Ageia's PhysX Needs a Killer App @ [H] Enthusiast

ChingChang said:
I do have two instances of F@H running, but if more CPU cycles were required by the program F@H should drop down to 0%..
Click to expand...

In my experience with the 4 dual rigs in my home that all run F@H, that is not the case. The prioritization works well on a single core but with dual core / dual processor both instances will not go to zero when something else that is demanding is running.
 
GotNoRice said:
In my experience with the 4 dual rigs in my home that all run F@H, that is not the case. The prioritization works well on a single core but with dual core / dual processor both instances will not go to zero when something else that is demanding is running.
Click to expand...
When I am encoding video and doing other things at the same time, F@H drops down to 0%.
 
ChingChang said:
When I am encoding video and doing other things at the same time, F@H drops down to 0%.
Click to expand...

Yeah, but when you run the PhysX boxes demo it is only one process not many, the scheduling seems to work differently.
 
people are saying its not gonna hit mainstream and thats probably true - never will hit "mainstream". If they advertise it as a entusiast item (which they are) then im sure alot of entusiasts like ourselves will buy it.

After seeing some of those demos I want one, HOWEVER $250 is too much price needs to come down alot. No ones gonna throw down $250 for some extra physics effects or to play some generic shoot em' up with cool physics in it. They need at least 2-3 GOOD games that utilize the PPU for me to go out and buy one.

I like the idea of physics - it opens the door for a much more realistic environment its would be cool to get rid of all those "immovable items" in games. But are we ready for it? - not for $250
 
the one thing i'm hoping comes of all this physx crap is a game with completely destructable terrain. i want a game in which EVERYTHING can be smashed and exploded. combine that with a stunning graphical virtualization of the real world, spawn me with a rocket launcher and unlimited ammo, let me take down walls, buildings, and make craters. optionally combine in a MMO or single player environment, and and that my friends is the killer app.
 
Correct me if this has been mentioned in this thread before, i just didnt see it:

Since most games arent optimized for dual core yet, why not use the second core to run the physics instead buying an additional physX card??
 
IceyN1pPles said:
Correct me if this has been mentioned in this thread before, i just didnt see it:

Since most games arent optimized for dual core yet, why not use the second core to run the physics instead buying an additional physX card??
Click to expand...

It appears that they will do exactly that!

I think that people are thinking of Ageia too much in terms of what they are used to seeing from companies like Nvidia or Ati. Ageia doesn’t just have the PPU cards, they have the PhysX engine. Making the PhysX engine work well on modern CPU’s has always been something they have tried to do, in fact, the PhysX engine predates the PPU by a number of years, during which the CPU was the only option available.

However even a 2nd core is not capable of nearly the amount of physics the PPU is capable of. In addition, while Dual Core optimized games are rarity today, so are games that support the PPU ;) This is clearly a tech designed for future games coming out this year and beyond, and many of those games (such as ut2007 and any other unreal 3 based game) will support multiple cores.
 
pureevilmatt said:
the one thing i'm hoping comes of all this physx crap is a game with completely destructable terrain. i want a game in which EVERYTHING can be smashed and exploded. combine that with a stunning graphical virtualization of the real world, spawn me with a rocket launcher and unlimited ammo, let me take down walls, buildings, and make craters. optionally combine in a MMO or single player environment, and and that my friends is the killer app.
Click to expand...
I wouldn't count on too many games like that. If you can destroy everything, you can also alter how the gameplay progresses quite drastically. It might sound cool to be able to blow up everything... but what happens in a multiplayer situation when everything is destroyed and you have nowhere to hide from snipers, tanks, etc. and you're killed as soon as you spawn? The coolness factor of the game would quickly turn into frustration.
 
dotK said:
I wouldn't count on too many games like that. If you can destroy everything, you can also alter how the gameplay progresses quite drastically. It might sound cool to be able to blow up everything... but what happens in a multiplayer situation when everything is destroyed and you have nowhere to hide from snipers, tanks, etc. and you're killed as soon as you spawn? The coolness factor of the game would quickly turn into frustration.
Click to expand...

