Ageia responds to GPU physics

the holes you blow into the walls on red faction are all pre-modelled and stuff, no physics at all sorry....
 
He didn't say:

"*blows hole in the wall that aren't premodeled and I watch all the dust float around as the metal schrapnel fall to the ground and bounce a little bit, then some dust gets in my eye, I rub it, and water in my eye falls soso realistically to the floor and just misses the schrapnel then I see XSNiper turn back to face me, with the wind blowing through his hair ever so slightly, while the dust gets turned up when he walks over it*"

Or something, now did he?
He was vague, I gave him a smartass response.
Deal with it.
What he did, is possible on Red Faction.
I've done it plenty of times myself.
 
Im wondering how sustainable this will be? Right now there is A LOT of room for improvement in graphics to get to photorealism. But say you're "rendering" gravity on a PPU,
the accelaration rate is always going to be 9.8 meters per second at 1G. even in 50 years when computers are doing things we cant even comprehend right now, accelaration will always be 9.8 meters per second at 1G.

How much room for improvement is there really going to be? is this a one trick pony?
 
AFAIK BFG is making the PhysX cards for Ageia. It would be really nice if bfg manufactured cards that have both a GPU and PPU on the PCB. Obviously such a card would be more expensive than the same video card minus the ppu, however it should still be significantly cheaper than buying two separate cards, not to mention it would leave an extra slot open in your motherboard for something else. Its not too likely that putting both chips on one card would result in the card being choked for bandwidth - in general the bus a video card is on (things such as NV3 cards on PCI aside) has much more bandwidth than the card is capable of using, and when the top end cards start approaching about half that bus's bandwidth, the bus gets replaced with something that has even more bandwidth (think agp to pci-e).

Also, to those saying that PPUs are unnecessary because with the rise of multicore cpus, physics could just be offloaded onto a core. Look at HL2 - it has one of the best physics engines out of current games, but the physics in it are still very basic, and they devour cpu cycles - it takes something like 90% of the cpu time when I play CSS, although granted, my cpu isnt exactly very fast. Even a core on a cpu two or three times as fast couldnt handle the load of remotely complex physics. a PPU on the other hand is, iirc, about 400 times as effective as a cpu at processing physics.
 
CodeWaste said:
Im wondering how sustainable this will be? Right now there is A LOT of room for improvement in graphics to get to photorealism. But say you're "rendering" gravity on a PPU,
the accelaration rate is always going to be 9.8 meters per second at 1G. even in 50 years when computers are doing things we cant even comprehend right now, accelaration will always be 9.8 meters per second at 1G.

How much room for improvement is there really going to be? is this a one trick pony?
That's what you think, in reality gravity decreases exponentially as the distance from the earth increases. Although we don't need PPU for that, there are alot more calculations to do in many games.
 
Loctus said:
That's what you think, in reality gravity decreases exponentially as the distance from the earth increases. Although we don't need PPU for that, there are alot more calculations to do in many games.


Yes but i said at 1G. if you're moving away from the earth then you are leaving a 1G envrionment.
 
CodeWaste said:
Yes but i said at 1G. if you're moving away from the earth then you are leaving a 1G envrionment.

It's not really at 1g, 9.8m/s/s is 1g.
 
Dr. X said:
It's not really at 1g, 9.8m/s/s is 1g.

No. When you are standing on the surface of the earth, you are experiencing gravity at 1g. And 1g is 9.8 m/s^2 not 9.8m/s. Gravity is not a function of velocity, but acceleration. And don't bring semantics into physics. OT.

edit: I see you changed it to /s/s.
 
What I'm trying to say is that physics processing will always be a drop in the bucket compared to general graphics processing. By the time we have super realistic physics, we'll have super realistic graphics and the power needed to handle the physics relative to the rest of the game will be similar to how it is now. If you guys had ever done any game programming, it would be simple to see how primitive physics calculations really are. Not only are they far simpler than graphics calculations, but they also don't require the same memory as graphics operations. Consider that textures are all loaded into video memory - all the physics card would have to do is perform operations on objects in the scene, thus not even close to the same memory. I guess something like 1MB would be overkill for a physics card. Think about it - your TI-83 calculator for school can do all the operations a physics card would have to do, even if you were trying to model fluid dynamics. That's a major reason for having video cards in the first place - video memory. A physics card wouldn't even need it's own memory beyond a cache, since it would just be a processor. Since it is just a processor, why not just use a core on a multi-core CPU?

