I just tested TAL Noisemaker under Linux and I realized that CPU goes up to 40% when playing it. That did not seem right to me, so I booted into Windows and the performance of that VSTi is about 7%. Also, under Linux, the CPU of that VSTi does never go below 5%, even when not being played at all!
Why is there such a huge difference of VSTi performance under Linux compared to Windows? That does make using VSTi pretty useless under Linux.
Is this related to the way the Linux port of Noisemaker is implemented (also I thought it was just recompiled for Linux), or is this due to the fact how Renoise handles Windows and Linux VSTi?
Are you using the same audio latency settings in Windows and Linux?
I found, at least in linux using jack, CPU utilization is much higher using lower-latency settings.
I actually tested with ALSA and a much higher latency (512 samples) as under Windows (where I use ASIO drivers with 128 samples).
I actually have noticed this too. Some LADSPA plugins also quite easily overload CPU. I wonder if this is maybe a general problem with linux dlopen?
I am still very curious about this. Maybe Taktik can jump in and shed some light on it?
Haven’t come across this too much, noisemaker runs generally around 15-30%, although I have had an occasion last week where a bunch of LADSPAs freaked out the system…
Since my laptop broke, I was trying Noisemaker on my netbook and the problem becomes even more apparent. Under Linux, some presets can’t even be played, because the CPU reaches 100% after a couple of notes being played.
That is kind of ridiculous. Something is really going wrong here.
I am still wondering if this is due to the Linux version of Renoise, or due to the Linux version of Noisemaker.
Surely, someone knows the answer?
Very unlikely that this is somehow Renoises fault when the plug works just fine on Windows. The platform dependent parts of the of the VST API are only about how to load that thing - the library. Everything else works the same way on all platforms. Well, but you never know. Would be great if you could contact the plugin devs for help. Maybe they have a clue…
What can make a big difference, as said above already, is the audio buffer size. If you are using ALSA, not the total latency affects CPU usage, but the single buffer size. At sizes below 64 samples this will be very noticeable already - aka, the smaller the buffers, the bigger the overhead / CPU usage in general - not just for plugins.
fladd: I don’t think it’s Renoise’s fault either, but you could try another VSTi-host in Linux (such as energyXT, not 100% sure if it supports vsti though) to see if there’s any difference cpu-wise.
Energy XT does support linux vst plugins.
XT2 has a lot of bugs, so even if the plugin does also not work in there, this is not prove that it is the plugin’s fault. I will check later nevertheless.
Just tested it, the same happens under XT2. But as I said this is not surprising since most things don’t work there.
Are there any other VSTi hosts for Linux?
It may also be juce framework which is less than optimal under Linux.
Oh, I didn’t know that Noisemaker was built with Juice! That might indeed be the cause. It is a shame, however! This is such a nice synth and I would like to use it on Linux as well. There are so few synths for Linux (at least in a VSTi/DSSI format) and the ones that are there are extremely buggy. I don’t understand that. I was hoping for at least one great synth when I heard that Noisemaker was ported to Linux.