I was also wondering if it's normal that my audio starts to crap out and stutter after using 5 instances of nebula even though my CPU performance meter says i'm only at 15% ??? Sorry if these questions are newbish, but I used to only run one or 2 instances for guitar cab emulation or an analog reverb effect.
Performance meter in your OS task manager or performance meter in the DAW?
CPU metering in the DAW is (always going to be?) broken with Nebula. The DAW can't measure how much Nebula does, because Nebula does a lot on the background.. or something like that. Use your OS task manager to see total CPU usage in all processes.
Remember that the 'reverb' version has higher latency, but because of the bigger buffers it can use multiple cpu cores better, so you might be able to squeeze more instances out of it. But because of the latency it is not usable for live-tracking / recording.
And if you put 5 to 10 instances on a single track, yeah, that track might be all out of CPU power. Most DAWs these days can calculate multiple tracks in parallel (on multiple CPU cores), but if you put a lot of instances on a single track you might run out of power before your CPU is even maxed out.
Right now I'm using 3 tracks + the master. On one I have superior drummer and a nebula console prog. On another I have kontakt 5 with rickenbacker bass, a nebula eq and a nebula console, on the next I have my guitar track with a nebula console and a nebula cab and on the master I have a nebula console and a nebula reverb. That's 7 instances of nebula in real time with no stutter, an average of 2 per track. My DAW says I'm at 35% and Windows says I'm using 50% cpu, if I add but one more instance of nebula, all goes to crap and I would nead an additional 3 from my understanding for proper console emulation (a line in on each track). Is this normal?
There are more factors at play than CPU performance in a computer, so who knows where the limit is hit. It's clear that you hit a limit.
Once more, you can't talk 'in instances of Nebula'. A real reverb or room emulation (long captures) takes up exponentially more power than a preamp or EQ emulation for instance.
What kind of CPU do you have? There are things like (for a lack of better words) 'fake cpu cores' that may paint a better picture for CPU-power left than is really the case.
A quadcore Intel i7 has hyperthreading, so is reported as having 8 cores. Unused parts in the CPU logic can than be used as 4 extra cores. But it is still a quadcore. If all the processing is using the exact same part of the CPU-silicon, there is nothing to benefit from the 4 extra cpu cores, so they are useless. Windows for instance sees 4 cores used, and 4 unused, and reports CPU activity at 50%, while in fact your CPU is maxed out. Those newer AMD cores that are built up around 'modules' have a similar effect. Every module has '2 integer cores and 1 floating-point core'. Your CPU has 4 modules, so 8 integer cores and 4 floating-point cores. Windows sees 8 cores. But Nebula is doing only floating-point math, so you only have 4 cores available. Windows reports 4 cores unused, and thus CPU at 50% while it is maxed out (because AMDs claim of it being an octa-core are not really true).
That shows that there are more factors than CPU that affect how much more instances of stuff you can do before your DAW starts choking. ASIO buffer being the clear example of this.
And yes, you hit weird stuff like this with Nebula more often than with other plugins. It just takes more of your whole PC to work with Nebula. Ram and cpu are both put to the test, and with all the multi-core calculations going on there is a good chance that you hit weird bottlenecks in your system, specially if you try to run your DAW at low latency.
Ahh i see, i have an intel i7 3770k so the remaining 50% are fake cores... Thanks for your help jorismak you're very helpful around these forums. Do you have an answer to my previous question on how to do proper console emu in reaper using vst instruments?
Just to clarify that the cpu usage you see in a task manager might not mean your PC is free to do whatever, there are other tasks.
And I _imagine_ (I'm not the developer of course ) that Nebula is for the major part floating-point operations. It's primarily 32bit or 64bit floating point as I understand.
I didn't own any Bulldozer-or-later AMD cpu's, but yes, the 'modules' consist of 2 integer cores and one floating-point core. Please correct me if I'm wrong, but the way I see it if an application would use 100% floating point operations you only have 4 cores to do the work. There is always some logic and integer work to be done in a program (memory operations and branch logic and stuff like that) but the major part of Nebula's CPU hit is floating point (the same with any other audio plugin I imagine).
triggerthehorizon wrote:Do you have an answer to my previous question on how to do proper console emu in reaper using vst instruments?
Not a fan of going off topic (but this thread has seen 90 topics already I believe ), so I _try_ to keep it short.
First, from reading AlexB manuals he explains it pretty well. But the way I see it (and once again, people please correct me if I'm thinking wrong):
'Console' way, is to put an 'input' preset first in your chain. This is assuming you have a 'recorded' signal, you would put the console input on it first. In case of a VSTi instrument (kontakt, synth, stuff like that) you would place it directly after the instrument (the part that generates audio). In case of guitar amp sims, you would place it directly after the sim (or after the speaker sim to be more precise).
Now, you could use a) a preamp-preset (from another library or something) followed by a line-input program or b) use a mic-input program.
Then put the mixbus-preset on the master bus. I place it directly after things like glue/bus-compression. If I go into the 'mastering' phase, I put any mastering plugins (and master tape and brickwall limiter) after the console mixbus. But any glue/bus-compression is still in front (if I leave them in at all).
The other way of doing it is more like using an analog summing mixer. The (line) inputs would go _last_ on a track, after all the rest, and the mixbus-preset would go _first_ on the master bus. To get the effect of going out to analog, back into digital into your master bus.
That being said, people in the analog domain often record with things like 1176's or hardware EQ's before the console. So they record (for example the guitar part) directly commited with the compressor and / or EQ in the chain. So with Nebula / my DAW I regularly put a preamp preset after my amp sim, followed by 1176-mojo and then some broad tone-shaping EQ, into line-input of my console (still VBC or CLC in my case). If happy with the take(s), I bounce / render that track to a file, and start mixing with it in another project with Reaper's default surgical EQ's and simple compression.
During mixing I put the group-bus things on the matching groups/busses, the glue-compression on the master bus at the start of the mixing session with mixbus-preset after it on the master.
I don't know if Nebula (mixbus/stereo) programs can capture things like crosstalk.. so the next part may be 'just in my head', but I put an instance of SleepyTime's free Crosstalk plugin after the Nebula mixbus-console preset, on its 'pro mode' at around 25% / 30% crosstalk. I have the feeling it just gives that little bit extra glue, but as I said, might be in my head .