I wish there were an easier way to explain this for people who haven't done any 3D game programming, but unfortunately it's hard to put into layman's terms. Suffice it to say a lot of what you guys are talking about is stuff that isn't even physics, and would never be handled by a physics card anyway.

things would react according to realtime calculations of the interaction between objects, which brings a whole new level of immersion and realism.

And again, I will say this already does happen. HL2 was not scripted events. Max Payne 2's primitive physics were real physics, not scripted - go into the funhouse, in the back room where they have the "physics room". Throw something at the stack of cans and watch how they react. That is physics, and it isn't scripted. Dropping something into water and having true 3d ripples is NOT as computationally intensive as people think. You have an algorithm that processes the impact on the water, and adjusts the locations of the polygons that make up the water's surface. If you knew the amount of power it took to draw all those polygons in the first place, you'd see the physics part is nothing.

I still stand by my original statement that physics are within sight of reaching true reality, something graphics has a long way to go to catch up. That physics can be so powerful right now on software tells me that with tomorrow's more powerful CPUs, the idea of a physics card is a joke. We might as well have AI cards. Hell, let's modulize every aspect of computer games and give them their own cards. I'm sure you guys would eat that up.

As for those demos, come on - you trust a demo made by the people who make the card itself?
 
jdLordHelmet, I agree with you in every way. While the only physics I've really worked with is Karma, which is about as primitive as it gets, I have also seen demos of Havoc physics, and it's just ridiculously simple.

My opinion of this entire thing is that it's pointless at this point in time. Remember the Novodex Rocket demo, or the Artificial Studios Mansion video (or even their playable demo material, it you got to see it)?
The advanced physics never really added to the atmosphere. The coolest thing these physics do is make lights swing around, which takes no processing power in comparison to creating realtime soft-shadows on every object as the light swings around.

Graphics and physics has always been about cheating. I don't care if the grass moves according to some laws I don't understand, as long as it looks real.
 
Call me a crazy rebel thinker here, but am I only one asking: Why?.

Think hard before attempting to answer these questions, because I have an answer to each that makes a PPU seems less and less important.

Why is being able to realistically blast through walls a good thing?

Why is grass blowing realistically in the wind a good thing?

Why is being able to model the trajectories of stones in an avalanche a good thing?

And the over-arching "trick" question you should bear in mind,: Why do we need a PPU to do them?
 
Nuzzles said:
He didn't say:

"*blows hole in the wall that aren't premodeled and I watch all the dust float around as the metal schrapnel fall to the ground and bounce a little bit, then some dust gets in my eye, I rub it, and water in my eye falls soso realistically to the floor and just misses the schrapnel then I see XSNiper turn back to face me, with the wind blowing through his hair ever so slightly, while the dust gets turned up when he walks over it*"

Or something, now did he?
He was vague, I gave him a smartass response.
Deal with it.
What he did, is possible on Red Faction.
I've done it plenty of times myself.


Well even blowing up a wall like in Red Faction takes a good amount of CPU. Since this kind of interaction will require an object to be an entity. Now if you take into account most CPU's today (high end ones) if we have over 5000 entities in a single level, the cpu has to rememeber position for each one, this hurts the CPU a great deal (most engines this will bottleneck the cpu to around 20 FPS (keep in mind Red Faction doesn't use nearly this amount of entities in games, but for newer games like games coming out on Unreal 3 engine will.

For non entites this is taken care of by the GPU. But anyways, # of objects plays an important roll in over all speed. In a perfect world it would be great to build an entire level as one unit and have each poly culled. But being a single unit breaking up a level by breaking up vertices for culling would be a very slow process (another draw back to this is it would have to be done by the CPU).

We don't have enough graphics performance as it is, and when lighting algo's take a step towards true spherical harmonics, it is going to require alot more from GPU's.
 
