CS Source Benchmarks Cont. and Final Analysis
At a resolution of 1600X1200 things will become significantly more demanding on the graphics subsystem. Let's have a look.
There's one word for the lead Radeon X800 XT posts up in these two tests, "ouch". The ATi card boasts a 13-14% lead without AA or Aniso on and a whopping 47% lead with 4X AA and 8X AF enabled. It's amazing to see how polarized the Doom 3 performance results are versus our Counter Strike: Source Stress Test results. It's like the two game engines are at opposite ends of the spectrum when it comes to graphics processor architectures.
However there are a couple of things to remember with the CS Source Stress Test numbers here. First, this benchmark is just a graphics "stress test" and it wouldn't be completely fair to say that the frame rates we recorded here would scale the same way with a full Half Life 2 game driven benchmark run. There's definitely a strong correlation, but we need the full release game, as we have with Doom 3, to make the proper assessments in the metrics. The other obvious notable here is that the GeForce 6800 Ultra, at all settings and resolutions, was able to keep things in the "playable" range in both Doom 3 and the CS Source Stress Test, whereas the X800 XT PE fell closely into unplayable ranges in Doom 3. Again, however, ATi's lead in the Source engine, at least with this early take on it, puts them firmly ahead of NVIDIA at this point in time.
The obvious conclusion we can draw, when looking at the benchmark data for Doom 3 and the Counter Strike Source Stress Test, is that both of these game engines favor one graphics processor over the other. We do have a keen sense of the obvious here at HotHardware, don't we? However, let's look a bit deeper at the situation. Since the early days, John Carmack has been a faithful devotee to OpenGL as his standard for game engine development. The reason being, is that the folks at Id feel it gives them the ability to get closer to the "metal" of the modern GPU and take more control over these highly programmable graphics processors, when handling modern complex shader programs and the like. Not to mention, NVIDIA has always claimed that the NV40 was "built for Doom 3", so we guess that it all should come as no surprise.
Conversely, we also know that what has been biting NVIDIA in the backside for so long, is how well ATi has been able to execute on the delivery of a VPU that adheres so closely to a strict DX9 architecture. In short, since all the way back to the R3XX architecture, ATi's DX9 performance advantage has been their strong suit. As a result, most developers taking the DX9 code path, have found their game engines running better on ATi hardware, just like Gabe and the folks at Valve Software have with the Source Engine. We say "most" because Epic's Unreal Engine is a bit of an anomaly in that regard.
The real story with game engines and these new high end graphics processor architectures, is probably more about the one constant that never seems to change over time, resources and money. Game Devs with fewer resources and smaller budgets, probably have less of an ability to hand tweak their games for optimal performance on these highly programmable graphics cores. As a result, a more simplified strict DX9 path may be attractive and that seems to be where ATi cores shine. The select few Game Developers that have the resources to get in and optimize game engine code for NVIDIA's highly flexible NV40/NV45 architecture, like Id Software and Epic Games, who are also both very effective game engine marketing machines, will probably deliver engines that will favor NVIDIA's architecture. Although the raw fill rate and processing power of the NV40 isn't anymore more robust than the R420, it's strength lies in it's highly flexible and programmable shader engines.
So, we have a running bet here at HotHardware now. Which new leading edge game engine will be most successful in the coming months; Doom 3 or Source? We've heard rumblings that the barriers to entry (ie licensing fees) are significantly higher for Id's Doom 3 engine, versus Valve's ala cart menu offering for Source but the devil is always in the details, isn't it? So 6 - 9 months from now, when the market is playing Doom 3 and Source based titles, what's the landscape of benchmark numbers going to look like then? What impact will Epic's Unreal 3 engine have in late '05, when it is supposedly going to make it's grand appearance? And will this round of graphic cards even matter all that much, since most likely ATi and NVIDIA will both have their next gen cores queued up by then? You can make yourself dizzy trying to factor the possibilities actually. One thing is for sure, we'll be here at HotHardware trying to sort through it all for you in the months ahead.