Oh man...just think where 3D gaming is going to go after this

[Heavy_Equipment]

Multi chip and multi card solutions are also coming, meaning that you will be able to fit more frame rendering power in a single tower case than Pixar's entire rendering farm. Next year.

I had originally estimated that it would take a few years for the tools to mature to the point that they would actually be used in production work, but some companies have done some very smart things, and I expect that production frames will be rendered on PC graphics cards before the end of next year. It will be for TV first, but it will show up in film eventually.

John Carmack

When Toy Story, and Ice Age type movies are on the brink of being produced by a single PC's hardware, (as opposed to the supercomputer->software approach being used today) the gaming market can only take the ball and run with it.

Were on the verge of seeing 3D cards that will make what we are looking at today, look like vga.

[MeltMan]

woah

[bradford010]

lmao!

....more frame rendering power in a single tower case than Pixar's entire rendering farm. Next year.

IMHO, he's full of it. I happen to work with 3D and know the power of a top end render farm is far beyond anything you could hope to fit in a single machine. And the supercomputer approach was first abandoned years ago, and that migration is almost complete with only the stragglers remaining on the likes of SGI rather than move to x86 boxes running linux.

[Techguy]

bradford010, but then again, this IS John Carmack annd he seems to of learned a thing or two about 3D.

I'd immagine that he also has a heads up as to what's comming down the line so he can develop doom3 and Id can make quake4 with those cards in mind. I believe he had radeon 9700's to play with, so I'd immagine he had acces to info on some other new cards.

Also, isn't running a bunch of PC's together sorta the whole point of a super computer?

[bradford010]

Yeah, I mis-read the supercomputer bit as meaning moving from the likes of SGi to x86 boxes.
Whereas now the definition of supercomputing is clustering for greater overall power, as you said.


After reading a little more of what he's said (as various other comments have popped up) I've come to the conclusion that Carmack's a fairly crafty fella.
I think the technology he's talking about is something ATI just used in some marketing, in a similarly misleading way. The hardware can be used to render to screen in realtime a scene an artist is working on. Albeit in a severely cut down form. But he's wording some of his statements to imply that we're going to have the computing power of NASA sat in our AGP slots come next year.
Then what confuses me is statements like

you will be able to fit more frame rendering power in a single tower case than Pixar's entire rendering farm. Next year.

If you just stop and think about the feasibility of what he's saying.

That in under a year, the rendering power of a room (and it's a big room mind ) full of dual PIIIs/PIVs with several GB of RAM apiece, can be leveraged into a card small enough to fit in an AGP slot.

Forgive me if I'm being skeptical, but I just don't see that sort of leap happening in under a year.

I'm not suggesting he's wrong. I'd be more likely to go with the opinion that he's talking bull, and he knows it.

[jtroutma]

In a year...... I am skeptical as well. However in say 3-5 years....... very probable (maybe in only 2 years )

[bigben2k]

I've also done some render work, 3D animation type, and I 'll believe it when I see it.

Playing an MPEG movie and Rendering are two seperate things.

[bradford010]

Just realised something else he may be talking about as well.

The cards may be able to render a single subject frame photorealistically. Which of course requires a lot less power than, say one of the huge crowd scenes in LOTR, including crowd simulation, dynamics etc etc.

[Heavy_Equipment]

I'm not implying I know a whole lot about 3D rendering, but I do know the trickle down effect.

Whether it's "Next year" or not, I can't say either way...but when it gets here, the 3D gaming market will definately be pretty hot.

In all honesty, I'm sure the gaming technology for example, ATI or Nvidia or 3DLabs is developing, is feeding the professional market, as well.

As for Carmack having cards to play with...he was raving about the R300 at the last QuakeCon, back in February, mind you, from the engineering specs. I read somewhere he was consulted through the development of the card, so I'm sure he had one in his hands long before any others.

The cards he mentioned being on the horizon are already here to a degree...look at the 3DLabs Oxygen GVX420, while not as fast as what Carmack is alluding to, these cards are monsters.
A full length AGP card (18" or there abouts), with a PCI daughter card.

Think about it this way, a few years ago, the 3Dfx Voodoo was IT. Look at the strides the video card industry have taken in just the last couple years. We are getting excited now about the early reports on the R300, and the speculation on the NV30 is heating up. These cards are going to be old news in the spring, and the next batch is going to be faster. This time next year, these cards will be two generations old, and GPU's will be approaching 1GHz.
I don't think John Carmack is too far from the truth.