I imagine a world where blowing the hinges off of a door to breach is a reality. You're simply not blowing the hinges off, you're splintering the wood around the door and realisticly warping the steel hinges, perhaps bending the door in with a well placed kick by the knob.

Instead of bullets dealing damage, they have actual ballistic characteristics, and your body has the consistancy of flesh and bone, and said bullet actually tears bits away as it hits you, or cracks your plate, if you're wearing armour. Maybe taking a headshot with a helmet will actualy send you ass over tea-kettle, knocking over anything in your path.

MIllions of particles of dust are kicked up behind you as you dash through the streets of some god forsaken country, when you stumble through the next archway, a street dealers cart is pushed in front of you, which you immediately tumble over. The car is tipped over and out spill the 250 apples and oranges and taiwanese trinkets he was going to sell. Lucky for you, you fell just as the archway above you has a dozen 7.62 mm holes punched through it by automatic weapons fire from the third world militants behind you. Light streams through the holes and refracts off of the dust as you scramble for cover.

Yeah, so you can quit trying to stop progress because you're an amateur graphics programmer, if the coders for UE3 think it warrants inclusion, and so does Sony(ps3) and Nintendo(revolution (x360 got to market to quick for PPU inclusion)) then I think it could be a worthy inclusion. It's just like saying "Hey, this cpu runs wordpad, why would I ever need anything faster?"
 
LoneWolf said:
I'd like to see a company like Ageia produce a less expensive product, or PhysX processors that could be added to a graphics card, rather than a separate product.

Are physics in games relevant? Yes, but so is price. Most gamers are already shelling $150-400 for a graphics card;tell a gamer they'll need to shell another $200 for a PPU card, and I'll bet that over two-thirds of them will balk.

If less than a third of gamers owns one of these cards, developers will be less likely to code support for them. They're in it for money like everyone else, and if 60% of people out there can't play a game because they don't have the hardware, that game isn't going to make its development money back. Sure, a separate code-path could be written for computers that have the PhysX card, but what will the additional development costs be to do so?

Even if the Ageia PhysX product is groundbreaking, without games to support it, it's just a pretty piece of hardware. One might point to it as the possible 3dfx Voodoo of today, but back when the Voodoo came out, a basic 2MB/4MB DRAM-based PCI graphics card cost far less than the cards of today, so having both the graphics card and the Voodoo wasn't nearly as hard on one's wallet as this will be. Ageia's product is one that I hope will do some neat things --but I'd like to see it within the realm of affordability.


dont know if any one said this yet but Ageia's physics API has 2 code paths built in to it
for a game to use there hardware it just has to use there middleware if it doesnt find the card it just falls back to software
 
A PPU will definately be needed. Imagine the computational requirements for dropping a two story brick house around some one's ears, not just a pre planned drop but depending on damage to a whole host of structural components and subsequent beams, bricks and bits that form a building. Another example is people and ordnance barrelling through dense folliage, It all adds up real fast I can tell you.

.
 
I think a PPU is a wonderful product and needed although i will say this: they haft to find a way to get it in prebuild systems like dell and whatnot, OR perhaps lisence it to nvidia or ati to put on there videocards or maybie the mb companies. Sure, it might work like 3d cards did and people pick up on it but i have my reservations
 
pakotlar said:
No. When you are standing on the surface of the earth, you are experiencing gravity at 1g. And 1g is 9.8 m/s^2 not 9.8m/s. Gravity is not a function of velocity, but acceleration. And don't bring semantics into physics. OT.

edit: I see you changed it to /s/s.

its not 9.8, it depends where on earth's surface you are.
 
