ATI - PS3 is Unrefined

Bad_Boy said:
Yeah, I beleive EA used scans for fight night also. (which i beleive would be a different studo from the EA MOH team) I think the guy from MTV went to interview kudo, and kudo got his head scanned and eventually put him in the game. It seemed like a pretty fast and cost effective process to just do that for free? :p

I'll try to find a link, but I believe it was on "MTV overdrive". I hope somebody else knows what im refering to incase I cant find it.

I will like to see that link too. I've never seen that before.
 
mckmas8808 said:
I will like to see that link too. I've never seen that before.

he took a shot of the Overdrive Browser
http://www.beyond3d.com/forum/showpost.php?p=634487&postcount=24

unfortunately theres no direct link
try http://www.mtv.com/overdrive/ (use IE)
but for some reason i cant find the thumbnail anymore, they may have removed it :(

edit:
click the G-Hole, theres a thumbnail with the host and Kudo together so that might be it.

edit2:
yep it was
fightnightmtv2tc.jpg


Hope that helped, sorry about the wait :D
 
Last edited by a moderator:
Bad_Boy said:
he took a shot of the Overdrive Browser
http://www.beyond3d.com/forum/showpost.php?p=634487&postcount=24

unfortunately theres no direct link
try http://www.mtv.com/overdrive/ (use IE)
but for some reason i cant find the thumbnail anymore, they may have removed it :(

edit:
click the G-Hole, theres a thumbnail with the host and Kudo together so that might be it.

edit2:
yep it was

Hope that helped, sorry about the wait :D

Yep that helped. I wonder how long it took them to map his face to a character in the game? Pretty cool non the less.
 
mckmas8808 said:
Yep that helped. I wonder how long it took them to map his face to a character in the game? Pretty cool non the less.
Yeah, Id like to know that also. It couldnt have taken all that long though, considering he was wearing the same clothes that he walked in and the same clothes he was playing himself in. So under a day for sure, my guess a couple hours max? Not talking just about the face scan, but you notice he has a custom skin also, and custom "MTV" shorts including his name "Blair", and of course a little longer for the dev to code him in.

But anyways, seems very effective/cheap and also very fast.
 
Scanning someone is not a simple process...
First of all it takes a really expensive equipment to get high quality data quickly, and even that won't guarantee that the human you scan can keep from moving during the process. Maquettes are obviously less of a problem here. I don't know how much such a scanner costs, but it's probably somewhere between $10.000 - $1.000.000 or so. Most VFX studios don't have one, they go to a specialized company to handle this like www.XYZRGB.com. I'm not sure about their rates either, but it certainly isn't cheap.

The scanner delivers point cloud data; that is, millions of tiny polygons, with an occasional hole thrown in here and there for some fun. This data can't be used as it is; the common workflow is to use it as a 3D guide to build a much simpler and animation friendly model on top of it, then use software to generate displacement and/or normal maps to store the difference between the scanned mesh and the rebuilt mesh.

The rebuilding process can take various amounts of time; back in `96, ILM needed some 3 modelers and 6 months (!) to build Draco for the movie 'Dragonheart' because of the complex scales and horns. A human face on the other hand should not take more than 1-3 days for a single modeler. In case anyone is interested in details about the process, here's Factor5's GDC presentation on using scanned data (and Zbrush as an alternative): http://www.langsuyar.com/gdc/.

So scanning will certainly make the modeler's life easier, and the result a lot more lifelike - but it won't be enough to remove the manual work. And building a full ingame character will still require textures (can use scanned or photographed data as a source similar to the modeling process), shaders, and a lot of rigging work to animate it (anim controls, binding to a skeleton etc.).
 
How many transistors is there in the "final" DD2 revision of Cell compared to the first one?
DD1 was 234million/221mm if i remember right?
 
Laa-Yosh said:
It isn't, see my previous post...
My point is that many studio's seem to have the hardware needed. I'm sure if EA has one, Sony probably does. (Sony:Molina) And after looking at the video it couldnt have taken them more than a day, start to finish. Animation is probably another thing which was probably already premade+precoded before the Host came to the studio, but If all that was done before hand it wouldnt take long I suppose.

Whats interesting is if its so expensive, why would EA do all that work you say it takes for free? I guess free/cheap marketing would be one answer, or on the other hand it could actually less expensive than we think it is? (after they have purchased the needed hardware of course)

On the other hand your right,
I'm sure for smaller development companies, they will be forced to do it the manual way due to costs and work hours, but for Major studios I dont think it would be all that much of a problem If they have the capable hardware and man power.
 
Great thread!

Seems if Sony chooses their own part, they are damned, and if they choose an off the shelf GPU from a major GPU manufacturer they are still damned. If RSX is nothing more than a G70 with the faster CPU<->GPU bus I will be more than happy. A 550 MHz G70 is an incredibly powerful monster.

I wonder how many Japanese games does ATI produce? :D
 
overclocked said:
The one thats going in the PS3? Had NO idea i thought the final one was DD2.

No idea if it will be in the PS3 or not, I think it will, but there just isn't enough info and what exactly has changed between DD2.1 and DD3.1 ... at least I don't know anything, but I'm pretty sure someone here knows ... if he's allowed to tell us about it is another question.

Fredi
 
overclocked said:
The one thats going in the PS3? Had NO idea i thought the final one was DD2.

