Ruby Video - HERE! and 2 STUNNING pics :D

[bloodbob]

It's lossy compression. By definition just taking a texture / normal map and compressing it will make it "worse". Same with DXTC. The benefit comes, just as with DXTC, from taking a higher resolution base texture (or using more normal maps in a scene to make the entire scene higher quality), with little to no performance penalty.

I didn't say worse did I said detrimental quality loss the simplest lossy compression of couse ist truncation. If XYZ company came out tomorrow said they can give 4:1 compression by simply truncating all the channels to 2 bit compents claiming it could be implemented with far few transistors then 3dc or DXTC do you think a quality comparision would be important?

I guess it's not obvious to some?

No its not obivous to the vast majority of users. Look at what extremetech is saying 3dc can do! This is why I'm the major point why I'm not happy.

The whole purpose of compressoin is to get the quality increase without sacrificing performance. Why are you arguing in circles?

Is that really true? I thought it was more to do with storage problems then anything atleast it is storage outside the 3d gaming industry. Also if this was true shouldn't it be standard to do all the benchmarking with compressed textures?

Yeah yeah and identical images the non PS3.0 is supposed to be http://www.pcper.com/image.php?aid=&img=farcry_before_bud2.jpg and damn PHPBB is a nazi on the image linking so it doesn't show up properly.

 

[Devourer]

Here you can find some more pictures as material to debate if you like.

 

[rhino]

Just for you guys, a pre-rendered still of Ruby.

For more info on the making of, check out RhinoFX. You can also watch the pre-rendered version of the demo. It makes a nice comparison. You can really see the spots where RT tech is close and the spots it still needs a little work.

 

[dr3amz]

impressive, and stunning how close ATI's version is to the original pre-render.

however, it kinda makes you want more!

could ATI have gone for even more detail in the tech demo?

but its hard to compare too far using a damn mov or mpeg!!! I NEED THIS CARD!!!!!! :?

I'm feeling very greedy, but i wanna see more lighting/shadows!

 

[Heathen]

Those lips are good enough to kiss. (Ruby's that is)

 

[thop]

rhino: Good work on Ruby, i especially like the facial expressions! I made some comparion shots of the pre-rendered and the real-time movie. You can see the differences, but i didn't really notice them so clearly when i saw the videos one after another. Very impressive if you ask me!

It's not 100% fair because the pre-rendered movie has more compression artifacts and motion blur, but anyway.

 

[flf]

Here's a side-by-side picture of Ruby and her sister Emerald. There always was a sibling rivalry between them, and that bitch Emerald just had to go work for that other company...

Which one is the corporate slut?

 

[Dave Baumann]

Just for you guys, a pre-rendered still of Ruby.

Mmmmm, swords... do I detect another episode of Ruby? Perhaps in about 6 months time...?

 

[Joe DeFuria]

Just for you guys, a pre-rendered still of Ruby.

Mmmmm, swords... do I detect another episode of Ruby? Perhaps in about 6 months time...?

Argh.....can't you at least let us have LAUNCH DAY without starting the next round of teasing?!

 

[Megadrive1988]

could someone post a direct link to what is for sure the pre-rendered movie?

 

[Nebuchadnezzar]

could someone post a direct link to what is for sure the pre-rendered movie?

That'd be awesome.

I like the pre-rendered Ruby MUCH more, I think ATI left too much detail out in the face, especially the eyes

Else it's simply awesome, ok maybe the reflective floor would have been also nice

 

[SlmDnk]

http://www.rhinofx.tv

Click on the Ruby pic and watch the movie. It's the pre-ren one..

EDIT:

Here is the direct link: http://www.rhinofx.tv/dc/dc_large.mov

 

[rhino]

well, the face has multiple uncompressed 2K textures on it in the pre-rendered version. That's just way too much data for a realtime card to be dealing with. So they had to compress all the textures a bit and it's most noticable in the face.
But the simple fact that it's so close and in realtime blows my mind. It's really easy to watch the ATi quicktime and see all the places where it doesn't quite measure up, but then you consider that our frames took many minutes to render and theirs take 1/30 of a second or less, it's puts in perspective.

 

[DemoCoder]

How much work was needed to convert the original source material into the real-time version? Is this using any XNA-like tools to make it easy to take artistic work and turn it into real-time material, or is it still a painstakingly laborious process to turn, say a 3dsMAX scene into real time. Has it been getting easier?

I'm wondering if the loop between visualization in a high-end tool, and turning that into reality is getting shorter and more automated.

Also, is there any layer compositing going on in those scenes?

 

[Megadrive1988]

I would imagine that real-time tech demos on ATI's next generation VPUs, beyond R420 - Radeon X800 (i.e. R5xx, R6xx, Xbox2 VPU, GCNext VPU)
will pretty much surpass the prerendered Ruby CG in many areas.

 

[DemoCoder]

I doubt it. Prerendered scenes can use vastly higher geometry and texture databases (without compression artifacts), plus real temporal AA and high degrees of spatial antialiasing. Jittered lighting/soft shadows also look better.

 

[JCLW]

Nice job to all at Rhino

Dude (does he have a name?) sounds a bit like Darth Vadar to me.

 

[rhino]

I can't comment specifically on how much work it was to turn the source into realtime because ATi did all that work themselves. I know that a lot of time was spent in writing shaders to approximate what we were doing. The geometry and the rig were much easier because we built those to their specifications.

There is some compositing going on in the rhino version.

It's quite possible that future realtime demos will surpass the pre-rendered scene. Even in the pre-rendered scene, we were still working within the limits of the ATi card. So the number of lights in the scene, the number of polys, the number of joints, the number of face targets, etc. are all much more limited than they would be in a true cg production.

But in terms of pre-rendered vs. realtime, as DemoCoder said, I think texture quality is going to lag far behind for quite a while. I don't even think the PCI-E bus is going to be able to handle massive numbers of 2k textures and the graphics cards themselves would need a gig or 2 gigs of ram. That's a couple generations away, at least. AA and sampling is a huge difference as well. I can crank the sampling up in mental ray, add a couple hours to a render, and get gorgeous AA. Trying to do that in real time is not quite as easy. And being able to do things like area lights, with ray-traced soft shadows . . .

 

[DemoCoder]

I just noticed that the pre-rendered version has real shadows, but the Ruby real-time movie appears to have "baked in" shadows.

 

[rhino]

Optico is the dude's name.