Rasha said:
I think a PPU is a wonderful product and needed although i will say this: they haft to find a way to get it in prebuild systems like dell and whatnot, OR perhaps lisence it to nvidia or ati to put on there videocards or maybie the mb companies. Sure, it might work like 3d cards did and people pick up on it but i have my reservations
I think you are right.
When I worked at Best Buy (probably 2 years ago...9800XT and 5900 Ultra were the top-end video cards at the time), I sold hundreds of GF4MX's, and 5200's, and etc. to pissed off people who took a copy of a game home for their kids and found out that their Intel Extreme graphics weren't so extreme. So now they are coming to me with an attitude saying things like "you want me to spend another $100 to play this stupid game?"
Now, if they take a game home, and find out that not only do they need to upgrade their graphics, but they need a new PPU too, they will be doubly pissed.
 
jdLordHelmet said:
Dropping something into water and having true 3d ripples is NOT as computationally intensive as people think. You have an algorithm that processes the impact on the water, and adjusts the locations of the polygons that make up the water's surface. If you knew the amount of power it took to draw all those polygons in the first place, you'd see the physics part is nothing.

I think you are missing the idea behind it all: i want to have the animated world behave just like the real world. For that, we need physics, i.e. we need to have to model at least the EM force and gravity.

Consider how the cars in NFSU handle and how their handling is completely unrelated to the real world. There are algorithms to approximate the handling of cars on the street, but until we are able to actually constantly calculate the interaction between the air passing over the car, the tires gripping to the road surface and all other forces acting on it, we will not have a real simulation. If you tell me that this is a 'trivial' matter, please explain why we still need wind tunnels to test the aerodymanic properties to Mr. Lance Armstrong riding his TT bike, since we should be able to simply calculate the interaction right now.

Honestly, I feel that part of the GFX community thinks that PhysX stepped on their toes, since traditionally, the graphics programmers had to consider deformation and animation in their aglorithms. I see a PPU as an extreme advantage to the GFX card, since it would be able to process all transforms that one would want on the vertices that you mentioned earlier. Instead of having the graphics subsystem handle the movement and placement of vertices, the PPU would do this, while the graphics card would only worry about displaying the resulting arrangement.

Now you could argue that the amount of processing power required to translate and rotate solid objects is minimal in comparison to rendering the actual image. And I understand that you would suggest that the number of polygons would have to increase dramatically in order to make the scene arrangement a difficult task. However, if the PPU is 'smart' enough to dynamically add polygons when deformation occurs and is able to calculate the results of collisions with greater accuracy than it is currently done (and we can all agree that most games are far from reality when it comes to things hitting each other), then the PPU will offer a signficiant benefit.


 
$100, $200, $400, whatever. I don't think it'll matter when the first company (like Dell) starts dropping PPUs into $500-700 PCs. Average user is taken care of. Enthusiasts don't come into the equation except at the very beginning, and that as unpaid beta testers.
 
drizzt81 said:
Now you could argue that the amount of processing power required to translate and rotate solid objects is minimal in comparison to rendering the actual image. And I understand that you would suggest that the number of polygons would have to increase dramatically in order to make the scene arrangement a difficult task. However, if the PPU is 'smart' enough to dynamically add polygons when deformation occurs and is able to calculate the results of collisions with greater accuracy than it is currently done (and we can all agree that most games are far from reality when it comes to things hitting each other), then the PPU will offer a signficiant benefit.

colissions is a big thing i would like to see improve no more weapons poking threw walls in CS:S and other games would be nice
 
perplex said:
its not 9.8, it depends where on earth's surface you are.

Look i was making a generalzation about accelaration, its really not the point. the point is, physics dosen't change. its not like its gonna be 9.8 m/s^2 today, but next week its gonna be 3.4m/s^2, no its always going to be 9.8.

So my question is, once we have a physics card with all the physics we need for games, whats the point of buying a new one when it comes out? its not like graphics, physics can't be improved with pixel shading, vertex shading, HDR etc. since physics can't change, is there ever going to be an upgrade path? Im asking if this is a one trick pony for consumers.
 
CodeWaste said:
Im asking if this is a one trick pony for consumers.
The only upgrade path I see them being able to head down is the performance one. Of course the physics itself couldn't be improved, but the amount of processes able to be computed could.

In that sense it functions the same as any processor does. The farther the technology goes, the faster it will perform. In the case of a PPU they could quite easily push through WAY more calculations by moving to a smaller nm PPU size, and other card refinements.

