$30 Physics Processors why not?

[PureScoundrel]

Why are there not any of these things around for $30 or so? It would be nice for those of us with integrated video.

 

[gilga]

Why are there not any of these things around for $30 or so? It would be nice for those of us with integrated video.

30 dollars would be nice but Im sure it costs at least that much to make, package and ship them. Not mention the marketing costs and research and development. 30 would be awesome but its nothing more than an unrealistic dream.

 

[Demon_of_The_Fall]

Ummm... If you've got integrated graphics isn't the GPU going to be the major bottleneck on anything the PC does rather than whatever chip is doing physics calculations ?

 

[SuperGee]

Why are there not any of these things around for $30 or so? It would be nice for those of us with integrated video.

hardware Accelerated Physics is for enhancing the current optimum experience games can gave with more PhysX.
So putting a PPU is a barganbin game'ish rig Killes the enhancment there will not be much left of it wenn games run in minimum requierments settings.

You don't play games as it ment to be played. Because that needs a decent G-card.

I played Graw and CF with a 9600Pro that sucks with or without PPU.
I played it also with a X1800XL on 1800x1440 playable. Without one the games is great with one it's a bit better.
And on a 7800GS-plus wich replaced 9600Pro I have the same experience as on the X1800XL.

So $30 for PPU make sens wenn G80 cost les then $50 and X1950 cost $30
and a BMW M4 20'06 model cost $228,-

I think you have to wait wenn Hardware Accelerated Physics becomes mainstream and there are budged to high-end favors.

For now its a midrang high-end thing.

Who have the hardware to support such game enhancing thing.

 

[PureScoundrel]

I understand. Thanks for your input. I was under the assuption that it would be more quickly adapted if it started out small for lowend and scaled to mid and highend. So that everyone would benefit from the technology and it would become more mainstream.

 

[THRESHIN]

for it to become more mainstream, R&D needs to be done to make these. before that happens, i think it would be nice to have a few games to take advantage of physx. before any of that can happen ageia needs to make their investment back from their original R&D for the physx card. for that to happen, people need to buy it at its high price tag. see what i'm getting at? its expensive right now for the same reason blue ray dvd players are. its new technology and therefore will carry a high price tag until it catches on and is further developed.

 

[HOCP4ME]

I understand. Thanks for your input. I was under the assuption that it would be more quickly adapted if it started out small for lowend and scaled to mid and highend. So that everyone would benefit from the technology and it would become more mainstream.

That seems to be a good idea, but rarely works in practice. When a new technology is released, most people don't actually want to buy it. They just want to see what it can do at its best. Then, they will buy it when the price comes down and it becomes more mainstream. If you released the low-end first, it doesn't seem nearly as revolutionary, and there is no hype and no "that's awesome!" comments.

GPUs are done in the same way. The high-end is relesed first, then the mainstream cards, then the budget cards. NVIDIA, for example, released 7800GTX and 7800GT before 7600GT and GS.

 

[mikeblas]

Why are there not any of these things around for $30 or so? It would be nice for those of us with integrated video.

Well, let's see. Say you only need one engineer to implement one of these things, and he has to work a whole year. He's really smart, so he gets paid $120,000 a year. You'd need to sell 4,000 of them, then, to cover his salary alone -- not to mention all the other costs involved.

Of course, only one engineer can't do the job. You'd need a couple dozen LSI and fabrication engineers, and they make more than $120,000 a year. At least a handful of systems architects, and they make more than $120,000 a year, too. Plus driver writers, and they make more than $120,000 a year. Then, documentation, support, shipping, marketing, and the actual fabrication. You need a couple of people to do procurement, and then some people to do testing, and then a few more people to handle government compliance issues (for the FCC and RoHS, and so on). Manufacturing managers, shipping managers, and project manages.

Boy, that adds up fast, huh? Let's say you can do it with a team of 80 people over a year. But some of them have lesser salary requirements, so we'll figure $80,000 a year on average. That's a spend of $640,000, and we haven't even ordered any boards or parts yet, or paid for any of the per-unit costs. We'd have to sell 21,333 units to break even.

