Single page Print

Counter-Strike: Source audio performance explored


There's more to performance than graphics
— 12:00 AM on August 25, 2004

OVER THE PAST WEEK we've seen a deluge of Counter-Strike: Source performance articles. For the most part, coverage has focused on graphics cards, but there's more to the Counter-Strike: Source performance story than pixel-pushing power. In addition to higher polygon counts, bigger textures, and fancy shader effects, CS: Source also features a surround sound audio engine that wraps in-game audio around five speakers.

Unlike the DOOM 3 engine, which currently processes positional audio on the CPU, the Source engine appears to take advantage of hardware 3D audio acceleration. In theory, leveraging sound card resources to crunch positional audio calculations should free up CPU resources and improve overall performance, but is there really much of a performance difference between audio implementations? We've rounded up nine different sound cards to find out.

Before we begin...
I'd like to thank TR reader Timm Stokke for getting the ball rolling. Timm pointed us to a couple of threads over at Shacknews suggesting that CS: Source runs at much higher frame rates when sounds are disabled. Of course, no one wants to run a game with no sound, but it raised the question: do some sound cards perform better in CS: Source than others?


Nine audio cards compared

To answer that question, I combed the Benchmarking Sweatshop and came up with nine sound cards to test. I've also thrown in some integrated motherboard audio—the VT8237/VT1616 south bridge/codec combo found on Abit's KV8-MAX3—for good measure. (Note to SoundStorm fanboys: Yes, I'm aware I didn't test with your beloved SoundStorm. Because SoundStorm requires a dated Athlon XP/nForce2 platform, I left it out of this comparison. As far as current Athlon 64 platforms are concerned, SoundStorm is dead. Get over it.)

When testing 3D audio performance, it's important to make a distinction between audio implementations that offer true hardware acceleration and those that emulate hardware acceleration in software. Emulating 3D acceleration in software dumps positional audio calculations back onto the CPU, so cards that offer true hardware acceleration should consume fewer CPU cycles than cards that fake hardware acceleration in software. Here's how the audio implementations we tested stack up in terms of hardware acceleration:

Hardware audio acceleration Software audio acceleration
Creative SoundBlaster Live!
Creative SoundBlaster Audigy
Creative SoundBlaster Audigy2
Creative SoundBlaster Audigy2 ZS
Hercules Gamesurround Fortissimo III 7.1
Hercules Muse 5.1
M-Audio Revolution 7.1
Mad Dog Multimedia Entertainer 7.1
Philips Ultimate Edge
VIA VT8237/VT1616 integrated

As you can see, Creative and Hercules are all over hardware audio acceleration. The M-Audio, Mad Dog Multimedia, and Philips cards all use variations of VIA's Envy24 audio controller, which lacks hardware acceleration for 3D audio. VIA's VT8237/VT1616 south bridge/codec combo also lacks true hardware audio acceleration, as do most integrated motherboard audio implementations.


A screenshot from the CS: Source demo we used for testing