Graphics card updates aren't just about new feature sets. They are also about making the core technology function better in itself.
 
jdLordHelmet said:
I say hardware physics will never be needed. The comparison of hardware graphics versus hardware physics is apples to oranges. Graphics have a long, long way to go. We still haven't reached photo-realistic graphics, and it will be a long time before we truly do. Physics, on the other hand, is far simpler to calculate, and much, much less computationally expensive than graphics. There is only so far you can go with physics, and the Havok engine is pretty close to it (at least within sight of it). Our processing power is increasing at a rate far greater than the computational requirements of game physics. Quite simply, the power needed for physics is a joke compared to graphics. Unless you are trying to give complex physics to every drop of rain in a storm, it just isn't going to be needed.



Watch this: http://www.gdhardware.com/files/ageia/BuildingDestruction/side_by_side.wmv

Look at this: http://softimage.wiki.avid.com/index.php/Fun_With_PhysX

And this: http://www.airtightgames.com/currentproject.html

Get this: http://www.novodex.com/rocket/NovodexRocket_V1_1.exe , play with the physics. Make some of those objects bouncy, or sticky, or play with thier friction ratios, set their breaking points, it is not just bricks, the PPU will enable water, and breaking walls etc. etc.

I don't think you understand what all this will allow. (it will allow fluid simulation, and breaking objects)

==>Lazn
 
Personally I'm really excited about a PPU. Finally when I shoot bullets at walls or throw a grenade into a building, I will see some damage to the building.
 
The first video of the houses are so completely biased its funny. First of all if you think a modern cpu can handle such a small amount of objects your kidding yourself, maybe not as much as the ppu house, but at least three to four times more than what they have it doing. Secondly the houses arent dynamically reacting to fracture or anything special like that. They have been chopped up into pieces by hand, and if you think of it like this you see how much of a waste of time it was in terms of real gameplay.

A equally representative video, yet less impressive. Would be composed of the ppu processing falling bricks while maintaining a certain frame rate with a cpu doing the same. The difference in this video is that the cpu can handle much more than they let it, and the bricks are in the shape of houses. (Which dont particularly look all that great to begin with)

Edit's as i watch em:

Now the fluid simulation video, which I havent seen before is very impressive, but would be much more impressive if the particles were smaller, but that is an effect worthy of some cash.

The metal building, which I also havent seen before, is an interesting usage of cloth simulation, although it looks hilariously unrealistic it has promise... maybe...

And finally the plane, seen it before, just more falling bricks, but this time with a more realistic count when compared to the house video.
 
jdLordHelmet said:
What I'm trying to say is that physics processing will always be a drop in the bucket compared to general graphics processing. By the time we have super realistic physics, we'll have super realistic graphics and the power needed to handle the physics relative to the rest of the game will be similar to how it is now. If you guys had ever done any game programming, it would be simple to see how primitive physics calculations really are. Not only are they far simpler than graphics calculations, but they also don't require the same memory as graphics operations. Consider that textures are all loaded into video memory - all the physics card would have to do is perform operations on objects in the scene, thus not even close to the same memory. I guess something like 1MB would be overkill for a physics card. Think about it - your TI-83 calculator for school can do all the operations a physics card would have to do, even if you were trying to model fluid dynamics. That's a major reason for having video cards in the first place - video memory. A physics card wouldn't even need it's own memory beyond a cache, since it would just be a processor. Since it is just a processor, why not just use a core on a multi-core CPU?

I wish there were an easier way to explain this for people who haven't done any 3D game programming, but unfortunately it's hard to put into layman's terms. Suffice it to say a lot of what you guys are talking about is stuff that isn't even physics, and would never be handled by a physics card anyway.



And again, I will say this already does happen. HL2 was not scripted events. Max Payne 2's primitive physics were real physics, not scripted - go into the funhouse, in the back room where they have the "physics room". Throw something at the stack of cans and watch how they react. That is physics, and it isn't scripted. Dropping something into water and having true 3d ripples is NOT as computationally intensive as people think. You have an algorithm that processes the impact on the water, and adjusts the locations of the polygons that make up the water's surface. If you knew the amount of power it took to draw all those polygons in the first place, you'd see the physics part is nothing.

