Sign in to follow this  

"Fast new lighting technique!"

This topic is 3452 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Recommended Posts

I've heard from a couple of different places now allusions being made to some spectacular new lighting method that supposedly: * Makes everything brighter and more detailed * Is lots faster, and can therefore render more/better with less beefy hardware The two sources I can think of that specifically mention this are Tim Sweeny when showing of the GoW2 engine updates, and again recently reading an article in PC gamer about Crysis Warfare. Both sources (and several others I've heard but don't remember) mention the above benefits, but neither give an explicit name to the technique. So now I'm curious as to what it is! My best guess is that they're talking about Screen space Ambient Occlusion (I'll save you the google), which can definately give you brighter scenes and "more" detail (better lighting anyway). If this is the case, though, I'm confused as to how it would deliver better performance. The way I see it is that anything that requires an extra pass over the scene is going to bog you down, not to mention multiple texture lookups instead of a constant term. The only way I can see this speeding things up is if you use it to reduce the number of "real" lights in your scene, since SSAO tends to produce brighter scenes anyway. So does anyone know more about it? Am I in the ballpark or not even close?

Share this post


Link to post
Share on other sites
Ambient Occlusion doesn't necesarily produce a brighter scene - it only modifies the ambient term, but the overall brightness of the scene doesn't need to be increased by using it.

I was just reading in PC Gamer about Crysis Warhead, and the term that they used was 'Global Ambient Lighting'. They also mentioned that machines that can run Crysis now will be able to use this technique without a framerate hit. This sounds suspicously like a modified SSAO that would approximate GI somehow. I remember reading the Siggraph paper from Crytek that said that they experimented with using SSAO on other lighting terms, but ultimately abandoned them because they didn't make sense.

SSAO uses the depth information only to modify the amount of visibility in the scene. I wonder if they are using the same type of technique to sum up the local scene colors to simulate global lighting. It's just a hunch, but the fact that they say that Crysis is the same performance bench mark makes me think that the algorithms used in Crysis are very similar...

Has anyone tried this type of color bleeding based on a screen space sampling kernel? Could be interesting to play around with and see what comes out...

Share this post


Link to post
Share on other sites
Quote:
Original post by Jason Z
Has anyone tried this type of color bleeding based on a screen space sampling kernel? Could be interesting to play around with and see what comes out...


wolfgang recently had a blog post on this, check it out.

Share this post


Link to post
Share on other sites
No, I don't think so. ATI's new hotness is cool, but it's very much about the hardware. Whatever this method is, it runs at the very least on XBox360 hardware, so I know it has to be a software solution. (Unless GoW2 plans on shipping a hardware upgrade in every box :)

Good guess, though, and I have been watching developments on the ATI stuff closely as of late.

Share this post


Link to post
Share on other sites
Based on what I saw from the Gears of War 2 talks and descriptions, I have a feeling you might be talking about a local irradiance encoding in SH coefficients used for ambient lighting. I implemented a similar thing for Tomb Raider Underworld as well in the form of an arbitrary irradiance volume, and I can see how you can argue that it makes things brighter and improves the look of the ambient lighting.

However, I wonder about the "lots faster" part. It doesn't really make that much difference, performance-wise, and although significantly higher quality, it's certainly not faster than a constant ambient term. Perhaps if you were comparing it to various other ways of faking indirect illumination, it's certainly quite fast, but you still take a small hit per frame for it.

It's also a technique that hits upon several limitations unless you combine it with several other additional tricks, some of which are clever and some of which are just filthy dirty hacks. Perhaps if we had about double the constant registers on the consoles, things might be a little smoother, but we don't and they're not.

Share this post


Link to post
Share on other sites
Quote:
Original post by Jason Z
SSAO uses the depth information only to modify the amount of visibility in the scene. I wonder if they are using the same type of technique to sum up the local scene colors to simulate global lighting. It's just a hunch, but the fact that they say that Crysis is the same performance bench mark makes me think that the algorithms used in Crysis are very similar...
Has anyone tried this type of color bleeding based on a screen space sampling kernel? Could be interesting to play around with and see what comes out...

Seems me,that you can make global lighting, using:
-normals & depth map in camera space
-colors from previous frame copy,for example
-colors from additional,rear view map
This map can be rendered with wide view angle,only for big shapes & sky behind,and not every frame(or distributed between frames / arithmetical mean )
I think also,that this map will pay for itself in case of using for coarse environment reflections too( bearing in mind clear boundary absence between global lighting ,specular and reflection)
I thought about it when doing realtime radiocity,now I remember only,that method of combined normals & depth seemed to me quite simple :)

[Edited by - Krokhin on June 29, 2008 1:17:41 PM]

Share this post


Link to post
Share on other sites
Quote:
Original post by Krokhin
Quote:
Original post by Jason Z
SSAO uses the depth information only to modify the amount of visibility in the scene. I wonder if they are using the same type of technique to sum up the local scene colors to simulate global lighting. It's just a hunch, but the fact that they say that Crysis is the same performance bench mark makes me think that the algorithms used in Crysis are very similar...
Has anyone tried this type of color bleeding based on a screen space sampling kernel? Could be interesting to play around with and see what comes out...

Seems me,that you can make global lighting, using:
-normals & depth map in camera space
-colors from previous frame copy,for example
-colors from additional,rear view map
This map can be rendered with wide view angle,only for big shapes & sky behind,and not every frame(or distributed between frames / arithmetical mean )
I think also,that this map will pay for itself in case of using for coarse environment reflections too( bearing in mind clear boundary absence between global lighting ,specular and reflection)
I thought about it when doing realtime radiocity,now I remember only,that method of combined normals & depth seemed to me quite simple :)
It also seems that this would be useable with a cube map or paraboloid maps around the viewer - then you have coverage for all scene directions. I would be interested to see some of the results for something like this. The screen shots that I saw from Crysis Warhead didn't look obviously screen space, but it was clear that the same lighting system wasn't used in the comparisons...

Share this post


Link to post
Share on other sites
Quote:
Original post by Jason Z
It also seems that this would be useable with a cube map or paraboloid maps around the viewer - then you have coverage for all scene directions. I would be interested to see some of the results for something like this. The screen shots that I saw from Crysis Warhead didn't look obviously screen space, but it was clear that the same lighting system wasn't used in the comparisons...

Don't look around-just try to do it :)
I work with clouds now,but if you decide to try-I'll add something
what I remember.

Share this post


Link to post
Share on other sites
Quote:
Original post by Krokhin
Quote:
Original post by Jason Z
It also seems that this would be useable with a cube map or paraboloid maps around the viewer - then you have coverage for all scene directions. I would be interested to see some of the results for something like this. The screen shots that I saw from Crysis Warhead didn't look obviously screen space, but it was clear that the same lighting system wasn't used in the comparisons...

Don't look around-just try to do it :)
I work with clouds now,but if you decide to try-I'll add something
what I remember.
[grin] Time is the limiting factor, my friend. If I could only find a few more minutes to add to each day...

Share this post


Link to post
Share on other sites
Quote:
Original post by Jason Z
Time is the limiting factor, my friend. If I could only find a few more minutes to add to each day...

I'm very busy also.
ps I wish you to prevent all problems :)

[Edited by - Krokhin on July 2, 2008 10:08:43 AM]

Share this post


Link to post
Share on other sites

This topic is 3452 days old which is more than the 365 day threshold we allow for new replies. Please post a new topic.

If you intended to correct an error in the post then please contact us.

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

Sign in to follow this