I was playing around with some old FutureMark benchmarks on my new GeForce GTX460… trying to get them to work under Windows 7 x64… so I was googling for some information. I happened to stumble upon this old thread: http://forum.beyond3d.com/showthread.php?t=14008
I vaguely remembered participating in that thread, so I decided to read through it. I was amazed by reading this nearly 6-year old thread again. It was about 3DMark05 and FutureMark deciding to use nVidia’s DST/PCF extensions for shadowmapping. Beyond3D made a lot of drama about that, because it was biasing the results in favour of nVidia and all that…
Well, I argued that FutureMark’s choice was very logical, and gave a number of arguments, something like:
- Shadowmapping is the shadowing technique of the future (Pixar RenderMan also uses it). Various developers including big names such as Carmack and Sweeney have already announced support for shadowmapping (including nVidia’s extensions).
- A few games (eg Far Cry) are already using shadowmapping (including nVidia’s extensions), and more are likely to follow.
- DST/PCF is likely to become a standard feature in all hardware.
Nobody seemed to agree. Apparently a lot of people back then weren’t even convinced that shadowmapping is the way forward. They also did not think that nVidia’s extensions would ever be adopted as a standard. They all seemed to think I was mad.
However, today the situation is something like this:
- Shadowmapping is used by nearly every game. Shadow volumes have died out with the Doom 3 engine pretty much, and only outdated engines such as the Source engine still use alternative techniques such as projected shadows, with obvious rendering shortcomings.
- Sweeney’s UnrealEngine 3.0 is the most used game engine at the moment. And it uses shadowmapping, with DST/PCF extensions.
- Another important engine in driving shadowmapping is CryEngine 2.0.
- Depth-stencil surfaces became a standard feature for Direct3D 10. This means it is compulsory for all Direct3D 10 (and above) hardware to support it.
So basically everything I said was right… and everything that everyone else said was wrong. Also interesting how Dave Baumann is trying to make a big deal out of this whole DST/PCF thing (while probably already being aware of the D3D10 specs, and perhaps the ATi hardware already in development, with support for the feature)… Making a lot of drama about it, Beyond3D stepping out of the beta program and all that… And trying to push ATi’s 3Dc feature (which FutureMark and ATi devrel had jointly decided not to use as it would not speed up their current DXT-based implementation, and in their case would introduce more aliasing)… Has anyone ever heard anything of 3Dc again, by the way?
Nope, Beyond3D a good source of industry information? Not quite. Dave Baumann and Richard Huddy good for AMD’s PR? Not quite. I’m surprised AMD never replaced them… they make a lot of beginner mistakes. Surely AMD can do better.