[bradford010]

What we have to do is just step back and take in the whole picture.

We're being told that the rendering power of multi-million dollar computer setups, that take up entire (larger) rooms is going to be made available through a card that fits in a AGP slot, at a consumer price point.

Within 6 months.





The Quadro line up from Nvidia (their current pro-graphics card line) could be considered at a consumer price point (albeit, this is a stretch, and we're talking about seriously wealthy consumers) and they come nowhere near this level of rendering power.

Even the current best offering from 3DLabs, the Wildcat 6210 does not come anywhere near close to the rendering power of a render farm. And for the price you'd pay for that, you could buy around two top end dual athlon systems.

Weta, the SFX company working on LOTR, just made another investment in expanding their render farm. Dual Xeons with 4GB RAM apiece. Joining the existing PIIIs they now have in excess of 400 machines in that render farm.
If the only number we consider here is the cost, would any company with any sense purchase that much hardware if it was possible that the same thing would be available at a fraction of a percent of the cost within a matter of months?

[Heavy_Equipment]

Firstly, no where in that quote does John Carmack say January..."before the end of next year" is roughly 17 months away.

Secondly, I'm sure the time frame and over all scope of the 'rendering farm in a box' analogy was meant as an exclamation point on the future of the industry...he was being emphatical.

If you just stop and think about the feasibility of what he's saying.

I agree with you.

Although, you could render Lord of The Rings on a single PC right now if you want to get right down to it...
It would just bog production down, and set the completion back by oh...10 years.

A half hour 'Toy Story' like TV show would be a lot less taxing than a 90 minute LOTR quality production.(I haven't seen Pixar do anything near that caliber anyway )

I wasn't trying to start a debate, with that quote, I was just using it in a 'makes-you-wonder' sense.

The future is bright, and I know we can all agree on that.

[bradford010]

Originally posted by Heavy_Equipment
I'm not looking for a debate, when it comes to this subject, in a battle of wits, I'd be unarmed,

ROFL

As for the time frame, AFAIK these quotes from him are starting to date a little, so I just adjusted.

But you're right, the future does look very rosy. I for one am very keen to look at what the next gen of cards will bring. Especially in pro graphics arenas.

[Heavy_Equipment]

oops, you replied while I was mid-edit.

That's gonna confuse everybody...

I did say what Bradford quoted about the debate, but I changed it after I re-read it... it's too late for that

[bradford010]

LMAO, yeah but if I change my quote, it's not nearly as funny!

[Heavy_Equipment]

I can humbly admit when I've bit of more than I can chew.

[airspirit]

I'm going to throw my .02 into this. They are designing the 9700s along the same lines as the Hammer. These are meant to be scalable on a single card (like some of those sick quad Voodoos). Combine this with AGPx8 and you have a disgusting amount of GPU power. My understanding is that, like the hammer, each chip has its own RAM. When these things start using next-gen RAM and we start seeing 4 chips on one board with, say 256MB DDR2 to each chip all pumping through 8x, that's a hell of a lot of power. It goes without saying that there are already rumors floating about the next generation after the 9700 based cards. Alot of rumor and speculation, but its all we have to go off of now. I doubt that the power will be like he says (what board has the kind of bandwidth to handle that? The Hammer based boards will approach it, though), but next year we are going to probably see a change in graphics card power that will dwarf anything we've seen in past years.

[bradford010]

Yeah.

oh, another thing I forgot to mention. Films are not rendered at anything near as small a scale as monitor resolutions. I can't remember what it is off the top of my head, but it's large enough to watch on a 30' screen without looking pixelated.

Let's just say 72dpi. 30 x 12 x 72 = 25920 px wide.

[ksw]

my 2 cents
there is something in those rumors i believe, if you look closely on the design of the radeon 9700, it looks as if it would be multi processor capable, but in need of a coordinator, a Central Processing Unit, meaning that for more than one GU, u need a coordinatuiing CPU, wich equals approx totaly, 2gpu + cpu, 100-150w power consumption. thats a heck of a lot of power.
do you agree?

[bradford010]

Hell of a lot of power, yes.

Enough to match 300+ PIIIs with 4GB RAM each, I dunno.