AIPU

@drizzt81
Yes it sound interresting.
But here is some thing to add to your points.
1 ) Game engine are for a long time more or less a total package wich include a AI API in a basic form or feature rich. So AI API are very common. But pure API are rare, mostly the are included in a licence game engine as a compleet feature package also including Physics.

2 ) AIPU comes with a API. This means Code reuse. AIseek AI gure's do the grunt work. Just like for all other AI API's. Dev's use it trough API intergrating it with there highlevel AI interacting with the Game mechanics. Code reuse means less work faster result. faster time to market no AI guru needed. But a middle common level AI team member.

3 ) The Dev community is more AI aware in the latest years. They know the importance but most dev's depend on there publisher.Deadlines crunching time. cutting features. And have to deal with them. As they work for them. So API are important Suporting a AIPU means not only better AIs but also at a larger scale and finer detail.
4 ) This AI awarness is only notaicable when after years Dev put this new balance focus of game design to practice and it take time to develop a engine and a game. So wait for the Nextgen game engines and how the AI will be used.
5 ) As the tools. New game engine like Unreal 3 uses UnrealEd3.0 were Level designers and sripter can graphicale drag and drop a AI desision tree. Next gen Game tools will incorporate more psysics and AI editing features.

AIseek cliames 100times more AI lowlevel Power. So they claim to what? Wich means larger or finer or secure low level AI. This is a heavy load. Where High level AI is based on top of it. Where the CPU don''t have to do a lot off the low level AI stuff, wich is done better and at a way larger scale. The High level AI load increases due to the underling higher low level AI load.
And further computing in the higher level AI desicion routines.

So it will not off load the AI computing of CPU it will mostly enhance AI using also so the CPU in tendem with the AIPU. For more enhanced AI in many way's.

While the progress of better CPU power because CPU scale up in IPC klok and cores. Stupid AI can be avoided in the future with more ease. But that does not mean throw the ancker out go no further we are happy now. So stop. Put more growing power in graphics again. Well I look further than stupid AI behavior the gameplay braker. AI can also enhance gameplay further and also immersion. An FPS is just one genre.

As a gamer I want at least decent AI. Better enhanced AI way beyond that it even better.
Bad AI ruins a game.
 
Immacolata said:
Ive tried to convince supergee, but he is adamant that such a device is a good idea.
Click to expand...
Yes I am more prepared to pay for a AIPU then for a CF/SLI solution. Graphics is important but not if other thing stay way behind :)
And for the record, I do not believe the future will be specialised ASICs, but a redefintion of what the CPU is. You will still have one product doing your calculations, but perhaps it will have the work subdivided so that the future "CPU" will really be a grid of processors each doing something.
Click to expand...
But the way I see it. It comes but it will be 2010 to 2015. So nice for UT2012.
Do you think for a Hard core gamer his gamerig and PPU will last that long. Within 3 years or less it will be replaced by something more powerfull to support the games wich come then.
I build a game rig for games coming in 2007 to 2009 afterthat the situation may be different. The possible glory day's of Asic's. Depends on the stance of games then. But if I want to play game in more optimum state in 2007 i need a PPU and or AIPU. If they brake through and enhance more then a few games.
While in a game in 2011 a Cell like x86 25nm beast with 4 cores and a array of 16 CoPU will do. It's the case of the near future vs longterm future.
But there could be competition off a PPU4 and AIPU3 the high-end solution while such CPU would be a budged solution.
Games will have Sliders. That option is also possible.
I doubt we will see a large boost of cards and addons. It is not in the interest of the cpu makers (and they really are the ones with the power), nor in the interest of the consumers.
Click to expand...
you forget the power off the dev's. If there is decent support for Asic. Those Games will attracked some gamers to buy stuff to enhance there game the way it ment to be played. It's just a desision to pay for that extra hardware or not. A choice that a few will do to be true, most not. But that is the way a new market get kickstarted. That it possible die or not over 5 to 7 years later. Thats is to far away to make sense for people who want to play nextgen games in 2007 in optima
At least not if you don't want to make building your own a computer even more of a hassle than it is today :)
Click to expand...
Well for CF and SLI you need a CF or SLI mobo. If you want some more Asic's you need a mobo that is flexible to support some.
Gamers who will keep taht option keep it in mind for choicing the best mobo.
 
aipu will never be feasible, devs cant make a game with a slider/option for AI Intelligence - it would change the gameplay entirely and make it take forever to debug and such
 
SuperGee said:
Yes I am more prepared to pay for a AIPU then for a CF/SLI solution. Graphics is important but not if other thing stay way behind :)
Click to expand...