I still stand by my original statement that physics are within sight of reaching true reality, something graphics has a long way to go to catch up. That physics can be so powerful right now on software tells me that with tomorrow's more powerful CPUs, the idea of a physics card is a joke. We might as well have AI cards. Hell, let's modulize every aspect of computer games and give them their own cards. I'm sure you guys would eat that up.

As for those demos, come on - you trust a demo made by the people who make the card itself?

You are missing the whole point of what a PPU is supposed to do. CPUs are general processing units. They can process any code. GPUs are designed with a specialization in graphics processing. The PPU would handle those calculations specifically for physics. They can still utilize DC CPUs, but they can offload AI to the other core.

With that said, look at how many objects within a game actually respond to interaction. The problem with using a general CPU is that they are not designed specifically to do such calculations, and while they can, a processor designed with that function in mind can handle many more of those calculations. To this day there are many, many objects within a game that are static and cannot be altered in any way. Only a minute portion of the objects will respond to interaction, even in a game like HL2 the number of interacting objects was small and it still puts a major stress on the CPU to do that.

Wouldn't it be a step forward if an objective played out according to the laws of physics. You're playing a WWII shooter and you have been given an assignment to blow up a bridge. As it stands now you simply plant the charges within an area designated by the programmer as a valid planting area. The bridge blowing up is scripted and occurs the same way time after time. What if you could plant the charges anywhere and the destruction of the bridge, or failur to destroy the bridge, happened based on the amount and power of the charges used, their planting location, the materials used to build the bridge, as well as other mitigating factors. Such conditions, as well as the rest of the real physics that happen are just too much for a CPU, even a DC one because such calculatioins are not a specialty.

While I'm sure this would in turn create additional work for the VC, the calculations needed for continuous, full world interaction physics are far more demanding than you think.
 
Lord of Shadows said:
The first video of the houses are so completely biased its funny. First of all if you think a modern cpu can handle such a small amount of objects your kidding yourself, maybe not as much as the ppu house, but at least three to four times more than what they have it doing. Secondly the houses arent dynamically reacting to fracture or anything special like that. They have been chopped up into pieces by hand, and if you think of it like this you see how much of a waste of time it was in terms of real gameplay.

A equally representative video, yet less impressive. Would be composed of the ppu processing falling bricks while maintaining a certain frame rate with a cpu doing the same. The difference in this video is that the cpu can handle much more than they let it, and the bricks are in the shape of houses. (Which dont particularly look all that great to begin with)

Edit's as i watch em:

Now the fluid simulation video, which I havent seen before is very impressive, but would be much more impressive if the particles were smaller, but that is an effect worthy of some cash.

The metal building, which I also havent seen before, is an interesting usage of cloth simulation, although it looks hilariously unrealistic it has promise... maybe...

And finally the plane, seen it before, just more falling bricks, but this time with a more realistic count when compared to the house video.


Get the rocket demo, do the "big bang" explosion, it is 16000 bricks falling, your CPU will get ~3-4 FPS at best. The PPU is supposed to handle 30000 of them at over 60 FPS. We shall see if they can actually hold up to that promise, but that is what Ageia has been saying.

==>Lazn
 
CodeWaste said:
Im wondering how sustainable this will be? Right now there is A LOT of room for improvement in graphics to get to photorealism. But say you're "rendering" gravity on a PPU,
the accelaration rate is always going to be 9.8 meters per second at 1G. even in 50 years when computers are doing things we cant even comprehend right now, accelaration will always be 9.8 meters per second at 1G.

How much room for improvement is there really going to be? is this a one trick pony?
What about air friction, terminal velocity, and other obstacles? Yes, you can simply add 9.8m/s every second, but that would be a gross oversimplification of the phisics required, which is just what we want to avoid.
CodeWaste said:
Yes but i said at 1G. if you're moving away from the earth then you are leaving a 1G envrionment.
Well, if you're on the Earth's surface, you aren't falling, now are you?
jdLordHelmet said:
What I'm trying to say is that physics processing will always be a drop in the bucket compared to general graphics processing.
No, the CPU is doing all the real work. Without this math, you don't have a game. A physics processor can affect gameplay, not just graphics.
 