What happens? I think some people who played Red Faction should comment.
 
any chance someone can turn this into a bench mark for stuff with or without the pyhsx cards....wondwer how small you make the boxes (more calculations with more boxes) before the physx card starts to get saturated, then compare that to a regular CPU...

I tried it on my Core Duo Laptop, ran pretty smooth and I tried to throw the ball so that it wrecked all the blocks at once...it did make my fan come on and peg my both my CPUs at 100% if only momentarly....kinda neat....
 
Razor6 said:
What happens? I think some people who played Red Faction should comment.
Click to expand...
Yea, the office map with all of the upper floors blown out. It was kind of fun, but the game play really needed a lot more work, hence why I quit playing shortly after I got it. Also the destruction was limited server side, which could probably be pushed pretty high up now with so much memory in the average gamers machine. Their was also some serious issues with cheating from what I remember.
 
I wonder if Nv & Ati will steal this market away from Aegia should the physics cards take off?
 
Razor6 said:
What happens? I think some people who played Red Faction should comment.
Click to expand...

Actually, altho it's really possible, not everything in red faction was destructible. If the level designer didn't want you blowing a hole, there's usually an excuse to stop you without really making it look out of place. You blow up a wall, and it looks like it worked, but when the smoke clears, it's actually a concrete covered steel wall, or metal gratings that you can see through but not penetrate :eek:

In the event that you made a custom level where everything's destructable, then all the cover gets blown. You can always make foxholes :D Destructable terrain includes tunneling remember?

If they put this in a persistent world like Huxley tho, they're gonna have to make the level reset somehow, it'll suck if the tunnel you made disappears while you're inside it.


I've said it before and i'll say it again. Destructible terrains are the GPU and CPU problem. Not the PPU. PPU only does motion and stress vectors (may include polygon vertices to simulate water) but actual polygon modification, which is essentially what terrain deformation is, is all done on the CPU which then updates the GPU. The most non-GPU computational work you'll get is if a squadron was carpet bombing the whole area. But once everything settles, there'll be no more computations, but all the new polygons will be carried by the GPU. Break a 12 triangle shape (a cube) in half (which does not even need a PPU) and you'll get 24 triangles. Shatter it into a dozen pieces (which again, is not done on the PPU) gives you 144 triangles. Extruding a crater shape on a formerly flat surface can instantly turn a 2 triangle floor into a 50 triangle one. The PPU can dictate where each piece goes but that is particle animation NOT TERRAIN DEFORMATION.
 
I agree with the editorial. They need a killer application or almost literally a great game to get it off its feet. Basically like what the original quake did for 3dfx's voodoo monster 3d addon cards.... those were the days :D

edit: Just saw celfactor.... Jedi Knight anyone? :)
 
HOLY CRAP!!! i just watched the cell factor movie!! God LORD I THINK I JUST NUT MY SELF. Im sold. I mean hell, i upgraded froma dual p3 850 system to a dual core x2 system JUST so i could play Doom3. I dont think $250 is too much to ask for that kind of game play.
 
Once again I have to chime in with the ppu being a processor not code. Sure having destructable stuff is nice, but having algorithms that accurately simulate stuff like wood fracturing or concrete crumbling are a larger problem than having a processor to calculate them. That house is assembled, it is not breaking apart in a dynamic way.

Edit: Yes that cell factor vid is nice ;o)
 
They don't need a killer app so much as they need a game that can demonstrate that somebody out there can code two versions of a game - one version that is still fun for people without PPUs, and one version that effectively is the exact same except which uses the PPU, albeit with actual immersion benefits and not in the "gee-whiz, isn't that nifty" kind of way. Which seems downright impossible, to me.
 
mashie said:
You still think in old game physics which is nothing but a great simplification of some very complex calculations.
Click to expand...

Then tell them to stop referring to Red Faction. I'm actually sorry i brought it up in the first place when the PPU was first introduced.

Here is just a sample of what difference proper physics can do. It not only calculate the trajectory of debris, it also calculates how things are broken apart. Unfortunately that scene is most likely a few years off for realtime calculations and rendering though.

