Sign in with
Sign up | Sign in
Your question

AGEIA

Last response: in Components
Share
December 13, 2005 2:32:09 AM

If any of you could ask a question form AGEIA, the new PPU company, what would you ask? Yeah, kind of wierd, but hey, just wanna know.

More about : ageia

December 13, 2005 8:19:17 AM

I would ask them why they haven't released any cards to reviewers. Is it because games wont support it yet? if it because they are still working to create it?

For $250 that card beter be packing on hell of a punch to make it worth it.... considering i could remove that card and upgrade my GPU and CPU by a giant margin with that $250
December 14, 2005 3:55:51 PM

I would ask this: do you have any plans to sell your company to nVidia or ATI?

That's what's going to happen; that or bankruptcy. Add in boards for physics will never take off. It's simple: people aren't going to pay $250 just so game physics are more "real."

The logical decision for a company like AGEIA is to sell their technology to nVidia or ATI so it can be incorporated into the GPU. From a business and profit standpoint, this is the only way AGEIA is going to make money.

Don't get me wrong... I think a processor dedicated to processing physics is a good idea- but not in the form of an add-in board. And certainly not for $250. The logical place for this is on the video card. Eventually ATI and nVidia are going to create their own solutions and when that happens... no startup company like AGEIA will have a chance.

-mpjesse
Related resources
December 14, 2005 6:26:01 PM

Off loading physics to the Aegia card will increase performance. When people pay $600+ for a graphic card that gives a small performance increase, or even a thousand for a large increase, whos to say they won't pay $250 so their computer can handle fancy physics.
December 15, 2005 6:32:20 PM

good point. but remember that people like that (including me) are a very, very small minority. certainly not enough for a company to be successful on.

-mpjesse
December 16, 2005 7:01:51 AM

Yeah thats true, but if Wikipedia is to be believed(along with the many other sites that pop up with 'aegia revolution' Nintendo's nexgen will have the PPU, so that adds the small majority of Nintendo fans(me and some others), and mpjesse's extreme speed gang. Aegia could do pretty well all alone.
December 16, 2005 10:08:22 PM

AGEIA could make money in making software physics engines. Lots of developers license physics engines like HAVOK.

I wasn't aware that Revolution was going to have AGEIA's PPU. That'll definitely bring in revenue. Nintendo will undoubtedly sell at least a few million Revolutions.

-mpjesse
May 8, 2006 6:18:37 PM

@mpjesse

I read your post with interest as I believe that Ageis could possibly be swallowed up by either ATI or nVidia although I'm not entirely sure they are planning for that.

I believe there IS a market and Viability for a dedicated PPU in todays PC market, even for £218 at launch.

For one thing, the card is so expensive because Ageis is currently only making money from the actual card sales, not from liscencing their SDK to developers. As market penetration widens (may take a few years to fully feel the benefits) they will be able to start charging developers for the use of their SDK because there will be a larger customer base of installed units which equals a lot of potential customers looking to buy games for their hardware. Until they can start to charge for the SDK, the price of the cards will likely stay on the high side, although considering there are people out there willing to shell out over £15000 for a Monster Gaming rig with 4 GPU's that don't output 4x the framerate, can you honestly say people will not pay £218 for a dedicated PPU?

I remember the days before 3D accelerators, when the first 3DFX card came out for £300 and the PowerVR came out at about the same, There were no/not many titles available at the time, but I for one bought a PowerVR and was then forced to buy a 3DFX as the PowerVR turned out to be a lame duck. To put it in perspective of the PPU, The first 3D accelerator was only just that, it was not a GPU and still people bought it in their droves as it promised so much at the time, fast forward and in hindsight those cards were only the first step on a rollercoaster evolution. 3DFX was eventually aquired by nVidia for their patents and technology. But look at the technology and the market for 3D cards now, unbelievable to think that there are 4xGPU solutions out there primarily for games, and they sell for over a grand...Who would have thought?