If you go through the rest of the back-of-the-envelope calculations, I think you can figure out why $30 is a silly price (at least initially).

 

[PureScoundrel]

That is all well and good but I think many of you are imagining the ppu out now for that price and that is not what I was thinking of. Let us say that a $30 part would be 1/4th the hardware of the current design. So there would be 4 different ppu adapters each with a different performance level. $30 for a wuarter of the featers and speed. $60 for half, $90 for 3/4th and whatever current price is for the full speed full featured adapter. They could modify the API to detect which level of ppu you have. All levels would still benefit the user. The low and midrange cards could be made cheaper requiring less parts. passively cooled, slower ram, etc.

 

[Demon_of_The_Fall]

That would just end up costing them far more, because all of the chips would cost the same amount due to how chip fabrication works, you'd just end up charging less for some of them. Doesn't seem to make sense to me

 

[mikeblas]

It doesn't make sense to me, either. You're not paying for speed; you're paying for engineering (and testing, and fabrication, and marketing, and distribution, and support).

 

[Chernobyl1]

I can think of a few reasons why they arent $30.

Its an unrealistic price point.
No competition (aside from NVidia but we are yet to see the graphics performance impact of using Physics on the G80).
Nothing (worth having) to run one on so no incentive for sales and no incentive for further competition.
Lack of sales prohibits a price drop, especially from $250 to $30, its just not gonna happen!

Maybe if AGEIA bring out a PCI-E part that works much better the price of the PCI card will drop substantially.
However, you may not want one then.

 

[Caffeinated]

Why are there not any of these things around for $30 or so? It would be nice for those of us with integrated video.

A person with integrated graphics and a free PCI slot to put a PPU could already better performance by spending that $30 toward a PCI video card. Granted, it won't be phenomenal, but I would have to imagine that something like an x1300 would have to be better than intel 900 or 950.

 

[pr0pensity]

If you have integrated video, you aren't going to be playing games, so why would you need a PPU? Physics can be done on the CPU, graphics can't, so integrated video is a necessity.

 

[pigwalk]

If a physics PPU card went for 30 bucks, or even under 100 people would actually buy them. Period. We all saw how far along it got at the 300 dollar price point, and even at 200 now it doesn't seem to be making much progress.

What's missing.... What's missing.... Oh yeah, Killer app anyone? But at 30 bucks, who even needs a killer app? People will just snap it up for another snazzy looking add-in card on the cheap.

 

[LuminaryJanitor]

Let's say you can do it with a team of 80 people over a year. But some of them have lesser salary requirements, so we'll figure $80,000 a year on average. That's a spend of $640,000, and we haven't even ordered any boards or parts yet, or paid for any of the per-unit costs. We'd have to sell 21,333 units to break even.

80*$80,000 = $6,400,000
=> 213,333 units

 

[pr0pensity]

If a physics PPU card went for 30 bucks, or even under 100 people would actually buy them. Period. We all saw how far along it got at the 300 dollar price point, and even at 200 now it doesn't seem to be making much progress.

What's missing.... What's missing.... Oh yeah, Killer app anyone? But at 30 bucks, who even needs a killer app? People will just snap it up for another snazzy looking add-in card on the cheap.

What would a $30 PPU do that a CPU can't?

 

[Chernobyl1]

to quote the PPU's design goal, Physics at speed.

 

[JaYp146]

Maybe Ageia could adapt something Physics-related that could run on a PCI-Express x1 bus? I still, to this day, haven't seen any popular add-on cards that use it ...

 

[Caffeinated]

If you have integrated video, you aren't going to be playing games, so why would you need a PPU? Physics can be done on the CPU, graphics can't, so integrated video is a necessity.

I'm not sure what you are trying to say here. Integrated video isn't a necessity, unless you don't have a discrete graphics card. I do, however completely agree with you in that if you have integrated video, you aren't going to be playing much of anything, and certainly nothing that needs a PPU.

If you have integrated graphics, it most likely sucks, and you'd be better off getting a PCI graphics card. How much is a physics accelerator going to help you if you have a shitty GPU? That's like getting a new sound card instead of relying on motherboard audio to offload the CPU. Will it help? Sure, but if your video card is awful, well guess how your gaming experience is going to be? Yeah, awful!