It makes perfect sense to use the PPU for destructable terrain. It will calculate how much of the ground that the force will break off based on the material properties and the force generated by the explosives. Not just some approximations as in Red Faction where the debris turned into magic smoke. With proper calculations you will get a raised edge from all the debris falling down as well as secondary damage on any other objects that happen to be too close.
Click to expand...

A PPU does not have any ability whatsoever to directly modify a mesh.
 
mashie said:
You still think in old game physics which is nothing but a great simplification of some very complex calculations. Here is just a sample of what difference proper physics can do.

It makes perfect sense to use the PPU for destructable terrain. It will calculate how much of the ground that the force will break off based on the material properties and the force generated by the explosives. Not just some approximations as in Red Faction where the debris turned into magic smoke. With proper calculations you will get a raised edge from all the debris falling down as well as secondary damage on any other objects that happen to be too close.
Click to expand...
They need to stop referring to things as being realistic. The "realistic" physics in that video are, by far, completely unrealistic. If they want to increase the immersion factor a good place to start would be by making scenes actually realisitc, not more of this "let's see how much crap we can have floating around in the air" thing that seems to be the trend.

If you can completely destory everything, there will be a point where the GPU will absolutely choke. When companies are trying to push every last bit of detail out of a GPU if you encounter incredibly complex scenes you're going to get absolutely awful performance. If GPUs could handle such things I'm sure there would be less of a reliance on tricks such as bump mapping in games today to increase game detail, but that's impossible at the moment and such a game would also be impossible. We are still very limited to what can be done in games due to the lack of graphical horsepower to render very complex scences at acceptable framerates.
 
Maybe UT2007 will let the PPU stand on it's own, but my mind keeps hoping for aomething better something along the lines of: Aliens VS Predator 3!

If AVP3 required a PPU, ageia would be set.
But I would like to still see what it could do for distributed computing.
 
menlatin said:
HOLY CRAP!!! i just watched the cell factor movie!! God LORD I THINK I JUST NUT MY SELF. Im sold. I mean hell, i upgraded froma dual p3 850 system to a dual core x2 system JUST so i could play Doom3. I dont think $250 is too much to ask for that kind of game play.
Click to expand...

But the fact is that you will not be getting that sort of gameplay online. It requires to much bandwidth to do multiplayer except on a LAN. So we see that awesome gameplay bascially have to be single player or LAN and that is not the kind of gameplay that hugely drives hardware upgrades IMO.
 
A majority of my gaming consists of multiplayer gaming. Yes, I've spent plenty of time playing single player games also (i.e. Oblivion), but at the end of the day I usually keep coming back for multiplayer games. When it comes to purchasing hardware, I am MUCH more likely to upgrade for a multiplayer game than for a single player game. A good frame rate means that much more than seeing "pretty explosions," which may in fact end up hindering your multiplayer gaming performance due to decreased visibility relative to non-PPU users.

A bunch of single player/LAN only "let's see how much junk we can launch in the air" type games will probably not persuade me to spend $250 on a PPU. Also, multiplayer games that use the PPU to just make more irrelevant debris fly through the air won't sell me on it either. I'd rather apply that $250 towards a Conroe and get significantly better multiplayer performance that will also give me better single player performance.

What the PPU will need to do in order for me to want to buy it is add physics that makes games more realistic and fun. If it can calculate how objects should break apart based upon the material it is made of and the force acting upon it, how it should slide based on friction generated, etc. then I will probably be eager to buy one. However, what I've seen so far are primitive tech demos that offer just the basics - it's hardly taking physics to the next level.
 
Amanda said:
I have to disagree. Oblivion is single player, and it has made a lot of people upgrade. Another top 10 "game": Flight Simulator 2004. Single player games have a large impact.

I'm not saying that online games don't though! They do have a huge customer base, and growing.

Online gamers usually turn down the eye candy and whatnot to get a higher FPS and quicker response speed, so I don't see online-only gamers to upgrade.

