buzz forums

onecircles: That's a pretty cool machine you got there. Better than mine. And I also run at 96000 Hz always, so your machine should be okay.

Maybe you're using WAY more machines in a song than I do? Maybe post a screenshot of one of your songs (the F3 page) that is using a bit too much CPU, and maybe we'd be able to tell from that if you're simply using too many machines.

I think we can fix this. There are sometimes ways to optimize the mix of a song without losing mix quality.

onecircles wrote:but some buzz machines can eat cpu. Jacinth and uranus come to mind.

you might be suffering from denormal issues here. uranus has an updated version that fixed it.

Is there a newer version of Uranus than b6? It just uses a lot of cpu with the room quality set at 32 %D I also discovered that pulsar can use my entire cpu if I get a lot of overlap going and max the oversampling. I definitely want to max my cpu. It's just a question of to what aim? I thought that doing the stereo channels would double the amount of data? I don't know anything about this pardon my ignorance.

Hey wait though, this is just the machines, before I set everything up. Now all the MixIO are before the guitar rigs, and the final MixIO has everything separated into proper channels. I also renamed all the machines to make them easier to follow in pattern view.

I just don't want you guys to think I'm some kind of chump up here.

i just tried uranus b6 and it definitively has heavy denormal issues. once it doesn't get any more input (silent parts) the cpu meter freaks out to almost 60%. what you can do is keep uranus constantly processing with like -200dB whitenoise or something.

there is a newer version from 2012 which might not have this problem anymore, but i didnt test it.

http://www.buzzmachines.com/machineinfo.php?id=1049

Woah thanks for looking into this for me. What is denormaling really? I've never heard it mentioned.

I originally thought it was another name for DC Offset but looks like I'm wrong - not seen/heard about this effect either!

EDIT: Can any of these code sources be used in a machine before the Master (or even in the Master) to prevent this - http://ldesoras.free.fr/doc/articles/denormal-en.pdf (Pages 5-9)

Mono:


Stereo:


Surround:

mcbpete wrote:EDIT: Can any of these code sources be used in a machine before the Master (or even in the Master) to prevent this - http://ldesoras.free.fr/doc/articles/denormal-en.pdf (Pages 5-9)

that's what i was talking about. btw that's firesledge (of buzz fame) who wrote that paper
no, you can't simply insert something before master to fix this, you have to either recompile the machine with appropriate compiler settings or if its an effect you can keep it constantly processing ("big enough numbers") by feeding it some (inaudible) audio. you have to avoid the floating point calculations going into high precision mode, so this can't happen after the fact that a "buggy" effect already processed your samples (and caused your cpu to spike).

onecircles wrote:Woah thanks for looking into this for me. What is denormaling really? I've never heard it mentioned.

About denormal pac told me this:
viewtopic.php?f=2&t=1989