I have Gigastudio 2.2 in Win98, and a Sound Blaster Audigy. When my sequencer plays and that Gigastudio comes at about 64 voies of polyphony, notes play, but some of the notes cut short.... is it a hardware limitation of the Sound Blaster Audigy?
Actually, assuming 48KHz samples @ 24-bit, 160 voices would be about 31MB/Sec sustained throughput.
I don\'t think PCI would be bothered at all by that, assuming you didn\'t have a lot of other things going across the bus. (I\'m assuming the sample data comes off the disk, across the PCI bus and into memory, and then GSt takes the data and sends it back across the PCI bus as a few channels of data to your sound cards.)
However, I am very sure your right about the disk issue, assuming all your samples were on one disk. Getting 31MB/Sec sustained throughput from one disk (IDE anyway) is pretty much the limit right now. I\'m currently trying to get around this by putting samples on multiple 1394 drives, and it is apparently working... (I do have some clicks and pops I desperately need to get rid of...)
Mark I dont know where you were taught math, but I get 160*48000*3 to approx. 23mb/sec. And since GS doesnt handle 24-bit, and I think all samples are 44100 hz, the real demand for 160 voices is 14mb/sec.
I use a Sblive for gs as my only other card is a sw 1000 yamaha card..which i use to run audio smoothly in sonar.. I dont have any problems using the live for Gs only that its limited to only 2 outputs. The main reason behind this (as the sw1000 is a gsif card)
is that in sonar i need the best card for the audio as it seems gigastudio never has any problems with polyphony
I used to get full 96 note polyphony from my Sound Blaster Live. The SB actually used less system resources than my current card. The latency was pretty good considering it\'s not a GSIF card. The card was on the noisy side though.
Your right. I slipped a digit and multiplied by 4 instead of 3. And while I agree that *2 would be more appropriate today, you can\'t blame a guy for dreaming!
Still, the disk drives are the issue, and not the PCI bus. I\'m find from doing measurements that the demands are significantly worse than just looking at the sustained throughput required by the above (previously incorrect) calculation. The issue is very easy messed up by seek times on the drive, because the samples are not guaranteed to be called in order, and because thee\'s no way for the system to cache what it doesn\'t know it\'s going to need.
Also, we are talking MB, and not Mb or mb here. Correct? Hey, at least you called me on it!