Good luck enjoying your dead end technology that devs will not develop for. Just like the ageia PhysX.

The possible glory day's of Asic's. Depends on the stance of games then. But if I want to play game in more optimum state in 2007 i need a PPU and or AIPU. If they brake through and enhance more then a few games.
Click to expand...

There won't be a glory day of third party ASICS. It will be Intel or AMD that will make a new kind of CPU which is build from many discrete units if anything happens in that direction. Besides as discusse 100.000 times now, AI is genereal purpose calculations, no need for special AIPUS to handle it before there's a very well defined code library to run AI. And there isn't. Yet. Remember 3d-graphics and 3d-libraries was invented before the 3d-accelerator. There isn't yet a shelf-product library for AI software used in games as far as I am concerned. Lets have one of those first before senseless ramblings and fantasies about AIPUS.

you forget the power off the dev's.
Click to expand...

HAHA! Good one. These days the power of the devs are measured in board room meetings where needle striped financial types estimate the profitability of all things gamey. They see AIPU and they think expensive extra time spend developing for a minority market segment EQUALS idiocy. Won't do it. Let some other schmuck try it. WE are innit for the money. Keeping it Simple, Stupid and as profitable as possible.

If there is decent support for Asic. Those Games will attracked some gamers to buy stuff to enhance there game the way it ment to be played. It's just a desision to pay for that extra hardware or not. A choice that a few will do to be true, most not. But that is the way a new market get kickstarted. That it possible die or not over 5 to 7 years later. Thats is to far away to make sense for people who want to play nextgen games in 2007 in optima
Click to expand...

A terrible lot of ifs and buts are hanging on this AIPU thing. Don't forget, we are moving towards ever more consoles and less pc for gaming. The addin market on pcs is diminishing year by year.
 
Immacolata said:
Good luck enjoying your dead end technology that devs will not develop for. Just like the ageia PhysX.
Click to expand...
I use PPU for some games. And more to come.
But it's to early to say if ageia will made it. in its new born Physics hardware market. Wait and see for two years.
AIPU I think the Dev support must first start. so is in a very more early stage.
Where Ageia have decent dev support. wich can become more in time.
AIseek must start with that. And the acceptance in that world off Dev's is unkown by me.
No Dev's no games.
So first step to succes starts with Dev support even before the hardware is out.
There won't be a glory day of third party ASICS. It will be Intel or AMD that will make a new kind of CPU which is build from many discrete units if anything happens in that direction. Besides as discusse 100.000 times now, AI is genereal purpose calculations, no need for special AIPUS to handle it before there's a very well defined code library to run AI. And there isn't. Yet.
Click to expand...
Well I thing you need to do some home work. As there will be a AIPU wich means some things like low level AI is very common thru games and even in differnet genres. So there can be made specialised hardware for. And a API that merge gameengines to HArdware accelerators.
If yoy did your homework you will know that AIPU only accelerate the common low level AI. That very computational to do it correct in a timely fashion.
Remember 3d-graphics and 3d-libraries was invented before the 3d-accelerator. There isn't yet a shelf-product library for AI software used in games as far as I am concerned. Lets have one of those first before senseless ramblings and fantasies about AIPUS.
Click to expand...
Because the Render part of Game engines are the Core features. Wher emost dev's focus on. AI is a secundairy feature wich is often put in a total soultion to make a Game engine complete.

So i you wnat a AI lib you get more choice to choose for a complete Game engine with AI lib within.
HAHA! Good one. These days the power of the devs are measured in board room meetings where needle striped financial types estimate the profitability of all things gamey. They see AIPU and they think expensive extra time spend developing for a minority market segment EQUALS idiocy. Won't do it. Let some other schmuck try it. WE are innit for the money. Keeping it Simple, Stupid and as profitable as possible.
Click to expand...
Well Gamers deside if something can be a sucses. But before that. It must first be picked up by the Dev's. No Dev support it will fail even before gamers come into play.
A terrible lot of ifs and buts are hanging on this AIPU thing. Don't forget, we are moving towards ever more consoles and less pc for gaming. The addin market on pcs is diminishing year by year.
Click to expand...
Well we got a pure Pc game Crysis.
And that only this year wenn those nextgen are fresh but will be out dated soner then later.
And I see it differently PC Game market is big. Console market is just bigger then them.
 