Yeah, probably. I think Barry Minor of IBM said a little while ago that the changes made from 2 to 3 were really just yield-related.
 
360 GPU claims to have "greater efficiency", and "unified shaders", as if those two phrases would be enough to settle the debate, but I wonder how many people would choose a machine with 48 * 10 horsepower motors versus (24 * 22 horsepower motors (pixel pipelines) + 8 * 10 horsepower motors (vertex pipelines)). One could argue all day the 10 horsepower motors are more efficient, and offer better load balancing, but in the end the G70 can still be more powerful. Sure it has less execution units, but each pixel execution unit looks quite a bit more powerful than each unified shader on the 360 GPU.

Which is more poweful, I don't really know, but I would certainly not place any bets against the G70, especially with it's incredible bandwidith with CELL, and no question in my mind CELL is on the order of 2 to 3 times more power than the Xbox 360 CPU when it comes to math throughput, and that's critical in video games, and especially graphics.

360 first generation games are full of aliasing, shimmering, and next to no anistropic filtering, and all we hear are excuses. So much for being superior.
 
Fredi,Titanio..

Yeah i mean the "big" change was to the DD2 right with the bigger PPE and all.
Witch brings me back to how many transistors and the diesize of DD2 then, 250million(?).
 
Laa-Yosh said:
Scanning someone is not a simple process...
First of all it takes a really expensive equipment to get high quality data quickly, and even that won't guarantee that the human you scan can keep from moving during the process. Maquettes are obviously less of a problem here. I don't know how much such a scanner costs, but it's probably somewhere between $10.000 - $1.000.000 or so. Most VFX studios don't have one, they go to a specialized company to handle this like www.XYZRGB.com. I'm not sure about their rates either, but it certainly isn't cheap.

Maybe not right at this moment. But is that because it's inherently expensive to do, or just because right now it's a specialist process with a small number of rich customers?

Many technologies required for this kind of work are rapidly becoming more widely available or cheaper because of increased demand. High resolution digital video, depth-imaging... and an awful lot can be done in software with a reasonable amount of source data. Software needs processing power and data storage, but both those things are rapidly becoming plentiful at affordable prices.

In years gone by a lot of broadcast video work was prohibitively expensive not because of the hardware itself, but because of the amount of storage required. These days you can throw terabytes of drive in a desktop PC without a second thought.

So I think the costs are going to come down dramatically on this stuff, just as soon as there's a bigger demand for it.

I also think you're going to see the asset creation industry (which already exists) really explode around the requirements for next-gen games and also cg for mainstream work (TV shows rather than big-budget movies). Bigger companies might do everything in-house, but smaller companies are going to go to a specialist and say "we need some characters" or "we need some car models".

Maybe it'll be like hiring an actor - except you'll hire a particular company for the type of virtual-character they have, and they'll customise it for you. Armies of pre-scanned virtual characters ready to be dropped onto standard rigs and licensed for your exclusive use. You needn't even use real actors - just scan random faces and combine them into believable hybrid characters. And/or apply some basic transformations to alter a character to be fatter, thinner or whatever. That way you can actually get many more characters than you've bothered scanning in the first place, again slashing the costs.

Think of it like sperm donation :) Offer members of the public a few quid in return for being scanned. People will be queueing up...

Almost all the talk about the bigger budgets and teams required for next-gen work is about the art. Out-sourcing is certainly one way in which that can be solved without having each development house add a zero to its art department headcount.
 
Edge said:
360 first generation games are full of aliasing, shimmering, and next to no anistropic filtering, and all we hear are excuses. So much for being superior.

Sadly the bottom isn't even reached if you've seen the latest namco pictures (Love Football and Frame City Killer).

Fredi
 
Bad_Boy said:
My point is that many studio's seem to have the hardware needed. I'm sure if EA has one, Sony probably does. (Sony:Molina)

EA has very large studios with hundreds of people working on many projects running in paralel. Sony is probably quite similar (although Imageworks is a VFX studio and is probably quite indipendent of the game dev studios Sony has).
Still, buying a scanner like XYZRGB's is most likely far too expensive for them, so if they have an inhouse one, that's probably not as detailed. Also note that XYZ's site opens with a Fight Night 3 screenshot:
http://www.xyzrgb.com/pic2.jpg

And after looking at the video it couldnt have taken them more than a day, start to finish.
Animation is probably another thing which was probably already premade+precoded before the Host came to the studio, but If all that was done before hand it wouldnt take long I suppose.

I wouldn't make any guesses from such a video; it's pretty obvious that it shows a highly simplified process. Have you seen any actual scanning in the clip, for example? All they show is a few short cuts of the photo shoot... They could've used a lot of pre-built assets modified a bit to look like the host, or they could've done most of the work prior to the actual visit. In my experience, all the work for such a character takes at least a few weeks...
 
MrWibble said:

I've been talking about the increasing need for outsourcing in nextgen games for a few years already :) Ever since I've seen the first images of Doom3, to be precise.

Costs will obviously move down, technology will advance, tools will improve and so on. Still, there are stages in the workflow where you can't replace an artist. Also, for a sports game or such where you need the actual celebrity, you can't resort to 'hiring' a pre-built 'virtual actor', you'll have to get that person to the scanner and so on.

I'll try to get a friend who worked at Weta on Kong to share some info on how much time and work went into their digital doubles... he's not there anymore so he might be able to talk ;)
 
Back
Top