Half-Life

Sound

Benchmarks

Running the CPU at 300 MHz gave the following results (only some cards)

Update: The new low Live results using LW 2.0 is, according to Creative, specific for Half-Life caused by an incompatibility with the new EAX 2 reverb engine and the way Half-Life implemented EAX. This shouldn't happen in any other EAX games. The LW2.1 improves the score significantly but is still slightly behind the LW 1.0.

Update 2:With the release of the 2030 and later drivers for the Vortex 2 the performance when you disable wavetracing is significantly less while performance with wavetracing remains the same. It seems like the disable of wavetracing isn't working as well as before. You can also see that the new optimized Vortex 2 ships used by the SQ2500 card performs 12.5% better than the old one, but it's still almost 21% slower than stereo.

You can see that the Monstersound only gains about 1% when wavetracing is disabled while the MX300 get a speed boost of 15%. The reason is of course that the Monstersound only supports occlusion not reflection and it's the reflection that requires a lot of CPU power. This can also be confirmed by disabling occlusion for the MX300 (s_occlude 0) which will give you slightly less than a 1% speed increase. Disabling reflections (s_reflect 0) will increase the speed with 14%. You can also see that the speed hit is the same when running the CPU both at 300 and 450 MHz. The reason is that Wavetracing is tied to the framerate, so the more fps the more accurate, and the more CPU intensive, Wavetracing is.

If we compare the scores without wavetracing we can see that the Live outperforms the MX300 with 18% and even the Monstersound outperform the MX300 with 7%. We can also see that $20 Trident card performs amazingly well. This is a surprising low score for the MX300. If we look at AudioBenchmark scores with 16 3D streams the difference between the Live and the MX300 is only 1% (3% CPU hit for the Live and 4% for the MX300).

After talking with the Valve software and Aureal about the scores they gave several reasons why the A3D 2 scores are lower than should be expected. The major reason for the lower score even when wavetracing is disabled is that unlike most other titles (e.g Heretic II, Sin and Recoil) they couldn't replace their sound engine with A3D 2.0 because Half Life sound engine had features which A3D 2.0 could not replicate so the A3D 2.0 engine was integrated into Half-Life's software engine. The feature they couldn't replicate was the reverb engine. Half-Life software engine uses streamed sources while the A3D 2.0 engine is much more efficient with static sources which are what most other A3D 2 games will use, according to Aureal. The reverb engine also cause about 2% CPU hit for a 450 MHz CPU and 3% with 300 MHz something the Live doesn't need to care about since it do its reverb in hardware. If you want to disable reverb when in A3D 2 mode you can do that by typing s_verbwet 0 in the console.

When it comes to the hit of wavetracing it should also be less in other A3D 2 games because of several more optimizations they have done since Half-Life was released. One of them is they realized that their was no need to calculate as many reflections polygons as they have in Half-Life but the number of occlusion polygons need to be the same. In Sin and Heretic II they have taken advantage of that fact and the reflection polygons/occlusion polygons ration should be about 1/4 instead of 1:1 like it is in Half-Life. The A3D 2 engine now also supports a feature where developers can change how often they want to update the reflections and occlusions like every other frame for occlusions if it pass 30 fps and even less for reflections so that the Wavetracing engine doesn't have to do twice as much work if the fps are above 60 compared to only 30 fps. I'll give more details about A3D 2 in an article that should be ready next week. We should hopefully soon have benchmarks from Sin and/or Heretic to see if the MX300 performs better there. According to Aureal the performance hit in Heretic is about 5% for Wavetracing.

Other interesting points about the benchmark is that the 4DWave-DX is almost as fast, 1.4% behind,  as the Live with the CPU clocked at 450 MHz, when clocked at 300 MHz performance the difference is slightly more, almost 4%. The WaveForce is about another 5% behind the 4DWave-DX but the WaveForce only plays 8 3D streams the rest are in software.   The reason the Trident and WaveForce card takes more of a performance hit with a slower CPU is that it do the reverb and 3D sound part in software instead of hardware.  The Live is almost as fast it can be with no performance hit at all for enabling EAX (and 3D sound) and the difference compared to disabling sound  is only 1.5%. The reason you see no performance hit by enabling EAX and 3D sound for the Live is probably because the small CPU hit that takes is about the same as the software reverb which you'll get when using stereo sound in Half-Life.

When I play the game I can't say I noticed that the MX300 performed well  below the Live. One reason is of course that the action usually isn't as intensive as in that demo and I would also guess that scene takes a worse hit of wavetracing than most other scenes in the game. Usually I also played the game at 800x600 or above in which case the fillrate of the TNT limits the performance for several frames instead of the almost only the CPU which is the case at 640x480 with a fast CPU.

Conclusion

When it comes to the sound score it's important to know that each category got a score compared to what we think is possible to do with that technology so you shouldn't compare the EAX score with the wavetracing score. The wavetracing lost a few points on the lack of any water effect, the muffled sound and especially the significant CPU hit.

EAX didn't get a perfect 10 because of the muffled sound, choosing the wrong reverb effect at some places and no way to choose how strong reverb you want, which is possible with a few other titles.

I think the MX300 produces the most compelling sound in this game, this is especially true if you use headphones or 2 speakers. With 4 speakers the Live is pretty close when it comes to 3D sound but I found the wavetracing to be more compelling than EAX. If you care a lot about your fps numbers you're probably best of with the Live and as said with 4 speakers you don't lose much compared to the MX300.

The Trident card's great performance, good 3D sound and reverb with 2 speakers and a lot of 3D streams makes me prefer this card in Half-Life over the Monstersound (or any Vortex1 card) if you use 2 speakers. With headphones you're better of with the Monstersound unless framerates are very important to you.

Music
score: 8

The music is not continuously playing instead they have background music (if we can call it music) using CD-audio just before certain events. The music is excellent and definitely thrilled me a few times but unfortunately it's only a few times the music is played.

Page 5

Game Reviews