If Ageis can actually sell enough units and get the prices down, PPU's will definately be here to stay, I could even envisage a PPU port being created, why not? Before AGP would anyone think there would be a Dedicated Graphics port? and by this time there may even be an AiPU out or on the way, If you sell it, they will buy... ;-)

The key for Ageis is to get their market share up before nVidia and ATI offer their Hardware Physics for free effort as, if it challenges the PPU concept strongly enough, it could be a travesty for Ageis. Why buy a card for £200+ if you can spend the same on a second GPU and get better visuals AND Hardware Physics? This could seriously tighten the screws for Ageis.

If it was feasable to put a PPU on the Graphics cards then I'm sure ATI and nVidia would jump on that as I believe the market is there for Hardware Physics and both companies would see it simply as another way to squeeze more cash from us.

I personally would prefer the dedicated PPU card take off at the level that GPU's have over the past 10 years. Anything that brings closer the true 'Virtual Reality' is good in my books.

Creative are making inroads with the APU concept with their fantastic X-Fi range of cards, you have to hear it to believe the differance. So perhaps the future of the PC architecture could be a Dedicated Processor for the 5 main tasks, GPU for Graphics, APU for Audio, PPU for Physics, AiPU for Artificial Intelligence, CPU for (Load Balancing) and anything thats left ;-). If each were sold as Processors similar to the way we upgrade our CPU's I belive this would be the best route to take. Then you would buy a motherboard with 5 Sockets, one for each of the Processor types, either a Shared RAM subsystem or different RAM for each Processor type. This would allow for the ultimate base for a performance PC

I am of course getting ahead of myself, but I simply wanted to paint a picture, that I believe to be feasable, of where the PPU could take us.

I strongly believe the PPU, in one form or another, will take hold. Will it be Ageis that lasts the distance? Only time will tell...3DFX the champion of the early 3D accelerator market was gobbled up by nVidia, will Ageis suffer a similar fate?

I look forward to your views on my ramblings ;-)

JKay6969
May 8, 2006 7:02:42 PM

First of all this card is coming out tommorow (May 9th) and I'm not buying it yet. I would wait till they have some good reviews about this product to wether it's worth having these card in your pc and most importantly spend $250-$300 for it. I think this is a great idea of a product, given time and experience I'd say this would be in every high performance gaming rig.
May 8, 2006 7:55:42 PM

If I could ask a question of Ageia I would ask....
Why did you saddle a good idea with the slow PCI bus?

I think dedicated physics hardware is a wonderful idea, that will one way or another be a central part of the future development of games.
Whether or not the Ageia PPU will have the time to mature, I can not say.
Whether or not Ageia will survive the comming year, I can not say.
Whether or not an independent PPU is the best solution, I can not say.


What I do know is, unless there is some major benefit to a PPU, the price must go down.

I cannot see there currently being a big enough market to support Ageia with the PPU at the price of $300.
May 8, 2006 8:18:30 PM

Sticking it on the PCI bus was a necessity bc there isnt any other slot to put it in. You have to remember that this card and processor has to be years in development, PCI-E is still only a couple of years old, and aside from a few BASIC cards, the only PCI-E cards are video cards

As far as speed is concerned, the PCI bus isnt very slow per se, it just runs at 33 mhz, which if you are thinking in terms of CPU speed, yes that is slow, but that PPU isnt running bloatware (windows, ... linux... any OS) its getting nothing but pure machine code, 33mhz is really fast when you stop to think about it


Further example:
My watercooling system has a safety computer that i made for it, it runs on an ADUC812 CPU, thats 11.092 mhz. Its WAYYYYY more than i need and i'm reading 8 analog inputs at 12 bit resolution 8 times a second, and then preforming 4th order liner equations on each digitized line. It then runs thru a comparison routine that checks to make sure i'm not too hot or cold. After i got the core of that program running swimmingly i added things such as date, time, runtime, and a datalogger for statistical analysis to the code and it still runs thru the protection code loop in a few hundredths of a second.

I will post my rig later in the appropriate forum, I still havent finished prettying it up.
May 8, 2006 8:22:09 PM

Holy sh1te that was a huge bump on these boards!
May 8, 2006 9:03:34 PM

Quote:
Nintendo will undoubtedly sell at least a few million Revolutions.


Remember. Its the Wii now. In the words of comic book guy. Gayest name ever.

