Introduction

Besides DOOM 3, 2004’s other highly anticipated shooter is without a doubt Valve’s Half-Life 2. Late yesterday Valve released the beta version of Counter-Strike:Source and a video stress test for Half-Life 2, providing a glimpse of how today’s hardware will run with Half-Life 2’s source engine. Keep in mind that the video stress test probably shouldn’t be considered the definitive “Half-Life 2 benchmark”, as this isn’t the final game. In fact, if you recall our Half-Life 2 performance articles from last year, the video stress test is somewhat similar to the “techdemo” benchmark demo from that article. And since CS: Source is still in beta, benchmarks from it should probably be considered as a preview of things to come as well. But that didn’t stop us from testing!

We’ve decided to split our test results into two categories: one for the video stress test and a second group for the beta of Counter-Strike: Source. We’re starting off with the video stress test first to bring you as much information as quickly as possible; then we’ll report on performance with Counter-Strike: Source beta. Let’s get started!

X800 XT Platinum 4xAA/8xAF

GeForce 6800 Ultra 4xAA/8xAF

The video stress test

As we just mentioned, Valve’s video stress test is partially taken from the techdemo we used at ATI’s Shader Day event in Seattle. The demo splices together portions of Valve’s E3 2003 presentation such as the scene with the Predator character in front of the flames. Bump maps are practically everywhere, the walls of the caves are filled with them. Valve’s pixel shaded water is also used extensively in the video stress test. Like Valve’s original techdemo, the demo from the stress test is mainly meant to demonstrate the incredible effects that the engine is capable of, not real gameplay.

X800 XT Platinum 4xAA/8xAF

GeForce 6800 Ultra 4xAA/8xAF

X800 XT Platinum 4xAA/8xAF

GeForce 6800 Ultra 4xAA/8xAF

Because of this, it's best to think of the video stress test more like a synthetic benchmark, ala 3DMark 03, rather than an indicator of performance in Half-Life 2, much like Unreal Tournament 2003’s flyby tests were back in the days. We’ve provided a few screenshots, but keep in mind that we weren’t able to use the benchmarking functionality that we had last year, where we could see our performance on a frame-by-frame basis, as well as capture screenshots from any frame in the demo used for benchmarking.

X800 XT Platinum 4xAA/8xAF

GeForce 6800 Ultra 4xAA/8xAF

For our testing with the video stress test, all tests were run three times at a given resolution, with the highest score kept and the rest discarded. We turned up the detail settings to their maximums on the high-end cards, this means that model, texture, and shadow detail were all set to high, while water detail was set to reflect all objects. We should also mention that we noticed visual artifacts on both ATI and NVIDIA cards after extended use. Much like John Carmack mentioned that DOOM 3 pushes previously unused portions of hardware, the same may be the case for Half-Life 2. Also keep in mind that we’re dealing with beta software.

In addition, although it “appears” you can run 6xAA on NVIDIA cards, this is definitely not the case. Valve should probably consider taking this setting out when a GeForce card is detected.