Suggestions for a CPU friendly alternative to Pigments?

[EvilDragon]

Your OS task manager is the one to watch, ultimately.

Reaper's resource monitor has a number of options when you right click on it, you should enable RT CPU as that's the main audio thread CPU usage, and is the most important metric.

[EnGee]

Yes, I always rely on RT measure in Reaper. Oh it's already installed in the Windows PC! Damn you Microsoft for such a pathetic menu/search in W11

Ok let's see what Reaper is telling me!

[syntonica]

Your OS task manager is the one to watch, ultimately.

Only if the DAW sandboxes the plugins so they show up in their own process. Otherwise, the DAW process includes the plugins. It makes for real fun debugging, too!

I can't remember if Reaper does it since it's own CPU monitor is so inclusive, I never bother with Activity Monitor (Mac). I only look at Task manager on Windows if the process is running to slow compared to the optimization level used.

[EnGee]

Well, there is a big difference (many times more than 5%) between Reaper and Live metres but S1 is similar to Reaper so I think I can rely on S1 to give me an indication (not precise I understand) of what CPU usage I have.

Anyway, why I'm bothered about CPU?! I don't have any problem running 10 of Massive X (or PhasePlant) instances plus one instance of XO or Addictive Drums. That's more than I need

[BONES]

I need a synth that fits the swiss army knife role Pigments has so I can fit it nicely alongside the synths I use for specific uses like DUNE, Legend, and Diva.

My first thought is to stop using Pigments and let the other synths you have pick up the slack. DUNE can do most of what Pigments does and it sounds much better. But if you are happy with the sound of Pigments, then maybe Hive would be a good alternative?

[Korg Supporter]

I need a synth that fits the swiss army knife role Pigments has so I can fit it nicely alongside the synths I use for specific uses like DUNE, Legend, and Diva.

My first thought is to stop using Pigments and let the other synths you have pick up the slack. DUNE can do most of what Pigments does and it sounds much better. But if you are happy with the sound of Pigments, then maybe Hive would be a good alternative?

Or maybe even try Surge XT, which is free? I recommend Hive for its amazing wavetable features and great-sounding filters. However, OP should also wait for Zebra 3 to come out. It seems like it will be able to do everything that Hive can do.

[BONES]

But will it be CPU-friendly? Because that's the crux of the issue - CPU.

[Korg Supporter]

But will it be CPU-friendly? Because that's the crux of the issue - CPU.

Urs promised that if you use Zebra 3 like Hive, it will be. Surge XT is also CPU friendly as long as you stick with mostly the legacy oscillators and filters.

[pdxindy]

Or maybe even try Surge XT, which is free? I recommend Hive for its amazing wavetable features and great-sounding filters. However, OP should also wait for Zebra 3 to come out. It seems like it will be able to do everything that Hive can do.

It depends on what features someone is using in Pigments. The biggest CPU user in Pigments is the granular synthesis. With granular and unison you can really crush the CPU. Neither Hive, nor Zebra, Dune, etc. have granular synthesis.

[Urs]

Alright, sent an email. Spire and ArcSyn are the best contenders that can do the same or similar stuff. I don't own Dune, so I can't replicate it. Here are the patches for reference if anyone wants to try them out.
https://drive.google.com/file/d/1NxxlkU ... share_link

So, I was curious and compared the "somewhat expensive patch" to the Spire patch. Spire uses about a quarter more CPU even though Hive uses 4 x 9 unison (36 oscillators) while the Spire patch uses 2 x 9 and 2 x 1 (20 oscillators). Hive furthermore is set to wavetable mode, which is more expensive even though one could have just used the square wave mode (PW via ModMatrix). In that case Hive uses another few percent less. So on my machine (Intel Mac) I can't confirm any advantage of Spire over Hive in terms of CPU usage.