HOWEVER, if this thing takes hold, it will be integrated into the video card or something else. Then, it'd be a part of the system and wouldn't have to be purchaced separately (sp). It would just be an in-game option to turn off, or be disabled on slower connections. Or they may have a cool programming technique to compensate for it...

If a single player game (such as Crysis) comes out supporting that, I'm sure many people would pick it up. It will look good, play good, sound good. And to get the best experience, some people will go all the way. I know I will. I'm building my next machine for FSX and Crysis. DX10 video card, PhysX, best CPU, sound card, motherboard, 4GB RAM, Vista.
Click to expand...


It's not a story based, or sim based game. Would you have expected to UT 2004 to be a success if it was SP only? Bot match MP fps's are fun for about three hours offline. Also, the game itself looked dumb. It was great as a tech demo, but players would spend all their time in the air playing with their ridiculously over powered control of the environment.
 
Amanda said:
I have to disagree. Oblivion is single player, and it has made a lot of people upgrade. Another top 10 "game": Flight Simulator 2004. Single player games have a large impact.

I'm not saying that online games don't though! They do have a huge customer base, and growing.
Click to expand...
The developers of CellFactor have said themselves the game is LAN-only because of bandwidth issues caused by the extensive use of physics in the game. I agree, though. Single player games will spur people to buy new hardware. See Doom 3 and HL2.

There is nothing out there suggesting Crysis will support PhysX.
 
Obi_Kwiet said:
It's not a story based, or sim based game. Would you have expected to UT 2004 to be a success if it was SP only? Bot match MP fps's are fun for about three hours offline. Also, the game itself looked dumb. It was great as a tech demo, but players would spend all their time in the air playing with their ridiculously over powered control of the environment.
Click to expand...

Keep in mind that the level in the videos that have been released was created specifically to show off the PPU.
 
After the long read through this thread I have a couple of things I would like to touch on.

First off, let me get it out there that I hope the hardware fails miserably, including ATI and nVidia's hardware solutions, with almost no support from games. Now let me explain why.

One thing that hasnt been direcly addressed yet is programming complexity. The reason PC games arent a lot better looking is because the games need to satisfy such a wide range of hardware, and throwing a PPU into the mix just doubles that range of hardware. I think with SMP support being worked into games, programmers are going to have a difficult enough time satisfying all the different types of hardware out there.

I think Microsoft has taken a step in the right direction (if the rumors are true) by making stricter requirements for WGF 2.0. Basically, complicating things for the programmers is NOT what we want. Even if the PPU does get supported by gaming companies, by the time they can use it to its potential in games, CPUs will be out that are almost as powerful.

Basically, I think the fact that they have sold their API to PS3 says that multicore CPUs will be capable of very impressive physics. And as for nVidia and ATI solutions, them adding support for physics calculations will make them into more of a general purpose processor, which completely defies the point of having a dedicated GPU in the first place.

So I say let the programmers deal with multicore, using the Havok or PhsyX engine, or whatever engine they want so that we can save our money for a decent CPU, and see how things go from there.
 
stelleg151 said:
After the long read through this thread I have a couple of things I would like to touch on.

First off, let me get it out there that I hope the hardware fails miserably, including ATI and nVidia's hardware solutions, with almost no support from games. Now let me explain why.

One thing that hasnt been direcly addressed yet is programming complexity. The reason PC games arent a lot better looking is because the games need to satisfy such a wide range of hardware, and throwing a PPU into the mix just doubles that range of hardware. I think with SMP support being worked into games, programmers are going to have a difficult enough time satisfying all the different types of hardware out there.

I think Microsoft has taken a step in the right direction (if the rumors are true) by making stricter requirements for WGF 2.0. Basically, complicating things for the programmers is NOT what we want. Even if the PPU does get supported by gaming companies, by the time they can use it to its potential in games, CPUs will be out that are almost as powerful.

Basically, I think the fact that they have sold their API to PS3 says that multicore CPUs will be capable of very impressive physics. And as for nVidia and ATI solutions, them adding support for physics calculations will make them into more of a general purpose processor, which completely defies the point of having a dedicated GPU in the first place.

