Is this an all or nothing choice? The AGEIA API seems to become pretty widespread, it allows for software rendering of rigid bodies. Games are being coded to the API. This would probably mean that we get a PPU setting in games like we do with GFX cards these days. Ie turn on FSAA etc.. We get the option in game to run up the PPU setting.

Consider the effects! You can have a arrow hit a target and punch thru or splinter into a few pieces. Your sword can lop off an opponents arm in melee combat with the Tarantino fountain of blood following if so desired. It means a bullet you shoot at a target behind a tree goes thru it if it is a armor piercing round and fragments if it is a hollow point round. It also means if you have enough ammo you can chop down that tree in front of your target.

http://techreport.com/forums/viewtopic.php?t=30273

We've checked it out and it doesn't look like it will be useful for us (at least in the current incarnation). Also, even if they did support our types of calculations, what they mean by "physics" is fairly different than what we'd call. It's probably closer to "physics inspired." For games, it's perfect, for simulations, it might need to be more realistic (at least for our needs).

Take a look at this article on TheInquirer.net

http://www.theinquirer.net/?article=23468

Key quote:

"This hardware add-in card presents 40,000 bones (rigid bodies) on screen at a time, while even Dual core AMD or Intel CPUs can calculate just few hundreds. When it comes to Fluids, PPU can render even more rigid bodies. Your game will look like never before"

Here is a site with some videos shown at E3.
http://personal.inet.fi/atk/kjh2348fs/ageia_physx.html

If PS3 is using the ageia API, then the hardware implentation for PCs might get adopted by the game industry as it will make porting from one system to the other easier

I think there's an important distinction to be made here between whether this particular card will be succesful, whether the concept of a separate card for physics will ever be succesful, and whether games of the future will have highly detailed physics.

I think that the first and third are easy: These $250-$300 cards will not achieve widespread adoption at the current price point. However, detailed physics will eventually become widespread in games.

The question is, when will detailed physics become widespread and what kind of hardware will be used to make that happen?

One possibility is that it would happen with a much less expensive add-in card devoted to physics, but I doubt that will be the outcome. I think the outcome is more likely to be that the functionality of this chip (not this chip itself, just its ability to do what it does) will either be integrated into future multi-core CPUs, or it will be integrated into more general purpose GPUs.

In the meantime, the only people who will really benefit from this add-in card are software developers who want to start work now on a game to release in 2007.

After reading the article, it looks like the author missed 2 things.

(1) A PPU is not required in a console like PS3. Why? PS3's Cell is already doing that job via AGEIA's API! Why add another chip when the Cell is doing pretty much everything for you?! (Cell = multi-core)

I'm not sure about the Xbox 360, though.

(2) The author goes on with the issue of the PPU being supported by a game upon release. Has this guy talked to Tim Sweeny of EPIC? Does he know that Unreal engine 3 supports PPU and multithreading via AGEIA's API?

Don't believe me?
http://www.dfwfrag.com/article177.html
http://hardware.gamespot.com/Story-ST-x-1611-x-x-x

(BTW, I suggest you get an Nvidia GF7xxx series if you play games using this engine. *wink wink*)

All it takes is one game to come out that demonstrates PPU to the mass market. It'll obviously start slow, but eventually picks up.

LoneWolf15 (#5) says :
Downsides I see so far:

(1) Increased power needs
(2) Increased noise (yet another fan)
(3) Increased heat
(4) Limited bandwidth (PCI bus in an age when we're trying to get away from it)
(5) Need for yet another molex connector
(6) High price

(1) It ain't that much.
(2) Depends how cheap the fan is, and if it has temp monitoring to adjust fans speeds. Nothing is stopping you putting your own solution.
(3) Its another 20W of heat...A typical CPU does 80W+...What's another 20W gonna matter? Heck, a video card does FAR more.
(4) I'm not sure on that. There's no hard data one can assess which backs your statement. Simply put, we don't know. So one can't really make such a claim as a "bad" if we don't know if it really is an issue.
(5) You can't use a 4-pin Y-splitter?
(6) While I do agree the cost is a bit much, you can't expect to pay $100 on technology like this. Of course, you could wait a bit, let it become more widespread and production ramps up. (Although ASUS is known to charge a bit, resulting in increased product cost).

Whether or not you folks actually want to buy the PPU is up to you. But after living through the 3Dfx era (yes I still got a fully working Diamond Monster3D Voodoo 1), I suggest you not be "first adopters" (guinea pigs) just yet.

