Here's a weird one: I have a very simple session in Ableton Live in which I'm getting a buffer underrun (crackle/dropout) every time the playhead moves. This could happen if I have a section set to loop or if I move it manually. It happens while playing back or while monitoring. The session has a single audio track with a bunch of plugins on it, including S-Gear. CPU usage is low, under 25% in task manager. I suspect S-Gear may be a culprit because I can turn on or off any of the other plugins with no effect, but if I turn off S-Gear or swap it out for a different amp sim, the dropout goes away.

I'm in Windows 10, and I've done all of the standard system tuning I know of. Power throttling is disabled, turbo boost is off in the BIOS (or on, doesn't make a difference), DPCs checked with LatencyMon look not too bad (typically under 100μs, occasional max 250μs from nvidia driver).
Computer specs should be sufficient
i5-4690k
16GB RAM
RME Babyface Pro

Help!

I have narrowed it down further and am pretty sure it's an issue with S-Gear (I've tried both v2.90 and v2.95) and certain DAWs. It occurs in Ableton Live and in Reaper, but does not occur in Bitwig Studio. I can get the dropouts to occur with a single instance of S-Gear on a single track and nothing else in the session. It occurs even if I remove all devices except for the amp (which cannot be removed), and bypass the amp at the bottom, so it essentially acts as a pass-through. It occurs regardless of which amp is selected. It doesn't occur if I replace S-Gear with (many) other plugins, even if the CPU load is much higher.

To really test it out, I put 25 instances of S-Gear on a track and set playback to a 1-bar loop. When the playback was set to loop, there were massive CPU spikes every time the playback head moved back, but when I turn off playback looping, the spikes (and dropouts) went away. You can see the difference in the attached CPU meter screenshot where CPU 1 is very spiky, but after I turned off looping halfway through it flattens out. This was taken in Reaper but Live looks similar. The spikes look small on the meter, but that is only because the meter averages over a window of a full second but the spikes are much shorter.

Thanks for reporting. Can you please send a message via support .

S-Gear is designed in such a way that any CPU intensive re-configuration is done in a backgound task (to avoid a CPU spike in the processing thread). Certain S-Gear parameter changes cause a controlled audio drop-out and this is intentional.

If the audio frame size is changed during playback then this may create some re-configuration overhead in the convolver which causes additional CPU load. It is possible that this is happening on a looped section.

Thanks for the reply. I'll send a support message now.