So I say let the programmers deal with multicore, using the Havok or PhsyX engine, or whatever engine they want so that we can save our money for a decent CPU, and see how things go from there.
Click to expand...

Yeah, and I say let the programmers deal with this software rendering mode, not using OpenGL or Directx that way we can save our money on GPUs too! ;)
 
Using a physics api makes things easier for the programmer as they dont have to write their own collision detection system.
 
Lord of Shadows said:
Using a physics api makes things easier for the programmer as they dont have to write their own collision detection system.
Click to expand...


and much more ;)

Either and SDK or API is great, thats why I think Ageia will do well if games that use thier cards come out soon. their sdk is free to use.
 
mashie said:
Well if they implement the PhysX engine they will suffer the nasty side effect of getting both dual core and PPU support automatically, oops. Guess they better use their own code to make sure such a horrible thing won't happen... :rolleyes:
Click to expand...

Their api is not PPU specific, thus them selling it to sony.

People in the PC industry seem to think that throwing enough hardware power at a problem is the best solution, but that is what is killing the PC gaming industry. The vast majority of people out there look at a PC and see how much people are spending on video cards alone and realize they could get a brand new console for that much, which as a result encourages more companies to make games for consoles instead of PCs.

I want PC games to stick around, and I think adding another 250 dollars to build a gaming computer is going to hurt the industry more than it will help it.
 
kraken0698 said:
Yeah, and I say let the programmers deal with this software rendering mode, not using OpenGL or Directx that way we can save our money on GPUs too! ;)
Click to expand...

Yeah the graphics analogy has been rehashed, but one of the [H] folks put it best when he said that the upgrade to dedicated GFX would make any non-gamer go "wow, thats way better", whereas this upgrade would take "explaining" as he put it. Basically, the upgrade in quality isnt significant enough to justify an entire revolution in computing IMO.
 
razor1 said:
and much more ;)

Either and SDK or API is great, thats why I think Ageia will do well if games that use thier cards come out soon. their sdk is free to use.
Click to expand...

I agree that these tools are great, but I would prefer that the game companies pay a "relatively" small price and buy the API and implement it for use with multicore CPU's like PS3 is doing, than force us to pay 250 dollars to see almost the same thing.

EDIT: Sorry those responses should have been put into one. :rolleyes:
 
Cell is not your average multicore CPU. In fact, its a lot closer to a PPU in design than a typical multicore processor. That's part of the reason why Ageia is working so closely with Sony.
 
Low Roller said:
Cell is not your average multicore CPU. In fact, its a lot closer to a PPU in design than a typical multicore processor. That's part of the reason why Ageia is working so closely with Sony.
Click to expand...

I was under the impression that Cell was for the most part similar to IBM's other PowerPC cores, which are not that dissimilar to x86. I havent payed that close attention, but I believe cell is still a general purpose processor, so I assume it is more similar to an x86 multicore processor than the PPU.

mashie said:
Exactly, games using the API will work on single core, dual core and PPU based systems. It is similar to the fact that you don't _need_ SLI/crossfire to play a game but it will improve it nicely. The same with the PPU in a PhysX based game, it will just make it even smoother and enable the game developers to push the envelop even further for those of us that want the best possible experience. If you want to see a building crumble at 10 fps on a dual core fine, I rather see it crumble at 60 fps on a PPU based system.

Another comparision is some people rather use AC97 codecs for the sound while other prefer a dedicated soundcard to squeeze out the last extra fps from the CPU. Anything that free up the CPU is welcome in my book.
Click to expand...

I definately see why this is attractive, but my point is this:

mashie said:
games using the API will work on single core, dual core and PPU based systems.
Click to expand...

I think programming physics so that it satsifies the small enthusiast crowd that owns a PPU, the quickly growing multicore crowd, and the single core crowd will be a lot more difficult than only programming physics for x86, and will thus overall seriously hurt the vast majority of gamers who dont want to pay the extra money.

Unloading work off the CPU is good, but where do you draw the line? If unloading physics succeeds, will we get another card for AI, then for lighting, then for whatever else is eating up CPU cycles?

