For a university talk I composed an overview on common rendering techniques in todays games (more or less the high-end ones). The material is a collection of stuff I have gathered in the last 5 years, although a lot is from recent siggraph courses.
Anyway it is not super technical as it is supposed to give rough outline on "everything"
I quickly translated from German to English, so mind if there are language bugs left. (e.g closely look at the thumbnail)
For a proper scientific talk the sources would have to be specified in more detail but anyway if you want to know where certain info came from you can always ask me here
http://www-e.uni-magdeburg.de/kubisch/Whats_cool_in_game_graphics_eng.pdf
Replies
The Wii is really the only system that needs an upgrade as far as visuals are concerned.
you used Turbo Esprit as a comparison to Need for Speed and Test Drive Unlimited ... except it wasn't a racing game, it was a free-roaming city-based drugs and guns game (with a play area as big or bigger than the one you quote for 2006)
i think the one you're after was probably Lotus Esprit Turbo Challenge
-caseyjones
It's also not as informative as I'd hoped, but it's a nice overview nevertheless.
if you want more detailed info on something specific just ask
I've added the important source links to the end of the slides now:
Siggraph
http://developer.amd.com/media/gpu_assets/Course_26_SIGGRAPH_2006.pdf
http://ati.amd.com/developer/SIGGRAPH07/Course28-Advanced_Real-Time_Rendering_in_3D_Graphics_and_Games_SIGGRAPH07.pdf
http://ati.amd.com/developer/SIGGRAPH08/Siggraph2008-Advances_in_Real-Time_Rendering_Course.pdf
GDC
http://ati.amd.com/developer/gdc/D3DTutorial08_FarCryAndDX9.pdf
http://ati.amd.com/developer/gdc/GDC2005_PracticalPRT.pdf
http://www.valvesoftware.com/publications/2008/GDC2008_StylizationWithAPurpose_TF2.pdf
http://www.valvesoftware.com/publications/2008/GDC2008_PostProcessingInTheOrangeBox.pdf
http://www.naughtydog.com/corporate/press/GDC%202008/UnchartedTechGDC2008.pdf
http://www.naughtydog.com/corporate/press/GDC%202008/PracticalSphericalHarmonicsBasedPRTmethods.pps
http://developer.download.nvidia.com/presentations/2008/GDC/GDC08_Ambient_Occlusion.pdf
http://developer.download.nvidia.com/presentations/2008/GDC/GDC08_SoftShadowMapping.pdf
Various
http://www.guerrilla-games.com/publications/dr_kz2_rsx_dev07.pdf
http://www.unrealtechnology.com/whats-new.php?ref=downloads
http://research.microsoft.com/~ppsloan/
http://www.punkuser.net/vsm/
http://www.punkuser.net/lvsm/
http://edusworld.org/ew/ficheros/2006/paginasWeb/making_of_sotc.html
http://game.watch.impress.co.jp/docs/20070131/3dlp.htm
basically the output for each pixel is a weighted sum of its neighbors and self.
imagine 3x3 weights & offsets with the middle pixel being the one we process.
different values yield different results, e.g if we do 1/9 as weight, we will get the classic "box car" filter (ie blur)
some more filter weights and results here
http://www.codeproject.com/KB/GDI-plus/csharpfilters.aspx
So in post processing you can use such filters to find "edges" say in depth buffer for basic silhouettes. You can blur an image for "glow" and so on...
Blurring a lot is same as "low-pass" filter in PS. To make this more efficient the image often isnt blurred at full res, but reduced in size along blurring, as we are not interested in sharp details anyway. That improves performance quite a bit, and doesnt affect quality much.
in rendering compute the texture offsets can be tricky, depending on the sampler type (there is rectangle textures and regular power of 2 ones) you can simply do +(1,0) for "next pixel right" or you would have to do +(1/width,0) ... anyway dont want to overcomplicate it hehe
We definitely had huge performance increases doing this. We use really small render targets and the results are very closely matched either way.