indeed Saltwater ... the entire system need to be perfectly configured In order to run dozen of nebula instances over LAN. for sure, focusing in no crashes, and some decent latency for still be able on mixing and tweak any eventual arrangement task! it demands a lot o inicial work to setup the whole thing. LAN settings also is crucial . In my settings, both Master and Slave machine's LAN adapters has "all" offload parameters disabled, and no jumbo frame! For mixing, I feel confortable with 1024 buffer sizer. At least I get no gaps, pops and clicks in the sound, even running in loop.
Also, in order to VEP5 to run smothly on Reaper (sorry, ..don't have any idea about other DAWs) this settings are extremelly important:
1) VEP Audio input instance: Check both: inform plug in when track channel count changes / ""hard reset on playback start""
2) VEP PRO: Uncheck all

 

Last edited: Aug 23, 2019

.

 

what sample rate are you running?

 

44.1KHz. Do you thing is better try higher sample rates? anyway, i found here a good literature for this subject. First of all , we need to understand that Nebula doesn't do oversampling, and from there, make the decisions for the project.

http://forum.acustica-audio.com/viewtopic.php?t=27724
http://forum.acustica-audio.com/viewtopic.php?t=28304

 

Last edited: Aug 23, 2019

hell no it gets way worse with higher rates!

 

Always use TRIM with SSD. It's not about speed, it's about balancing the use of the chips/transistor inside the SSD. So if you don't use it, the SSD will degrade (transistors die progressively) much faster.
Of course, as a result of degradation, both speed and space will end up decreasing faster.

This is interesting, because Nebula libs are literally sampled effects. So when it comes to oversampling it's like sampled instruments.

The technology is way more complex, of course. For every Nth kernel it uses (number of harmonics it captures) it has to oversample the input signal by Nth and apply the Kernel (it's kind of a convolution in fact).
So if it captures 6 harmonics, it has to oversample by x6 the input signal, apply the sixth kernel and then downsample and mix the others (from x1, that would plain eq matching to 6th passing from 2x, 3x,...).
But this is not oversampling the Nebula effect, it's just the way it works.

That's the main reason Nebula is so CPU hungry. And I'm not even talking about doing this for several input levels (the famous gain stage) and with tweakable controls.