I don't see a PPU helping this situation much.

A perfect example of where an enterprising individual could make a fortune, and would be a hero to the gaming community is if someone was able to develop a PCMCIA video card for laptops, most of which are stuck with godawful video. A PCMCIA PPU won't help here, either, unfortunately.

Physics might make more explosions, but if your video card can't keep up with rendering in the first place, rendering more crap isn't going to help it. I wish it would, but that doesn't make it so.

Intel 900 and 950 are absolutely horrible graphics solutions. nVidia's 6100 is a lot better, as is ATI's solution - they are actually modern GPUs. I guess a PPU *might* help a user with a decent cpu and one of these, but I still think if you have a free PCI slot, and $50, you will get better performance if you go address the problem, which is the video card.

 

[Krypton]

Well, let's see. Say you only need one engineer to implement one of these things, and he has to work a whole year. He's really smart, so he gets paid $120,000 a year. You'd need to sell 4,000 of them, then, to cover his salary alone -- not to mention all the other costs involved.

Of course, only one engineer can't do the job. You'd need a couple dozen LSI and fabrication engineers, and they make more than $120,000 a year. At least a handful of systems architects, and they make more than $120,000 a year, too. Plus driver writers, and they make more than $120,000 a year. Then, documentation, support, shipping, marketing, and the actual fabrication. You need a couple of people to do procurement, and then some people to do testing, and then a few more people to handle government compliance issues (for the FCC and RoHS, and so on). Manufacturing managers, shipping managers, and project manages.

Boy, that adds up fast, huh? Let's say you can do it with a team of 80 people over a year. But some of them have lesser salary requirements, so we'll figure $80,000 a year on average. That's a spend of $640,000, and we haven't even ordered any boards or parts yet, or paid for any of the per-unit costs. We'd have to sell 21,333 units to break even.

If you go through the rest of the back-of-the-envelope calculations, I think you can figure out why $30 is a silly price (at least initially).

That is why they outsource and offshore most of the roles you are describing and the smart chip companies are fabless. So in reality, they have 25 people, a shitload of venture capital, and agreements with Fab companies. The Top Brass gets rich (legitimatly or not) even if the product is a flop.

And yes, you could sell a $30 PPU and still make money. The suckas that are early adopters pay for much of the initial R&D, and as design matures, competition enters, you will be dropping prices quick. Product cycle is the same for almost all technology products.

 

[SuperGee]

And yes, you could sell a $30 PPU and still make money. The suckas that are early adopters pay for much of the initial R&D, and as design matures, competition enters, you will be dropping prices quick. Product cycle is the same for almost all technology products.

First as a gamer you pay Asus or BFG for a PhysX CARD With happen to have a PPU on it. Asus and BFG pay's PPU for a small bulk Price.
It are not high volume sales. It so new it's in low volume wich means relative high price.
Then its ASP is nothing avarage about it's just one chip in produktion one model one produkt one price.

If a ASus P1 128MB cost $30,- Ageia gets a fraction of that. PCB and memory and a bitty margin takes the lot of it. $5 to 10 for ageia?

Maybe with $200 they catch $30 to 50
And there low production volume on old prodution prosces makes the cost also higher.

Large volume sales like onbard audio chips Realtec can live with that. as it's sales in large numbers where even a small margin make big profits.

But if sales get higher and it becoming more sucsesfull. CF:Revolution and UT2007 could work as a katalyst.
Price can drop because more volume sales. $30 I don't know they must sell a lot of those things.
For now $30 would be a dump price.

 

[T-Minus10]

If a physics PPU card went for 30 bucks, or even under 100 people would actually buy them.

The problem with the current batch of PPUs is *not* the price point. People are willing to spend exorbitant amounts of money on CPUs and video cards. The problem with the current batch of PPUs is that they're not useful at *any* price point. If ppus become useful, they have a chance of becoming common. If they become common, they have a chance of being introduced at lower price levels. Until they are useful, paying $30 for a ppu is still throwing out $30.