I tried to directly compare the other preset to ArcSyn demo, but somehow I could not import the patch to ArcSyn. So I kind of rebuilt it in ArcSyn, set Hive's unison to 16 (as this is I think what ArcSyn does?), removed the 4th oscillator (ArcSyn has 3) and set ArcSyn's filter to "sideband up" (?). Here indeed ArcSyn has a CPU advantage, mostly because the sideband filter in Hive might be somewhat something different, with further parameteric control over the sound. Yep, apparently Hive's sideband filters take up quite some more CPU than those in ArcSyn. Good find! I'll put it on my list for evaluation if we can squeeze some more out of it.

When comparing ArcSyn to Hive with classic filter types, I could not see any significant difference in CPU performance on 16 voice chords. To the contrary, I found that 3 Oscs with 16x unsion sawtooth and 1 SVF lowpass performed with a tad less CPU in Hive when set to "clean" mode (where the filter is an SVF then). Similarly, when setting Hive to "dirty" and enabling some distortion in ArcSyn's filter section.

Finally, I built another simple patch 3 Osc max Unison + 1 filter from scratch/init in all three synths, set to max voices, hit a chord and got this readout in Reaper:

Screenshot 2023-02-12 at 17.32.02.png

So from my perspective, I can't really see how Hive is considered "not low CPU", even if of course every synth may have ways to use more or less CPU. But when set to real world settings I think Hive commonly does a decent job, and is as good as any low CPU synth.

That said, CPU make, RAM, audio interface, buffer size, samplerate etc etc all could make further difference, in case of which these really close results may shift in favour of one another.

[Korg Supporter]

Alright, sent an email. Spire and ArcSyn are the best contenders that can do the same or similar stuff. I don't own Dune, so I can't replicate it. Here are the patches for reference if anyone wants to try them out.
https://drive.google.com/file/d/1NxxlkU ... share_link

So, I was curious and compared the "somewhat expensive patch" to the Spire patch. Spire uses about a quarter more CPU even though Hive uses 4 x 9 unison (36 oscillators) while the Spire patch uses 2 x 9 and 2 x 1 (20 oscillators). Hive furthermore is set to wavetable mode, which is more expensive even though one could have just used the square wave mode (PW via ModMatrix). In that case Hive uses another few percent less. So on my machine (Intel Mac) I can't confirm any advantage of Spire over Hive in terms of CPU usage.

I tried to directly compare the other preset to ArcSyn demo, but somehow I could not import the patch to ArcSyn. So I kind of rebuilt it in ArcSyn, set Hive's unison to 16 (as this is I think what ArcSyn does?), removed the 4th oscillator (ArcSyn has 3) and set ArcSyn's filter to "sideband up" (?). Here indeed ArcSyn has a CPU advantage, mostly because the sideband filter in Hive might be somewhat something different, with further parameteric control over the sound. Yep, apparently Hive's sideband filters take up quite some more CPU than those in ArcSyn. Good find! I'll put it on my list for evaluation if we can squeeze some more out of it.

When comparing ArcSyn to Hive with classic filter types, I could not see any significant difference in CPU performance on 16 voice chords. To the contrary, I found that 3 Oscs with 16x unsion sawtooth and 1 SVF lowpass performed with a tad less CPU in Hive when set to "clean" mode (where the filter is an SVF then). Similarly, when setting Hive to "dirty" and enabling some distortion in ArcSyn's filter section.

Finally, I built another simple patch 3 Osc max Unison + 1 filter from scratch/init in all three synths, set to max voices, hit a chord and got this readout in Reaper:

Screenshot 2023-02-12 at 17.32.02.png

So from my perspective, I can't really see how Hive is considered "not low CPU", even if of course every synth may have ways to use more or less CPU. But when set to real world settings I think Hive commonly does a decent job, and is as good as any low CPU synth.

That said, CPU make, RAM, audio interface, buffer size, samplerate etc etc all could make further difference, in case of which these really close results may shift in favour of one another.

Thank you! However, I have a feeling that using ring modulation instead of sideband filters would take up less CPU.

[Korg Supporter]

Just an update, on my Dell XPS 13 (i7 6560U, 16 GB RAM, 41 khz, multicore), I can load up to 55 instances of Hive with the Acidity preset and 28 instances with a 6 note polyphonic patch that has 16x oscillator unison and a normal LP filter.