mte Posted June 28, 2016 Share Posted June 28, 2016 Hi, I've always thought the CPU on my macbook pro 2012 is the culprit for the glitches I had when using heavy patches like Ravenscroft piano sample library + some kontakt strings sample library + a large omnisphere patch together with quite some sustain pedal usage but found out this is actually not the case. The strange thing was that glitches almost never appeard until I released the sustain pedal and changed chords together but very quickly appeared at these conditions even at relatively low polyphony used. At these conditions plugins have to: - play release samples for the notes in first chord set - play new samples for the changed chords - play some other samples due to sustain pedal actions all at almost the same time. So there's quite a lot of loading involved here. So I've experimented a bit and found the REDUCING kontakt's dfd preload size to 6kB (from default 60) not only improved loading time (which wasn't really a problem since SSD is fast) but also almost eliminated glitching at the sustain repress and chord changing - thus enabling me to have quite a bit higher polyphony! So I'm thinking why is this? Loading samples from RAM should be faster than direct streaming right? So why does streaming give me better results then? What are the technical details about this? I'd like to know more so I can tune the parameters even better. Thanks, Matej Link to comment Share on other sites More sharing options...
Markay Posted June 28, 2016 Share Posted June 28, 2016 I have very limited knowledge with Kontakt but my recollection is that the DFD setting is designed to pre-load from a slower HDD. The SSD feeds up data at a faster speed making the DFD buffer an unnecessary intermediate step. Reducing the DFD buffer size means it fills and flushes faster minimising the delay in the data getting loaded into RAM. MainStage 3 | Axiom 61 2nd Gen | Pianoteq | B5 | XK3c | EV ZLX 12P Link to comment Share on other sites More sharing options...
Recommended Posts
Archived
This topic is now archived and is closed to further replies.