Today during lunch I took some time to play the Counter-Strike: Source beta. I've never really been a big CS fan, but I bought Condition Zero a while back to test how Steam handles game purchases. That entered me into the CS:S beta. I've found the traditional Counter-Strike gameplay much more compelling with the level of realism provided by the Source engine. The object physics really are amazing to behold.
After watching all the preview videos for Half-Life 2, I was anticipating the character models more than anything. Specifically, the faces. In Half-Life 2, facial animations and appearance achieve a stunning level or realism for a computer game that is markedly better than what Hollywood CGI has produced just a year or two ago. Source seems to outshine its contemporaries (Doom 3 and Far Cry) in all things human model related. This feature does not seem to be present in the counter-Strike iteration of the Source engine.
I can understand why the folks at Valve would choose to do this. I'd imagine that the self-shadowing luminosity calculations that make up eyes in HL2 would be very taxing when rendered on 40 fast moving, player-controlled characters. In fact, all of CS:S seems to be geared toward faster performance at the cost of some attractiveness. The textures do not seem to be as painstakingly built as those in HL2. There is some bump-mapping on various objects, but normal maps seem few and far-between.
The video stress test included with the game seems to offer much better looks than CS:S itself. It's more representative of the footage we've all seen of HL2, and framerates are in fact lower than actual gameplay in CS.
As far as I can tell, unlike Far Cry and Doom 3, Source does not offer a per-pixel lighting system. This would explain the higher than Doom 3 performance I get on my video card (that and Doom 3 is an OpenGL app, while Source is DirectX and I have an ATi card). This does cause the game to appear a bit antiquated in low light conditions. Far Cry and Doom 3 are really impressive in terms of the atmosphere they create with lighting.
In fact, I was watching John Carmack's Quakecon keynote and he talked about where he's going next with regards to rendering technology. He mentioned there were some flaws and imperfections in the Doom 3 technology, but he thinks real-time in-game graphics could match and even surpass pre-rendered CGI in the next few years, at least on systems running multiple PCI Express cards in an SLI configuration. At that point, it may make sense for movie studios to do their CGI work in real time.
Man, that would make for some great looking games.