Some tricks for slow CPU users..

if you have a slow CPU (PII if you’re not using any VSTi, PIII otherwise),
you could find these ideas useful:

  1. disable “pattern follow” option:
    (just belowe the pattern list)
    this will reduce the amount of CPU usage for video management, while prevent you from automatically follow patern, but you can survive to this

  2. always use the lowest sampling frequency possible:
    This could sound obvious, but using a frequency of 33khz is sufficient most of times

  3. cut tracks that you do not use:
    when you are focusing your attention on a track, but you can’t listen to it well because of CPU power shortage, muting other tracks doesn’t help because the muted channels are still rendered. So cut the muted ones and then use UNDO function before making the changes you need to the track you focused on.
    this tip has become obsolete starting from ReNoise 1.27: now you can mute/disable each single note column or entire track by using the labels on the top of each track or by using the main track DSP element in the Track DSP section.

  4. Use 16 bit mode, when working in windowed mode
    directX handles blitting a bit faster using 16 bits. (added by Pulsar)

  5. use the “working” quality level
    From ReNoise version 1.25, using the “working” quality level (adjustable in the “Config=>Audio” menu) will not switch off VSTi support.
    This is a great solution for a compromise between performances and quality.

  6. reduce the maximum number of voices when using polyphonic VSTi’s (added by Johan)

  7. Render VSTI tracks to sample to reduce CPU consumption (added by Vv)
    By selecting tracks that use VST instruments and then using the “Render to sample” feature inside Renoise 1.5 and higher upon the selection, you can reduce 20% to 70% or possibly higher (depending upon the VSTI cpu consumption efficiency and the amount of VSTIs you render to one sample) since playing samples won’t take as much cpu resources as when playing a VST instrument.
    Insert a new track to play the rendered selection in, without effects and mute the original VSTI track(s).
    You have now possibility to either mixdown your song or extend your mix with new instruments without cutting out possible important elements.
    Don’t forget to mute the track playing the rendered sample and unmute the VSTIs when rendering to disk. Also don’t forget to apply any specific mixdown you did on the rendered items to the VSTI track(s) as well before rendering to disk.

  8. disable HyperThreading (P4/Win)

HypertThreading is a virtual multiprocessor technology for P4 class CPU’s. It creates a virtual second CPU which should improve multimedia performances.

If you have a P4 with HT, you should disable it when using ReNoise: DualCore and other multiprocessor technologies work fine with ReNoise, but HT doesn’t. If you use HT, ReNoise will be limited on one virtual CPU, thus allowing only 50% of CPU resources to be used.

Most likely you can disable HT via BIOS; you will then need to make a repair installation of your Windows.

If you can’t disable HT via BIOS (f.e. with ASUS p4V800-X motherboard), you can still edit your boot.ini file, adding a /ONECPU option to the boot entry you want to modify. You can also copy a boot entry, paste it on another line, and add the /ONECPU option to it, allowing your system to be booted in multiprocessor or single processor mode. This is not as performant as with BIOS disabling, but it’s still better than using ReNoise at 50%.

improvement to what’s written above:

changing Windows kernel to the single-processor one may increase general performances.

you can specify the kernel to be used by the /KERNEL option.

Make sure you copy the file ntoskrnl.exe from your Windows CD (I386 folder) or
C:/windows/ServicePackFiles/i386
to Windows’ System32 folder.

here it is a sample boot.ini file (it’s the one I use):

  
[boot loader]  
timeout=5  
default=multi(0)disk(0)rdisk(0)partition(1)WINDOWS  
[operating systems]  
multi(0)disk(0)rdisk(0)partition(1)WINDOWS="XP without HT" /fastdetect /NoExecute=OptIn /ONECPU /kernel=ntoskrnl.exe  
multi(0)disk(0)rdisk(0)partition(1)WINDOWS="XP with HT" /fastdetect /NoExecute=OptIn  
  

this configuration starts Windows without HT after 5 seconds during which the user is able to choose Windows with HT enabled.

