[MUD-Dev2] [DESIGN] Ray traced environments

[Jon Leonard]

On Sat, Mar 17, 2007 at 05:17:35PM -0400, John Buehler wrote:
> Blue sky time.
> 
> Processors are fast.  They're getting more numerous per box.  We have
> graphics coprocessors.  Physics coprocessors are also out there.  It may be
> time to start talking about raytracing in games.
> 
> Here's one man's commentary on some current activity related to realtime
> raytracing.
> 
> http://www.pcper.com/article.php?aid=334
> 
> My questions to the list are:
> 
> 1. Are we still another 5-10 years away from rendering games with
> raytracing?

I doubt it'll ever be standard:  Raytracing is one technique of many,
and by itself is not all that interesting.  Effects like a wall
appearing brighter near a lamp (as compared to the darker appearance far
from the lamp) isn't modelable with classic raytracing.  (Radiosity is
the standard technique for that.)

> 2. Is there any real value in going with raytracing?  Are photo-realistic
> environments a niche feature for games?  Is this the point where the visuals
> just make the rest of the game look bad by comparison?

Improved visuals does seem to be a selling point, so I expect they'll
keep improving for a while.  Though if there gets to be a point where
essentially all players are running the games in a substandard visual
mode for speed, it may stop.

> 3. If raytracing becomes the standard for game rendering, how will it affect
> the industry?

Probably not that much.  The rendering portion of the software continues
to get a bit more complex and expensive, and so do the models, but
that's the trend anyway.

If the objective is actual photoreal rendering, then the proper
technique is to solve the integral equation for energy flow for the
specific scene.  (My preferred reference on this is Andrew S. Glassner's
_Principles of Digital Image Synthesis_).  That's not terribly
well-suited to current graphics-card design, so if that becomes more
common we'll see less emphasis on high-end graphics cards, or better
coupling between the card and the CPU/memory system.

> 4. Is raytraced gaming the killer application for all the
> multiprocessor/multicore stuff that's coming down the pike?

To be honest, I don't see any killer applications for the many-core
designs like the 80-core design Intel is talking about.  Code that
people are willing and able to rewrite for efficiency is already fairly
efficient (and saturating the memory bus), and code that isn't rewritten
won't do a terribly good job of using more than a few threads.

Jon Leonard