Motherboard builders are going to need to start making the boards with 5 pci slots again to hold all of these fancy pants stuff.
 
I noticed a Mobo with Via Hardware accelerated onboard sound with the Envy24 chip or something like that.
On addon away room for one more.
Well Lookin out for a very extendable Mobo would be important for the future update and specialized hardware.
 
Ever played a game like Civ 4, GalCiv, et al, on a "huge" map? Seems like there's a heck of a lot of things that need to be calculated in-between turns when you get far along in the game.

Could RTS games benefit from this tech more than any other game? FPS games might have you handling anywhere from 5-20 enemies at once(maybe 30-40 if you're playing seroius sam, but the "I" is kinda missing from seroius sam's "AI" :p). RTS games have hundreds of units in play at any given time.

The prospect of controlling an army of thousands of units is promising.
 
With PPU and AIPU, I hope they make enough slots on the MB bro....I may have to buy a bigger case for my 4 X next gen R600 system...... :D
 
SuperGee said:
So with all those Addon mayhem from
Ageia
AIseek
ATI GPGPU HavokFX

The Mobo choice get important.
Click to expand...

Why not place both the physics processor and AI processor on the same card? Neither requires a great amount of bandwidth. I wouldn't think twice if there were a $150-$200 card which offered both physics and AI acceleration in games. Heat would also I imagine not be an issue, at least in the near future an AI processor or physics processor would not approach the number of transistors already on video cards.

Actually, instead of calling it a PPU/AIPU they could simply name it a "Game accelerator." Write an API (added to DirectX) which is capable of offloading physics and AI processing to either seperate CPU cores or a dedicated game processor. Whatever abilities the card lacks could be implimented in software mode. Even better, all those who are morally opposed to a PPU could use those extra cores for software-mode physics processing.

At the core of all this pointless arguing is one false, though common, conclusion:

Physics must either be processed on a CPU core or PPU - if we pick one then the other can never ever be used for physics processing ever again.

This is just flat out wrong. It would be very possible to impliment an API which transparently would enable to user to take either path: PPU or Core Duo.

History will judge whether further CPU cores truly can equal the abilities of a physics card. If so people will not purchase PPUs because software mode physics would be nearly just as fast. The PPU camp was wrong. If not, then all the people who are now in the Multicore camp are wrong. No need for further argument for the time being.
 
Hvatum said:
Actually, instead of calling it a PPU/AIPU they could simply name it a "Game accelerator." Write an API (added to DirectX) which is capable of offloading physics and AI processing to either seperate CPU cores or a dedicated game processor.
Click to expand...

Why does everyone want to make stuff exclusive for windows? Why not have an open API, like OpenGL?
 
Hopefully it fails because I don't feel like spending over $1,000 on random cards to just be able to play games.
 
drizzt81 said:
Why does everyone want to make stuff exclusive for windows? Why not have an open API, like OpenGL?
Click to expand...

I'd love it to be an open API. I just used DirectX as an example so all the Microserfs on these forums don't get confused. Afterall, if they can't handel a physics processor there's no need to also ask them to accept anything as foreign as cross-compatibility, security, open standards or good programing practices.
 
dotK said:
Hopefully it fails because I don't feel like spending over $1,000 on random cards to just be able to play games.
Click to expand...

You'll never be forced to purchase a PPU. Physics processing will need, out of market necassity, to also have a software processing mode if it's going to become succesful. If you want your games to really shine you'll need a PPU - but if you don't get one you will not be any worse off then if the PPU had never existed in the first place.

Honestly, everyone wins with the PPU. I don't understand why half the people posting here are all deathly afraid of it. Your Software/Dualcore physics processing WILL NOT GET MAGICALLY SLOWER JUST BECAUSE YOUR NEXT-DOOR NEIGHBOR HAS A PHYSICS CARD.
 
Hvatum said:
You'll never be forced to purchase a PPU. Physics processing will need, out of market necassity, to also have a software processing mode if it's going to become succesful. If you want your games to really shine you'll need a PPU - but if you don't get one you will not be any worse off then if the PPU had never existed in the first place.