  1. Use 16 bit mode, when working in windowed mode

directX handles blitting a bit faster using 16 bits.

I remember hearing that 32-bit is often the fastest bit depth because modern gfx cards are optimized for that. Is this not true then? :unsure:

remember hearing that 32-bit is often the fastest bit depth because
modern gfx cards are optimized for that. Is this not true then?

its not true for renoise.

32 bit mode is not faster than 16 bit mode.

but it should be better/faster in some ways than 24 bit mode - because of 4 bytes alignment (32 bit) rather than 3 bytes (24 bit).
It is faster to multiply or divide something by 4 rather than 3!

therefore - the difference in speed.

however, it may also take some time to copy extra byte (alpha) so it can make some slowdown.
so the final score = speedup (on 4 byte alignment) - slowdown (extra byte)

p.s. many cards don’t even have 24 bit mode for the reasons above.

i assume that 32 bit processors can copy (blit) two 16 bit pixel at once, so therefore it should be faster.

Ok, thanks for the answers guys! :)

is it possible to delete just one track at the time… if i just use 10 tracks, i have to put 11 tracks in it… can i delete the 11th track in some way?

I had to read this about 20 times, but finally managed to understand what you meant (my fault, sorry :) )

Deleting the 11th track won’t be a benefit in this case, because unused (silent) tracks do not take part to the mix, so don’t worry.

PS: I’ve added something very useful in the first post of this thread.

ok thanx for the tip.

my apologies for not writing english very well.

I use an Pentium II, 350 mhz, with 196 MB internal memory. and i have made an song with 10 tracks used. But when i play the song it sometimes jams, it gets stuck for a little while. I there something more i can do about this? because 10 tracks arn’t much track. With madtracker i could use almost 30 tracks before the song crashes somethimes.

and also, when i register my Renoise, and i use the render option, will my song be rendered in a good condition, without any flaws?

Yes, rendering will never be affected by a too slow computer when it comes to audio quality. It will only be slower :)

hey IT-Alien … i think you are a the real one and only copy/paste/cut/undo freak arround the world!!! ;) :lol: :D :rolleyes: :)

AND

switching off patternfollow and 16bit mode takes no effect on win2k …

tested with a song with 10 tracks 1-4 columns per track. 12 instances of kontakt, some vst-effects (reverb), some Renoise-effects (filter, eq, automated device)

30-50% cpu … stuttering sound with 50% cpu on big patterns …

i think i should incease the outlatency of the asio-card

don’t know … <_<

It does make difference when you have an 8MB video card :)

oh interessting - ok my gfx-card is quiet lame … only an ATI fury pro but with 32mb … :unsure: … but maybe you are right …

but no problem my notebook is great and Renoise runs 10times faster ;)

As a side note, I use soundfonts through SFZ VST, and so far most of them take approx. 10% CPU, certainly more than even the most resource-hungry VST instrument. (The system is a 2.4GHz P4.) So the solution of yours may not be the best in some situations…

@kiste … maybe SFZ VST is’nt good coded … try to use other softsampler VSTi’s like sampletank, kontakt, kompakt and so on … by the way SampleTank2 (now available) is great!

I don’t think that it would change anything: the priority of access to control panel is not related to the speed of hardware access

it’s maybe a bit circumstantial, but for those who are using a lot of VST effect plugins, it might be useful enable and disable the respective effects by pattern automation (effect xf00 / xf01) according the the song’s requirements.
in plain language this could for example mean, that i disable a compressor, reverb or whatever kind of plugin on the respective channel until it is really required.

in my songs i have parts that never play together simultaniously, so this presumed, the method can save quite some CPU time… but of course it depends on the individual song scenario and arrangement.

that’s a great optimalisation trick Keith (+ experimentation potential :)

I’ve read about it before, but just did some experimenting with it for the first time. I like how you can switch different effects on & off, create fast gates or weird mixes of effects. Will be using this in the future.