Not sure if this helps, but this can also be a result of the coordination of audio latency/buffer settings chosen in the DAW, Audio Interface, and Neb Instance.
Using NebPro Reverb instance, dspbuffer setting matched to the audio interface and DAW can make the response improved for one rate over another. Too large a buffer on playback and interface compared to NebPro can make longer waits and spikes. All 3 being too short for samples and number of instances can pause or crash or glitch audio. Too long can make too much cpu and mess with calls to and from audio drivers.
I always recommend setting up a fresh DAW environment, load a single stereo track at the sample rate used for testing. Load a single instance of the NebPro program at that same rate.
Play it once with Neb bypassed and look at CPU.
Play it once with Neb on and look at CPU.
Now, adjust playback buffer audio interface buffer, and Neb mast dspbuffer to a higher number than current (if using a really small buffer, start by doubling it- if using a really huge buffer, start by dividing it in half). If the cpu improves with this change, reduce/increase it again. Keep doing this until you have improved as much as possible.
Now, do the exact same thing at 96kHz for a 96kHz program. This can suggest if you will gain something by keeping a record of amounts of buffer to use for different sessions at different rates.
Just as an example, with a moderate computer with decent hard drive speed and the minimum 4 gig ram for capping 32 bit can run more instances at 8192 than it can at 2048, but if the persons audio interface drivers are terrible, or commands between DAW playback and interface are terribly different from NebPro, then this may be incredibly inefficient.
Too low- almost guaranteed problems even on super fast 64 bit system. For instance, not many people with fast enough drive, memory, and processors to load 100 stereo tracks at 96kHz with a full 10 Kernel preamp program 10 times on each channel with a 32 sample audio buffer, but some get dozens and dozens of instances running smoothly at 4096 buffer on Neb, Audio device, and DAW.
Tim Petherick wrote: ianc wrote:
The cpu usage just fries my brain, so I'll take Tim's word about whats really happening, I've got a project open in 44k and the 44k presets usage is slightly more than the 96k of the same preset.
Other presets are the other way round, so I'll just put this down to differences in the presets settings as a possibility, there's not should a large enough distance between the 2 to get a headache about
The only reasons I can think of cpu being different is because maybe the total amount of samples may of been 1 or 2 less between versions and maybe a slight difference in kern length. I have never really compared cpu's of converted and native, so I'm not sure where this was seen, may of been a misunderstanding.