Being first adopters always results in you paying a premium...Then after a while, version 2 of the PPU comes out, which would likely to rectify the issues of the first...My general rule, never buy first generation. (2nd generation is always better).

Remember this joke :
"If at first you don't succeed, call it version 1.0"

Ahh, the good old days...I remember how much of an improvement adding the Voodoo 1 did on my old Pentium 150Mhz...Those were the days! :)

The more and more I think about it, the more this seems to parallel what 3D graphics went through almost a decade ago.

When PPU gets "real". CPU or GFX manufactors would have incorporated it. Kinda like FPU etc i old days.

My bet here is tho that nVidia and ATI adds it to their GPU´s as just another part.

For those yanking about Cell..Cell is just a a PPC basicly. And Xbox2 is PPC. both Xbox2 and PS3 can do this PPU through API. But none of the 2 will be as fast as the PPU card with a specialized processor.

Is the PPU something that will revolutionalize the gaming market like 3D? no. Will it revolutionize it like OpenGL and DirectX? No.

The PPU is more like having a surround sound system compared to a stereo. I mean..seriously..explosions that are more real and stones moving more real than random scripted. It´s not really something huge...

Most people wont ever notice it either. Since most people don´t even know how rocks should fall down a cliff or how explosions should be.

It spells HYPE.

Time will tell whether this becomes the next 3dfx (pre STB boondoggle, that is) or the next Transmeta:

--- 3dfx had a good idea (commodity 3D graphics) just as the market was getting the idea of moving in that direction, and rode on the wave's crest.

--- Transmeta had a good idea (a unique approach yielding low-power computing) just in time to be crushed by a tsunami (Pentium-M).

I have a hard time seeing PhysX catch on, though, unless they can get the pricepoint down to $100 or less and a couple 'killer app' game titles to go with it. Otherwise, there will be a Parhelia-style flash-in-the-pan product launch followed by a quick relegation to the novelty market when people discover that the tradeoff between ownership cost and real-life uses is far disproportionate, and better served by other configurations of other products.

They have a little tool which runs on most PCs
http://www.ageia.com/novodex.html
It runs nicely on this Pentium M without a PPU...

The question asked: Would you rather pay more for a PPU or for a second CPU core?
The PPU is really just for games. Assuming it does what is claimed for it on numerous games:
- If the PC is just for gaming, then the PPU makes sense instead of a second core in the short term.
- If the PC is for general use plus gaming: For AMD CPUs, they should get a dual core. For Intel, however, the PPU might be a better idea as dual core is not that interesting as it already has hyper-threading and the current dual cores are hot, power hungry, require a new motherboard, are fairly slow per core and somewhat inefficiently implemented. .

I hate the fact that what the PPU actually computes is never discussed. "Physics Processing". Humm, what's that? Probably that means 3D transformations. Probably that means bounding box intersections. MAYBE that means polygon/polygon intersections.
But I wonder how much the PPU is just a general simple vector processor with firmware to drive its primative computations?
In particular, a single Cell vector unit has 256K for a single SPE.. that's enough RAM to hold thousands of object bounding boxes and animated transforms, and the SPE has power to manipulate them. As much power as the PPU? We don't know. But the PS3 has SEVEN of these SPEs.
Anyway, until we see some technical details, the PPU is just marketing hype. Note I didn't say "tech demos" or even "actual game benchmarks" since both can be misleading since of course they'll be tuned and picked to show off the PPU is useful.
Now, what else I'm wondering, if the PPU is a general purpose computing pipeline... can we access it for other computes? Sigh, pointless to speculate until the marketing hype is turned off and the technical info is actually released.

Ryan Shrout makes the same points I have been making over and over again:

From the forums:

What people aren't getting is that at best, the addition of a PPU will give you only incremental benefits until market penetration is such that game developers can give up the uninstalled base and write games that take specific advantage of the unit. The mass market will not pay $299 for an incremental advantage, as a result, market penetration will remain low and developers will not take specific advantage of the card.

If a developer were to write a game that took advantage of this card's abilities to their best benefit, then you would have games that have deformable terrain and all sorts of effects. Unfortunately, things like deformable terrain are so integral to the engine, it would be very difficult to write a version that would still run on a computer without a PPU. So, to maximize the benefit of the card, the developer would have to write off a huge segment of the market. That's just not going to happen.