I'll buy the board if it gets good developer support. The biggest problem is where to put it. I might want to do dual GPUs at some point and on most motherboards you only have 1 PCI slot left which I'd use for a sound card. They need to start making motherboards with the slots a little farther apart. At least for dual GPU boards.
May 8, 2006 10:07:47 PM

Very nice setup you have there.

I, however, differ on my interpration of things.

I don't feel your comparison is justified because the PPU, right now, is basically a GPU coprocessor, and almost anything to do with video takes a huge amount of bandwith. Furthermore, if the PPU is ever to really shine it is going to have to do a heck of alot more than be a graphics coprocessor.

The PPU is supposed to be able to do 20 giga instructions per second.
Keep that in mind as you read the next part.

My justification that the PCI interface was a poor choice is that the PCI bus has a theoretical 133 mbs max (realistically 110~130), and that is shared accross the whole bus.
And I am pretty sure that max theoretical counts traffic both ways.
Which is to say if you have 120 mbs going up the pipe (toward the CPU) then you have next to nothing going down the pipe(toward the card).

If I am correct then, the PPU is being starved for bandwith both ways.
This is all the more true if the PCI bus is being used by anything else.

They could have gone with PCIe since it is two years old by now and what I assume to be their intended audience will have PCIe motherboards.
The PCIe 1x standard allows 2.5 gbs either way simultaneously, for a max of 5 gbs total. Realistically it is less but should be much more than enough.

About the only advantage I can see with going for PCI is it will be cheaper for the company to make than PCIe. And they actually have a PCIe 4x version slated for release latter this year, though. So until then the PPU is a waste.
May 8, 2006 10:28:42 PM

price wise, i mean 200-250 is the retail price, etailers and ur local custom shops will sell them for much cheaper as they do with gfx cards
May 9, 2006 6:20:01 AM

Quote:
Nintendo will undoubtedly sell at least a few million Revolutions.


Remember. Its the Wii now. In the words of comic book guy. Gayest name ever.
Yes, it is now, but when mpjesse was writing that, it was named Revolution. It was half a year ago.

This be an old topic.
May 9, 2006 1:00:41 PM

Quote:

I don't feel your comparison is justified because the PPU, right now, is basically a GPU coprocessor, and almost anything to do with video takes a huge amount of bandwith. Furthermore, if the PPU is ever to really shine it is going to have to do a heck of alot more than be a graphics coprocessor.

The PPU is supposed to be able to do 20 giga instructions per second.
Keep that in mind as you read the next part.

My justification that the PCI interface was a poor choice is that the PCI bus has a theoretical 133 mbs max (realistically 110~130), and that is shared accross the whole bus.


Yes, your thoughtas are correct albeit they still seem a bit biased toward relating cpu and other hardware performance to a microprocessor

Understand this, I am NOT trying to insult your intelligence as you are also not trying to insult mine.

Now, for those that don't know how the processing inside of a chip works, and for those that just want a bit of clarification, 20 giga - 20 billion instructions per second, sounds impressive, now lets break it down.

32 bit chip - executes 32 instructions simultaneously
20 billion by 32 is 625 million operations

now assuming ti does all of those in one second that 625 MHZ
thats one heck of a chip, my video card doesnt even touch that one

now heres the kicker, how do we get that chip on the 33 mhz pci bus a 100 mbs to perform with any decency, well that takes hetradyning (fancy name for frequency conversion) and coding, remember, our AGP bus is only 66mhz....

yes video takes a TON of bandwidth because you have to render everything

but this physics engine... no rendering, its just tellign the GPU how something should be rendered... now insted of passing advanced complex polygons, we are back to hex code, and really doesnt need a whole lot of bandwidth.

Is the pci bus enough to run this card, in my mind undoubtedly, however, I wouldnt use it simply because I'm already loading my pci fairly heavily with a scsi controller and an audigy 2 sound card

If the scsi card wasnt there then I would probably be one of the first people to give it a go

Yes, they should have put it on a pcie platform for the various reasons of advancing tech and speed and such

I think that was an engineering call because it is far easier to design a part that utilizes a parallel bus, rather than a serial bus, trust me its just easier to physically do
!