leileilol wrote:

Coders: those who code the format, player, and also happen to be a musician

Hey ! what about your avatar, this seem to have many strings, what do you know about music ?

Scali wrote:

The PAS was actually quite a decent alternative to the Sound Blaster. It was compatible with SB (also digital sound), but was of considerably higher quality than the real thing (which wasn't that hard).

No the first PAS , the PAS 8-Bit was not compatible to SoundBlaster, hence the combination link between the PAS and the (SoundBlaster compatbile) ThunderBorard from MediaVision.

The PAS was the first card introducing stereo FM sound via its dual OPL2 synthezsiers. Great sound. Great stereo panorama. The PAS is a legend, Creative copied the PAS abilities with its SoundBlaster Pro 1 (CT1330A series) but with lower quality. The historic relevance of the original PAS is great for incorporating Stereo FM music in adventure games where the Sound Blaster Pro could only supply monaural music experience.

Right, I was thinking about PAS in general, mainly the PAS16, which seemed to be the most popular option around here.
It was compatible with ThunderBoard/SoundBlaster.

do you think it make any difference to use a PAS or a SB16 for the soundtrack of OMF2097, everytime I see both in a video game setup, I think they perform exactly the same.

I like how you present the 1992 GUS. I was 12 years old when the GUS went out and as a (french) Canadian I was quite proud this was a "local" product, yet it took me about 5 month of work lawnmowing to put the money to buy one at 275$. I remember my dad offered me the 768kb upgrade, it was a surprise.

I think you are unfair for the Pentium and AWE. The AWE could mix 32tracks at 44.1khz fetching 28mb of onboard waveform samples. That would bring an early pentium to his knee, maybe he could do it, but then there would be no horse power left to render that 24bit 3d image. My point is, the awe32 still had his reason to exist in 1994. But yes, it was about the end.

ElBrunzy wrote:

do you think it make any difference to use a PAS or a SB16 for the soundtrack of OMF2097, everytime I see both in a video game setup, I think they perform exactly the same.

The SB16 perhaps not. Early SB's (SB1.0, 1.5, 2.0, Pro) have quite a distinct colouring to their sound (very fat and compressed). PAS16 has a more neutral/sterile sound, which you also find on later SB cards and many clones.

ElBrunzy wrote:

I think you are unfair for the Pentium and AWE. The AWE could mix 32tracks at 44.1khz fetching 28mb of onboard waveform samples. That would bring an early pentium to his knee, maybe he could do it, but then there would be no horse power left to render that 24bit 3d image. My point is, the awe32 still had his reason to exist in 1994. But yes, it was about the end.

You have a strange idea of CPU cost.
Even a 486 can play 32 tracks at 44.1 KHz without too much trouble, eg in FT2 or with Cubic Player.
A Pentium is much faster still.

I've read that in a publication. Dont remember having done precise benchmarkin myself. Anyway most modules didnt had 32 channels. You know what, I'm not sure at all about that. Of course it's hard to compare software mixing with hardware mixing quality. I'm sure there is info out there to settle that.

ElBrunzy wrote:

Of course it's hard to compare software mixing with hardware mixing quality. I'm sure there is info out there to settle that.

Why is it hard?
There are plenty of players around that support software and/or hardware mixing.
As I already said, FT2 and Cubic Player are among the better ones. Good balance between sound quality and performance.
There are also various players that are much slower and/or sound much worse. Many Windows-based players have very poor performance, especially on 486 machines, because they are coded poorly, and rely on the FPU.
You'll want players from the DOS-era to get properly optimized mixers for 486/Pentium.

Scali : I start to think we dont have the same idea of hardware and software mixing, I would be curious to hear what's your definition ?

ElBrunzy wrote:

Scali : I start to think we dont have the same idea of hardware and software mixing, I would be curious to hear what's your definition ?

...Not sure if trolling or srs...
I think we can have only one possible definition? Hardware mixing is where the mixing is performed by dedicated hardware. Software mixing is where the mixing is performed by the CPU using a software routine.
With hardware-mixing, the quality and performance are 'fixed' by the hardware implementation. All popular cards/chips (eg GF1, InterWave, AWE) have quite decent quality/precision (16-bit or higher), and apply some kind of interpolation during resampling to minimize aliasing.

With software, the quality and performance depend on the specific implementation. So some players sound better than others. Some perform better on a specific CPU than others... And then there are players that neither perform nor sound well.
Early players only supported 8-bit mixing (or even less, pre-degrading the samples before mixing, so the result of mixing is implicitly 8-bit. Reenigne's mixer for the 8088 MPH endtune does the same, only aimed at ~6 bit PWM, so even less precision), and no interpolation, so they don't sound anywhere near as good as hardware mixing on the aforementioned chips/cards. More advanced players can sound comparable to hardware mixing (or even better, in rare cases), but at the cost of more CPU.
Hence my point of picking 'best in class' software mixing players for a comparison. In which case the comparison is very easy: just run the software and compare.

Thanks for your explanation. Please believe that I'm not trolling. But then again, I'm an enigma wrapped in a rid... joking, srs I'm not trolling!

I just once read somewhere on an article that to achieve the mixing quality of a emu8k or an interwave (maybe it was an emu10k too, I remember it wasnt very specific in the article, an emu10k would make more sense too) you would need an pentium class computer. At the time it was early pentium-1 60/66mhz era. I never bothered to make the test and accepted that information as true. I think I remember that by that time when my friend who didnt have a codec soundcard couldnt achieve very good quality in software, but then again they mostly had 486 at the time.

I have mikmod in debian sarge running OSS who give around 5~20% cpu usage on a SMP ppro-200mhz cpu playing a small 4channels chip tune. I'll try to play a 32channel 44.1khz .s3m on it and configure mikmod to act like a hardware codec and see how it perform. I'll give you the results soon.