SUMMARY: Last month during QuakeCon it became clearly apparent that computer audio has become somewhat of a forgotten component in the computer industry when talking to gamers and listening to companies at the gaming event. We’ll present some benchmark numbers of five different sound solutions as well as provide commentary along the way on our thoughts of computer audio solutions and what should be done to improve things using nVidia's SoundStorm APU as an example.
Processor(s): AMD Athlon XP-M 2600+ @ 3900+ (2.6GHz) Motherboard(s): ABIT NF7-S Memory(s): 2x 512MB Buffalo 2-2-2-5 BH-5 @ 221MHz Video Card(s): Gigabyte Radeon X800 XT PE 256MB (Supplied by Gigabyte) Hard Disk(s): 2 x 200GB Western Digital “JB” in RAID 0 Sound Card(s): nVidia SoundStorm via ABIT NF7-S (hardware accelerated), Terratec Sixpack 5.1+ (hardware accelerated), Sound Blaster Live! Value (hardware accelerated), Sound Blaster Audigy2 ZS Platinum Pro (hardware accelerated) and Cmedia 8738 PCI (software) Operating System Used: Microsoft Windows 2000 SP4 Drivers Used: nVidia nForce 4.27, ATI Catalyst 4.8, Cmedia v0644, nVidia Audio driver 4.42, Terratec 5.12.01.3057, AUDIGY2_1_84_50 and SBLive! 1.03.001
We tested every game with a resolution of 1280 x 1024 with AA 4x / AF 8x forced on in the Windows control panel. Vertical Sync was disabled for all testing. Maximum sound detail was enabled where possible to make sure the SoundStorm was working the hardest – encoding Dolby Digital on the fly for true positional surround sound in all games. We also provided results with sound disabled to give you an idea of the overall impact of sound in games on frame rate.
We quickly discovered that providing you guys with a 100% true and accurate measurement of sound card frames per second performance would be a tough job. While we cannot be 100% accurate, we opted for a pure real-world testing environment instead. Except for UT2004 and Q3A (our testing showed a good and consistent difference in results) we did not use any pre-recorded timedemos as we want to provide a true indication of performance – as tedious as it was. Instead for the rest of the games we fired up FRAPS and measured the average frames per second while we actually played the game each time.
This is the part where we cannot be 100% accurate – we played the same stage in each game, shot the same amount of rockets, jumped the same amount of times and took the same line in the drag race but each time there were small inconsistencies which cannot be helped. This method of testing took us much longer (an entire weekend in total) than it would have if we used timedemos but the results are certainly truer than they would have been if we just timedemos.
The fact that it was such a time consuming and tough job for us is another testament to the fact that computer audio is a forgotten component because there is hardly a fraction of the audio benchmarks available on the web as there are graphics benchmark software. Everyone makes a big deal out of a few frames per second difference between graphics cards, but why is there never any mention of the differences that can be achieved in that same few fps between various sound card solutions? Some new audio benchmark software would go along way to helping this situation! Any developers out there listening?
Let’s get started and see how the mighty SoundStorm shapes up against the competition. It is VERY IMPORTANT to remember although the SoundStorm is hardware optimized, it is working in several cases at least twice as hard as the other sound solutions due to the on the fly encoding of Dolby Digital from an original stereo format in all of the games. Plus the SoundStorm also supports a full 64 path sound environment meaning more sound effects were processed at the same time over other unsupported cards (Cmedia and Sound Blaster Live! Value).
