me any of these being down decently with estimated physics on the CPU?

How is that twisting your words? You said something stupid, I asked you to prove it and I gave examples where estimated CPU physics don’t work. Yet your response is I am twisting your words. Talk about childish.
If you are right why not show my examples working as estimated physics on the CPU?

“Levels are fully destructable as games would take 5 years rather than 2-3 to make. This is why we've only ever seen 1-2 "physx/highly destructable" levels per physx supported title, there is no time to make more.”
Like I said before there isn’t a single highly destructible level physic games with only 2 maps that I can think off. That’s true anyone can see that them self. I am not twisting your words. People have made their own maps up without taking x50 longer. http://www.youtube.com/watch?v=o8_vdUPGmAQ is a full game that didn’t takes years longer to make due to destructible maps.

“Things like Call of Duty have been done in things like Warmonger.”

I don’t know Call of Duty, show me some vids please. Does it do it on the scale of Warmonger as it the scale that matters. The main benefits of hardware physics are increasing the amount of physics going off at once.

“You think a more realistic physics engine simply means any wall you design into the game will explode realistically?”

The only person harping on about a more realistic physics engine and more realistical physics is you. The main benefit of hardware Physics is the scale. Instead of having on the CPU 10 trees falling down from an explosion, you have 100+ on the PPU/GPU. Instead of the CPU having 20 boulders falling down an avalanche you have 1,000 on the PPU/GPU. The boulders don’t act more realistic it’s the scale. An avalanche with 20 or less boulders just looks poor. Games dont need 100% perfect realistic physics engine's. What games need are more physics on a larger scale. Well apart from liquids and cloth they could do with being more realistical but they dont need to be 100% life like.

“No, you would be completely wrong, if you HAVE to have realistic physics engines to have destructable terraine, explain how games for decades have had "some" destructable items without a completely realistic physics engine.”

You don’t have to have a realistic physics engines to have destructible terrain. No one ever said that. As you dont need a realistic physics engines to have destructible terrain I dont need to explain "games for decades have had "some" destructable items"

“Please tell me how these destructable items get into a game, also please tell me why absolutely everything in all the proper Physx games proper levels, doesn't explode, only parts do... if its done by the engine, and nothing needs designing, and its ALL the engine, why can't everything be broken.”

Since you’re not going believe a word I say why don’t you do what other people have done. Download the free Physx SDK and look for yourself. You can design you own maps and try out the physics. Download Cellfactor and try the map editor.

Everything doesn’t explode in all proper retail Physx games as that can break gameplay and isn’t fun. Most games put people on rails and don’t work if the player can go anywhere. Most good games have a player on rails but doesn’t feel like it as they have limited free movement. Most FPS shooters don’t work if you can wonder all over and just blow holes in everything.

 

[Pottsey]

“PhysX seems dire though. I'm running it on a GTX 280 and it lags and also looks so poor its unreal. (On Unreal Tournament 3 this is)”
Did you to copy the two updated DLL files and put them in the UT folder? That fix’s most people’s slow down problem.

The way it looks has nothing to do with PhysX. Do you mean the mean the way the physics act or the way the game looks?




drunkenmaster
“point out why absolutely every last thing you posted was completely incorrect......”
That’s impossible and you know it as your wrong on so many parts. Can you answer my estimated CPU question? There is no way I am completely incorrect down to every last thing.

 

[Rroff]

Physics processing only uses one part of the GPU rendering pipeline - even with fairly complex graphics there is usually idle time and unused capacity that can be used effectively for physics calculations without having ANY performance impact on the rendering process- yes if you have complex graphics and complex physics there is going to be some slowdown on both sides... but pretty much all the games out now and the immediate future only make use of a few simple physics objects and a handful of particles - which shouldn't hit the rendering performance at all and infact can increase it due to the pressure taken off the CPU.

Hes correct in a way about exploding walls tho - while a proper physics implimentation would allow the game to build the entire world geometry so that it could be deformed (a bit like fluids) we don't currently have the hardware to facilitate that - so the traditional method would to be to build a unique instance of a wall and its exploding counterpart made up of lots of smaller parts - you could then attach physics to those parts to make it look a bit more realistic when the wall breaks up.

If you look in the source files for the COD4 SDK you will see they have templates for several geometry instances where the solid sections have been sub-divided like mad into smaller chunks with physics attached to those chunks so if you shoot this specific geometry in the game chunks of it will fly off or crumble until it all breaks up.

 

[snowdog]

Here's Wikipedia's list of Physx-supporting games.

http://en.wikipedia.org/wiki/Physx#Title_support

Hahaha, Christ. I count three games that are good for a few days, and one actual good game.*

*This is just this poster's opinion, before someone gets butthurt.

Seconded...