The PPU needn't be a $200 addon card. They can sell a gimped version as a $20 chip on your gaming card. I see the PPU as more of an upgrade to the CPU than the GPU. You can play Doom 3 on your Geforce 7800 or your Radeon 9000 and it's the same game. A physics processor can make it an entirely different game.
 
pakotlar said:
No. When you are standing on the surface of the earth, you are experiencing gravity at 1g. And 1g is 9.8 m/s^2 not 9.8m/s. Gravity is not a function of velocity, but acceleration. And don't bring semantics into physics. OT.

edit: I see you changed it to /s/s.

Actually, I never changed anything. It was m/s/s from the start. Perhaps you didn't read it correctly.
 
I'm not very good at speculating about things (well, not in English language anyways). I'll let AGEIA release their card first, then games and then judge about the whole PhysX issue.

But I say this: it sure is interesting to lurk and read these threads from all kinds of hardware forums :)
 
my opinion only....
we dont need another fork in processing types and solutions and hardware types.
we need more multi core multi threaded applications using *general* programmable processors.

you buy system in the future. no more CPU's and GPU, and PPU's
just a modular motherboard with the DVI ports (video becomes just another port)
and the user can buy 1 or AS MANY AS THEY WANT general processors.
the OS's job will be to dynamically configure the general purpose processors to do a mix of contemprary "CPU, or GPU or PPU" work.

Users who want to upgrade, simply buy another generic programable processor and stick in the mobo. reboot, the OS now has more resources to work with, but the overall allocation process doesn't change. Your game or app or whatever creates threads based upon whats its doing. This is my audio thread. This is my AI thread. This is my graphics thread. This is my physics thread, etc ad nasuem. The OS programs the configurable processors to handle the type of load more efficently, and then assigns the threads to them. Ideally, even the individual threads should be multi threaded, so if you have 8 general processors, the graphics "process" could spawn say 2 threads, both on different general processors. Yes, timing would be a bitch, but it already is with multi core GPU's.

However, in this idealistic model, consumers just pop in some more general purpose processors when their system starts lagging. No more OMG i need a new mobo for the new interface of GPUs or PPUs or whatever.

of course, this will NEVER happen because then too many companies would be out of business.
 
i just thought of something, this has been done before. Anyone remember the optional co-math processor that was availible as an addon to computers like 12-14 years ago? It looked like it would fit in a ram slot, but had a slot of its own on the MB. what it actually did was assist the processor when making large calculations.Same idea, just a different application.

Anyways the whole idea soon became irrelevant due to the fact that faster CPU's were developed.

any one else see this happening to the Ageia?
 
CodeWaste said:
i just thought of something, this has been done before. Anyone remember the optional co-math processor that was availible as an addon to computers like 12-14 years ago? It looked like it would fit in a ram slot, but had a slot of its own on the MB. what it actually did was assist the processor when making large calculations.Same idea, just a different application.

Anyways the whole idea soon became irrelevant due to the fact that faster CPU's were developed.

any one else see this happening to the Ageia?

That is because the CPU ate the FPU, but it still is there. If the CPU did not have an integrated FPU it would really suck at games.

==>Lazn
 
So maybe in a couple of years we weill see the Cpu eat the PPU, and it will be integrated onto the chip? just like the memory controller in AMD?
 
Lazn_Work said:
That is because the CPU ate the FPU, but it still is there. If the CPU did not have an integrated FPU it would really suck at games.

==>Lazn

and at some point the GPU will eat the PPU in the same way
but it will be some time befor that happens but mark my words in the next 5 years you WILL have some form of dedicated physics hardware in your gaming PC like it or not
 
CodeWaste said:
any one else see this happening to the Ageia?

It really depends how much physics are pushed and accepted and also how processors in the future work (we may be looking at 4 core cpu's for computers :D)
 
Some day we will have computers that are just one chip, with the CPU, GPU, Soundcard, RAM, Hard drive, and everything integrated. But not any time soon.

==>Lazn
 
Back
Top