I just think given a simpler task of coding the API only for x86, that devs will actually give us comparable physics, as they wont be having to take into account 2 completely different hardware setups. As people have shown , the demo can be run on dual Opties, and even HL2 has fairly impressive physics that run on single core CPU's.
 
stelleg151 said:
I just think given a simpler task of coding the API only for x86, that devs will actually give us comparable physics, as they wont be having to take into account 2 completely different hardware setups. As people have shown , the demo can be run on dual Opties, and even HL2 has fairly impressive physics that run on single core CPU's.
Click to expand...

That's exactly what mashie's trying to explain, though. With PhysX, all the programmer has to do is program to the API. The API, then, delivers that to the platform, whether it be single core, dual core, or PPU. Their implementation actually *simplifies* the programmer's work.

 
mashie said:
The Cell used in the PS3 has one general purpose core and 7 or 8 very specialized SPEs, more info here.



Well the people running fx5200 videocards can't run 8xAA and 16xAF either so what prevent the developers to give you low, medium and high physics settings?
Click to expand...

PPU guy: My grenade bounces better than yours! Nye nye! :p

Software guy: My bullets don't have drop! Nye Nye! :p

Regardless of video settings, gameplay remains the same. You can't say the same for physics, especially in an application that's supposed to make a $300 card worth it.
 
Sly said:
PPU guy: My grenade bounces better than yours! Nye nye! :p

Software guy: My bullets don't have drop! Nye Nye! :p

Regardless of video settings, gameplay remains the same. You can't say the same for physics, especially in an application that's supposed to make a $300 card worth it.
Click to expand...

But that's still only considering multiplay. The reason behind the Cellfactor trailer (i.e. lan multiplay with a bunch of people "battling" eachother with junk) is that that truly is the easiest way to demonstrate that kind of physics interaction - no story needed, no special gameplay elements or setup, just stick 4 guys and a pile of crap in a room and let them go at it.

I firmly believe that the "killer app" whatever it is, is going to be first and foremost, a single player experience. As an example, even as simple as it was, who didn't think something along the lines of "Hey, now That's cool!" in HL2 when getting to the part where you have to stack up the bricks on that teeter-totter, or pull the plastic barrels down under that cage to float the ramp for the airskip.

Multiplay will be there, but I believe that it will not come into it's own right at first.

 
mashie said:
The Cell used in the PS3 has one general purpose core and 7 or 8 very specialized SPEs, more info here.



Well the people running fx5200 videocards can't run 8xAA and 16xAF either so what prevent the developers to give you low, medium and high physics settings?
Click to expand...

Cool thanks, I had forgotten about that minor:D detail about how the cell works :)

But my point regarding the PPU is just that, the reason PC games dont look as good as they could is because they have to program them so they can run on such a wide variance of video cards, and adding another card that they have to do that with will only make the situation worse.

uzor said:
That's exactly what mashie's trying to explain, though. With PhysX, all the programmer has to do is program to the API. The API, then, delivers that to the platform, whether it be single core, dual core, or PPU. Their implementation actually *simplifies* the programmer's work.
Click to expand...

I admit I dont know exactly how the API works, but I would be very impressed if the API could do a half decent job of accuratly adjusting the amount of physics according to each clients unique hardware setup.
 
stelleg151 said:
Yeah the graphics analogy has been rehashed, but one of the [H] folks put it best when he said that the upgrade to dedicated GFX would make any non-gamer go "wow, thats way better", whereas this upgrade would take "explaining" as he put it. Basically, the upgrade in quality isnt significant enough to justify an entire revolution in computing IMO.
Click to expand...

That's just the thing. With physics, there is that wow factor. If I were to look at someone playing a game that has 20 boulders crashing down a hill and then at someone else who has 10,000+ boulders crashing down the same hill, I would say wow!

The boulders are just one scenario. I mean the possibilites are endless. With fully destructible buildings, vehicles, foliage even terrain. Not to mention liquids, like water, lava etc.

But hey, that's just me. :)
 
You must log in or register to reply here.
Back
Top