3D graphics were different. First of all, 3d could be rendered by software, so developers could write for the lowest common denominator AND the highest common denominator without much additional cost. Additionally, 3d became integrated into the video cards at little or no additional cost at the mainstream level. 3d essentially became free, which increased market share to the point where developers could write specifically to take advantage of 3d and do things that were not possible in software.

This last part is what the PPU is missing. Until it becomes essentially free, developers will not be able to write games that require a PPU.

I won’t spend a dime on it until its absence in my machine really starts to affect gaming. I would gladly spend $300 on anything else in a computer, heck I would even go watercooling.

if it were me, I'd pay up to $150 for this...and I'd be glad to get something like this in my system, or even just let the manufacturer know that there there is demand..I'd consider it a donation to the advancement of gaming technology

This reaffirms what I said in the forums, if this thing takes off it will not change gameplay much, at least not for quite awhile. What it will do is allow more eye candy because graphics are scalable and hardware physics acceleration can be a beautiful thing. Gameplay isn't scalable however.

I agree with what most people are saying, this tech should be lumped into GPUs if only for bandwidth reasons alone...

Does any one remember how much was 3DFX monster when it originally appeared? (just to compare to $299)

No tangible benefit whatsoever when you purchase a PPU. Come to think of it, it sounds just like the kind of thing only rich people would have. I mean, seriously, a card that enhances games only? :D I'll just get a PS3, which doesn't need to be set up and can already do physics calculations on its own.
What do you expect of a guy who can't even afford a Logitech MX310?

Would you rather pay more for a PPU or for a second CPU core?

Second CPU core. More than just games get a benefit.

Expect it's missing one cruical piece, a killer-app like "GLQuake".

I don't feel the situation is analogous. We were getting 3D from all directions, including rather half ass implementations like the Virge.

We got more than one killer title and we had several existing titles that made the transition.

You also had a very work related use in CAD, etc.

That helped fill the market. I really don't see that here. I see a gamers toy alone that serves no other function. At least a 3D capable video card can spit out a Windows desktop. What does this $300 item do without tailored software? Nothing, except create heat and raise your power bill some some trivial amount.

Even 3dfx realized integration was the way and that the add in board wouldn't fly forever.

This product remains targeted wrong. $300 put into a dual core CPU you get a product that has more uses than just gaming, that has true tangible benefits today. Not to mention a product that fits within the existing paradigm.

The existence of companies like Alienware and their success I suppose proves there is some fringe element out there that spends money, but catering to them doesn't win you vast market penetration or even make you a Dell.

The whole product seems derived from the idea that gamers will throw down and spend money on anything that someone claims will make games better.

They may get the fringe to run out to do it, but that won't pay the bills. If they want to win a market, then they went the wrong direction.

Um, so why the heck havent ATI or nVidia bought AGEIA yet? These PPU cards are never going to be bought by anyone but the same people who spend the money to get a good 3d gaming card anyways, so why not have them on the same board. Looking at the design of the card it looks identical to a graphics card.

Whichever graphics company puts physics processing on their gpus first will be ahead of the game, but its when there is some competition in that area that it will really take off.

Downsides I see so far:

Increased power needs
Increased noise (yet another fan)
Increased heat
Limited bandwidth (PCI bus in an age when we're trying to get away from it)
Need for yet another molex connector
High price

I think it's a great concept. But in an age where a good 3D card pretty much starts at $200 (a good budget card being another animal and starting at about $110 for a PCIe model) $249 to $299 is quite high in price for the average PC gamer. It certainly won't be an impulse buy, so that can't drive sales like a $99 part might.
In the age of the 3dfx Voodoo, most vendors were able to code a patch for their games (Tomb Raider, GLQuake, etc.) to add Glide ,Direct3D, or OpenGL support, without much issue. I don't see it likely that it will be as easy this time around, which could greatly hamper sales. I'll be waiting until a cheaper part based on a 90mm process, with PCIe support comes along, and at that point we'll see if there's enough game support to make it worthwhile.

Somehow I don't think PC gamers are going to run out and buy $249 cards to get enhanced physics in some games. Since developers can't require a PPU, they can't design the games around the chip - only offer extra effects for users who happen to have one. The only way this is going to work on the PC is if ATI and NVidia license the tech, and include PPUs on their cards, making the userbase larger.

Though what we really need are general purpose coprocessors, à la Cells. 3D graphics, 3D sound, and 3D physics all need similar operations, and I don't see why one architecture couldn't handle them all.

w00t....reads review ...errr. preview !