cpu savers
Posted: Wed Feb 08, 2006 8:29 pm
people frequently ask me how they can keep their processor load down, so i'm gonna do a little post. it will be by no means complete, as i'm about to go run errands and have limited time, but i'll be back.
1. busses, especially for reverbs
-many fx, especially reverbs and things that add modelled harmonic content to a sound just GOBBLE cpu resources. you should really only have one or two reverb busses in a tune while you're writing. come back later and go bananas on your mixdown, by all means, but while you're writing, don't go puttin reverb and chorus inserts on everything or you'll run out of processor really fast. a delay or beatrepeater doesn't use that much because it is just copying the audio data that is already there, but a reverb has to do extensive real time harmonic modelling and it really is a lot of math.
2. don't chain too many inserts on any one track
-each time you add another effect to a stream you are removing what you hear fromt the original sound by another algorithm. if you get 6 or 7 plugins on a sound, and you feel it's integral to the piece, consider routing the output to another channel and sampling yourself. you can always freeze the big fx chain for later editing if you want to change something... which brings me to:
3. sample yourself
-seriously. you should be sampling yourself all the time. if you are just in the digital realm the sound quality loss is truly minimal, and you'll be able to make all types of sounds that would use the whole processor just to play once. this is especially important for complex one shot sounds like splashes and fx.
4. use a sampler with a large number of sample mapping locations and multiple outputs.
-if you use a sampler like kontakt when you produce, you can just dump in a ton of samples, one per key, and then jam out your song with midi. this has the advantage of being able to switch samples very easily (for writing fast), and you can re-use a sampler multi for another tune later if you don't use all the same samples ad nauseum. when you use bigger samplers, and play in your midi, you will find your songs have a more 'real music' kind of feel, and it will also save you gads of processor not to be running a million impulses all the time.
i find that even if you keep it simple, and your fx chains short, having nine or ten impulses running in a track really bogs you down. if you have only one sampler vst, and you take advantage of all the outputs and internal fx, this isn't a problem.
5. when a tune is almost done, and you're out of headroom, bounce out all of the tracks seperately and then start a new ableton set with them.
-this one has multiple advantages. first off it's like freezing every vst, so your computer only has to play a bunch of audio files. even if you have 20 tracks your processor load will stay way down.
-it will also sound a lot cleaner. when you are listening to a bunch of math happen in real time ableton (and any other sequencer) usually cuts a few corners to make itself more stable and responsive. when you render a piece of audio, ableton just does the math at it's own pace and gives you a perfect output file. a lot of people who swear by doing their mixdowns in cubase or whatever get their sound quality difference here. this is true of any good audio program, they all do the math perfectly when rendering, but cut corners when playing back through the soundcard 'live'.
-if you want to make major arrangement shanges or edits, this is really nice too. especially if you have done something like mixdown your drums into one or two files. it is really easy to do hiphop style drum edits/cutups. it is also sometimes nice to use the finished bounces as source material for further manipulation.
ok showertime... more later.
peace!
d
1. busses, especially for reverbs
-many fx, especially reverbs and things that add modelled harmonic content to a sound just GOBBLE cpu resources. you should really only have one or two reverb busses in a tune while you're writing. come back later and go bananas on your mixdown, by all means, but while you're writing, don't go puttin reverb and chorus inserts on everything or you'll run out of processor really fast. a delay or beatrepeater doesn't use that much because it is just copying the audio data that is already there, but a reverb has to do extensive real time harmonic modelling and it really is a lot of math.
2. don't chain too many inserts on any one track
-each time you add another effect to a stream you are removing what you hear fromt the original sound by another algorithm. if you get 6 or 7 plugins on a sound, and you feel it's integral to the piece, consider routing the output to another channel and sampling yourself. you can always freeze the big fx chain for later editing if you want to change something... which brings me to:
3. sample yourself
-seriously. you should be sampling yourself all the time. if you are just in the digital realm the sound quality loss is truly minimal, and you'll be able to make all types of sounds that would use the whole processor just to play once. this is especially important for complex one shot sounds like splashes and fx.
4. use a sampler with a large number of sample mapping locations and multiple outputs.
-if you use a sampler like kontakt when you produce, you can just dump in a ton of samples, one per key, and then jam out your song with midi. this has the advantage of being able to switch samples very easily (for writing fast), and you can re-use a sampler multi for another tune later if you don't use all the same samples ad nauseum. when you use bigger samplers, and play in your midi, you will find your songs have a more 'real music' kind of feel, and it will also save you gads of processor not to be running a million impulses all the time.
i find that even if you keep it simple, and your fx chains short, having nine or ten impulses running in a track really bogs you down. if you have only one sampler vst, and you take advantage of all the outputs and internal fx, this isn't a problem.
5. when a tune is almost done, and you're out of headroom, bounce out all of the tracks seperately and then start a new ableton set with them.
-this one has multiple advantages. first off it's like freezing every vst, so your computer only has to play a bunch of audio files. even if you have 20 tracks your processor load will stay way down.
-it will also sound a lot cleaner. when you are listening to a bunch of math happen in real time ableton (and any other sequencer) usually cuts a few corners to make itself more stable and responsive. when you render a piece of audio, ableton just does the math at it's own pace and gives you a perfect output file. a lot of people who swear by doing their mixdowns in cubase or whatever get their sound quality difference here. this is true of any good audio program, they all do the math perfectly when rendering, but cut corners when playing back through the soundcard 'live'.
-if you want to make major arrangement shanges or edits, this is really nice too. especially if you have done something like mixdown your drums into one or two files. it is really easy to do hiphop style drum edits/cutups. it is also sometimes nice to use the finished bounces as source material for further manipulation.
ok showertime... more later.
peace!
d
4 sure dude, i've been doin that alot of late, and it saves alot of frustration. Sure it takes a bit of extra time 2 do, but when its played back without CPU spikes, clicks & pops etc, i feel alot better about live, and u can then twist the audio in any way u c fit plus use that extra CPU headroom 4 extra FX etc......
