Ned Bouhalassa wrote: > > Following a posting on osxaudio regarding problems with EXS, I decided > to do a test of EXS' polyphony: > > My questions are: what about the 1,000 voices from the G5 unveiling? > How about just 150, without drop-outs/crackles? Is this related to my > use of the factory drive, as opposed to using another, audio-dedicated > drive? Can anyone confirm any of this? > > Here's my setup: G5 Dual/2gig ram + Panther + Logic 6.3.1 + internal > 160 gig factory drive. Hi Ned, 1000 voices could only be achieved with DFD off. Samples loaded into RAM in 32 bit mode would be the most efficient way for the EXS24 to operate in terms of CPU cycles consumed per voice. The maximum number of streamed voices would probably be obtained with a setup where you had many drives operating in parallel with the instrument audio files distributed over the set of drives -- whether in RAID 0 or just distributed across the set. A bunch of SCSI or SATA controllers with multiple drives attached would be the ultimate solution -- whatever is most CPU efficient. All samples streamed from disk must be converted to 32 bit format and this consumes CPU so there are CPU limits associated with disk streaming. Such a setup would definitely deliver more voices than your single sample drive setup but would chew CPU cycles ( esp with Software RAID controllers) are and would eventually overload the PCI bus. The people who are getting terrible disk streaming performance -- your sample drive should be defragged, your samples should be in the fastest section (first partition) of a dedicated sample/audio drive which is preferably at least 7200 rpm. If there is a choice of cluster size when formatting you may get better performance with larger cluster sizes (I have no evidence on this wrt Macs). In recent tests on a P4 2.4 PC with the EXS24 (and also with the freeware Sfz sf2 player) and the free Maestro grand piano (gig format converted to SF2 and EXS24 using CDExtract) I am getting over 60 sustaining voices (pedal down, every note playing a different audio file, no repitching of samples) from a Seagate Barracuda ATA 5 120 gig drive without clicks or pops. Regards, Murray
Message
Re: [EXS] Disturbing test regarding EXS polyphony
2003-10-28 by Murray McDowall
Attachments
- No local attachments were found for this message.