Honestly, everyone wins with the PPU. I don't understand why half the people posting here are all deathly afraid of it. Your Software/Dualcore physics processing WILL NOT GET MAGICALLY SLOWER JUST BECAUSE YOUR NEXT-DOOR NEIGHBOR HAS A PHYSICS CARD.
Click to expand...
If it catches on, you will eventually HAVE TO buy a PPU if you want a decent gaming experience. The same holds true for the AIPU.
 
But that means it must become a big sucses with very large Developer support then it will be more often a requierment. If it have a meager sucses, that it can stay alive in the game market it will never come further then a optional hardware for games. With a exception here and there.
 
dotK said:
If it catches on, you will eventually HAVE TO buy a PPU if you want a decent gaming experience. The same holds true for the AIPU.
Click to expand...

... and if it catches on then that would mean that the idea would have caught on with people. Get it? Uncle Stalin isn't going to go force half the gamers in the world to go purchase a PPU. If half the gamers in the country have one, it will be because half the gamers in the country decided to get one.

Of course if you're an old-timer nostalgic for the days of software physisc, you'll be in the minority. Then again why everyone else should forgoe anything? If Hydrogen fuel cell vehicals come out I'm not going to refuse to purchase one because I'm afraid how it will make all the people with old cars feel. The world doesn't change to suite our desires, unfortunately.
 
He makes no claim that the world follows (or should follow) his desires. He is just saying that he hopes it does. This is perfectly normal sentiment.
 
That makes me think about XP vs Vista 3D card needs.
3D accelerator GPU ars so common. If you buy a PC. It will have a 3D card. Even Nongamers will have one. this Makes Most PC at least light Game rigs.
Now with Vista for optimal OS use you need a 3DCard. this means They put in Vista ready rigs at least a Dx9 GPU solution also optional a Dx9 onboard solution.
This GPU are often the budged ones. X1300

Wich make me think if PPU will be requierd and for that it must be a Sucses first. By then it would have a full Produkt line. Onboard to High-end wich means a $50 to 100 solution would be possible while the enthausiast go for that $450 version.
Wenn It will be requierd but $100 you won't notice much on a $2000 PC.
The Budged / Component shifting game.
 
creative should just come out with a physics, AI and sound card all in one.. then maybe it will catch on
 
mjz_5 said:
creative should just come out with a physics, AI and sound card all in one.. then maybe it will catch on
Click to expand...
nice, so you have a card that can't do three things.
 
drizzt81 said:
nice, so you have a card that can't do three things.
Click to expand...

not to mention that it will have proprietary connectors for everything which will never be released to the public :p
 
SuperGee said:
By then it would have a full Produkt line. Onboard to High-end wich means a $50 to 100 solution would be possible while the enthausiast go for that $450 version.
Click to expand...

The difference(the big one) between physics cards and graphics cards is this; The textures, polygons et al. have nothing to do with the gameplay. I don't remember the exact r_picmip value I used to use when I played RA3, but it was enough that the game looked like it was incredibly old. This didn't affect the gameplay except to make it easier for me to see things.

Physics, on the other hand("gameplay" physics, not effects physics), doesn't share this benefit. If you want to have a level with a bunch of boulders bouncing around that can kill/hurt the players, all players must have systems capable of computing the scene accurately. How does the $450 card's performance "match-up" to the performance of a $50 on-board version? If someone with a sub-standard system wants to play quake 4 multiplayer, they can. They might have to turn the graphics all the way to minimum, but that won't actually impact the gameplay itself. If someone with a $50 onboard physics solution wanted to play a game that took advantage of the $450 solution, the gameplay would be affected at a massive scale.
 
jimmyb said:
He makes no claim that the world follows (or should follow) his desires. He is just saying that he hopes it does. This is perfectly normal sentiment.
Click to expand...

That's fine. But he should at least recognize the fact that if it catches on and he still refuses to buy one he's pretty much the Amish equivalent of a computer gamer. An old timer that's lamenting the "good old days" when people still rode horses and dentistry meant pulling your tooth out with a pair of pliers.
 
Spewn said:
The difference(the big one) between physics cards and graphics cards is this; The textures, polygons et al. have nothing to do with the gameplay. I don't remember the exact r_picmip value I used to use when I played RA3, but it was enough that the game looked like it was incredibly old. This didn't affect the gameplay except to make it easier for me to see things.

Physics, on the other hand("gameplay" physics, not effects physics), doesn't share this benefit. If you want to have a level with a bunch of boulders bouncing around that can kill/hurt the players, all players must have systems capable of computing the scene accurately. How does the $450 card's performance "match-up" to the performance of a $50 on-board version? If someone with a sub-standard system wants to play quake 4 multiplayer, they can. They might have to turn the graphics all the way to minimum, but that won't actually impact the gameplay itself. If someone with a $50 onboard physics solution wanted to play a game that took advantage of the $450 solution, the gameplay would be affected at a massive scale.
Click to expand...

Why does that matter. Gameplay would be effected, so you have different people playing a slightly different (more or less detailed) game. Do you feel it's unfair that someone would have a fundementally different gaming experience instead of just a higher resolution or greater texture quality? Life isn't fair, if you're poor then you're going to have to deal with a smaller number of large boulders instead of many small ones. I'm not going to forgoe a greater level of detail because someone else is stingy.
 
For multiplayer games that certainly wouldn't work though, since everyone needs to be playing in the same game world.

For single player games it can work to a certain extent (i.e. 10 vs 20 boulders), but there are still limitations in how far the game can vary depending on hardware (i.e. a fundamental physics puzzle would have to work on all hardware configurations).

The approach that Havok FX takes according to their webpage is to accelerate effects physics in hardware (since these don't affect gameplay), and to compute gameplay physics in software. Note I'm not commenting on whether this is a good or bad approach; that topic is another discussion altogether.
 
jimmyb said:
For multiplayer games that certainly wouldn't work though, since everyone needs to be playing in the same game world.

For single player games it can work to a certain extent (i.e. 10 vs 20 boulders), but there are still limitations in how far the game can vary depending on hardware (i.e. a fundamental physics puzzle would have to work on all hardware configurations).
Click to expand...

What sort of stupid developer would center their game around the movement of boulders? That's not even fun... Blowing things up on the other hand is. And that does not need to work on all hardware configurations.

jimmyb said:
The approach that Havok FX takes according to their webpage is to accelerate effects physics in hardware (since these don't affect gameplay), and to compute gameplay physics in software. Note I'm not commenting on whether this is a good or bad approach; that topic is another discussion altogether.
Click to expand...

For single player games it would work to an infnite extent, if done correctly. If you are poor or a paleogamer you can still not have a physics card but simply miss a great deal of depth. Those with money/not afraid of new technology can experience fully destructable buildings, large numbers of physics interactions, more accurate fluid effects, more interesting weaponry. Those without physics cards can continue gaming in the immobile universally invulnerable environment of today's games.

Still, please remember that technophobes are not any worse off for the physics card simple existence. They'll simply be missing out on a great deal of depth and accuracy, but developers could very easily make games playable for those without physics cards. In fact if the Technophobes simply stopped reading about physics cards, lived under a rock and went to Gamestop once a year to buy their games they would have no reason to even believe in such a thing as a physics card.

Bottom Line: Physics cards are great for people like me who have a bit of money and welcome new technology, and they don't hurt the people who are either stingy or phobic of change. This isn't an "Will everyone have physics cards or will no-one have them" question, rather it's a "How many people will have physics cards and how good will they be" question. I don't see the dillema.
 
Hvatum said:
What sort of stupid developer would center their game around the movement of boulders? That's not even fun... Blowing things up on the other hand is. And that does not need to work on all hardware configurations.
Click to expand...
On certain levels that's true, like blowing up a house and crushing the enemies inside. My point was that the successful completion of a game cannot be contingent on the completion of a physics puzzle that will not run on all supported hardware (i.e. the game runs without a ppu, but it is impossible to complete).
 
jimmyb said:
On certain levels that's true, like blowing up a house and crushing the enemies inside. My point was that the successful completion of a game cannot be contingent on the completion of a physics puzzle that will not run on all supported hardware (i.e. the game runs without a ppu, but it is impossible to complete).
Click to expand...

I agree completely, but that seems like a very rare design element and a somewhat oddball situation. It's a bit stranger, but even today you have the same thing: A developer couldn't make the successful completion of a game contingent upon the user's computer being able to correctly render the advanced features in DX9 or OpenGL 2 (It's possible, you could have a puzzle that involves identifying objects behind warped glass or something). ;)
 
I could see the use of an AIPU packaged WITH a game. Like have a Final Fantasy XVII logo on it or something :p

It could certainly be fairly useful, but I don't think that it would be for a large amount of games.
 
You must log in or register to reply here.
Back
Top