Korg Collection 5

[planetearth]

Watch the plugin boutique video with Dan. He explains the CPU usage there. Short version: it's a complex model and it runs all voices pretty much all the time based on polyphony and there's 4x OS by default.

It's a great walkthru and worth watching anyway.

Do you have a link to the video? I checked out the offer for Korg Collection 5 on PluginBoutique.com, but only saw a link to a generic video on the new collection. It's the same one available on Korg's site and YouTube. It's just a two-minute overview of the instruments, not an interview with anyone.

Steve

[Funkybot's Evil Twin]

Do you have a link to the video? I checked out the offer for Korg Collection 5 on PluginBoutique.com, but only saw a link to a generic video on the new collection. It's the same one available on Korg's site and YouTube. It's just a two-minute overview of the instruments, not an interview with anyone.

Steve

Here you go!

[planetearth]

Do you have a link to the video? I checked out the offer for Korg Collection 5 on PluginBoutique.com, but only saw a link to a generic video on the new collection. It's the same one available on Korg's site and YouTube. It's just a two-minute overview of the instruments, not an interview with anyone.

Steve

Here you go!

Thanks! This is considerably longer (and more informative) than the one I saw!

Steve

[EvilDragon]

Then why did their Wavestate which is not an analog emulation also have the same problem which has since been fixed?

There can be any number of reasons for unintended CPU churn. It's not always denormals.

[danatkorg]

I see that a poster has some questions about CPU usage.

It's absolutely a heavy plug-in. That's the price for doing a real model of the 2600, with all of its saturating opamps (many of them!) and idiosyncratic modules. Even the envelopes are computationally expensive. Hopefully you'll find that the quality is worth the CPU cycles!

All allocated voices are also computed all the time, for two reasons. One is that it's a 2600, and so note-ons are just one way to trigger sounds (though self-triggering patches are most useful, in my experience, with only a single voice). Another is that all states are maintained, just as they are on a real analog synth. For example, envelopes continue to run, and if retriggered do so from their current positions; oscillators and LFOs continue to run, so their phases maintain their relationships; filters continue to resonate, and if retriggered the resonance sounds instantly, allowing sounds like "Kick Techno Res" in the factory set; etc. All of this contributes to the plug-in's notably "analog" character.

For all that, it's really not particularly stressful for modern computers. On my two-year-old Mac Studio Ultra, I can use more than 50 instances simultaneously. The sequence I played in the Plugin Boutique livestream a few days ago (see link below), starting at around 13:15, uses 14 instances of the ARP 2600 without breaking a sweat.

The manual has this note on page 1 (literally):

For best results, use a modern, powerful computer
The ARP 2600 is a complex instrument, and this software model is designed to deliver both unprecedented accuracy and stunning audio quality. Together, this complexity, accuracy, and audio quality require a significant amount of CPU processing power. The factory sounds are stored with lower polyphony settings in order to play well on a broad range of systems. Modern, relatively powerful computers should be able to play up to 16 voices per instance at the default 4x oversampling. If you encounter CPU performance limitations, try setting the Oversampling Mode (see page 16) to Always Use 2x.

On page 7, you'll find this:

13: Poly # Notes
This sets the polyphony when KEYBOARD is set to POLY. The value is saved in the Performance. Since some 2600 patches don’t require any note-ons to make sound, the specified number of voices runs all the time, and uses CPU resources regardless of whether or not notes are being played.

Page 16 has info on Oversampling Mode, as referenced above:

Oversampling Mode
[Use Trim Pot Setting in Performance, Always Use 2x, Always Use 4x]
This lets you globally override the per-Performance Oversampling setting (see “Oversampling” on page 58). Use Trim Pot Setting in Performance: Performances use their individual Oversampling settings.
Always Use 2x: Oversampling is always set to 2x, regardless of the Performance settings. Use this to optimize CPU usage.
Always Use 4x: Oversampling is always set to 4x, regardless of the Performance settings. Use this to optimize audio quality.

The Troubleshooting section also references the points above, including another note that all allocated voices run all the time.

And, the online FAQ has the same info: https://support.korguser.net/hc/en-us/a ... erformance

Btw, regarding wavestate native: I think the user is referring to the very early version of the plug-in, which used a multi-threading method that didn't mesh well with Apple Silicon. (It works very well on the wavestate/modwave hardware, on the other hand.) We solved that issue by making the audio portion single-threaded. This isn't an issue with any of the other plug-ins.

[danatkorg]

I tried the ARP2600 demo:
Blurry at 120%!
Except when I set Cubase to 125% scaling or Windows to 125% for the standalone version. -> not blurry

Thanks for mentioning this. I filed an issue in our database a few days ago. The graphics assets are actually quite large, 3212 x 1868, which with normal plug-in headers and footers almost fills a 4k screen. This scales quite well, both up and down, on HighDPI monitors. However, if the monitor is not HighDPI, then it seems like the graphics are first scaled down to 50%. After that, increasing the scaling - as one might do on a monitor larger than HD but not quite HighDPI - seems to start with the 50% scaled version and then scale it up again, resulting in the blurring that you describe. Presumably using the application or Windows scaling bypasses this and uses the full-size assets. I have to think that we'll be able to find a solution that simply scales down the full-size version. In the meantime, I'm sorry for the inconvenience.

[T-CM11]

Thanks for mentioning this. I filed an issue in our database a few days ago. The graphics assets are actually quite large, 3212 x 1868, which with normal plug-in headers and footers almost fills a 4k screen. This scales quite well, both up and down, on HighDPI monitors. However, if the monitor is not HighDPI, then it seems like the graphics are first scaled down to 50%. After that, increasing the scaling - as one might do on a monitor larger than HD but not quite HighDPI - seems to start with the 50% scaled version and then scale it up again, resulting in the blurring that you describe. Presumably using the application or Windows scaling bypasses this and uses the full-size assets. I have to think that we'll be able to find a solution that simply scales down the full-size version. In the meantime, I'm sorry for the inconvenience.

Thanks for looking into this Dan!

ps. I think the software makes a DPI "guess" based on resolution. In my case, when I type in the following command "wmic path Win32_DesktopMonitor get /format:list"; Windows doesn't know the size of my desktop monitor (in cm) (this is of course needed to know the DPI).

[pummel]

do all the vst's in this collection use a common filter?

[danatkorg]

do all the vst's in this collection use a common filter?

They do not. The ARP 2600 alone has three different resonant lowpass filters to choose from.

[zvenx]

Three or two low pass filters?
rsp

[danatkorg]

Three or two low pass filters?
rsp

Three. 4012 and 4072 up there (LP 1 and 2), and then the mode-sweep SEM-style filter below. There is also one other audio-rate low pass filter, but it’s not resonant and has a low max cutoff…bonus points for naming it!

[bmanic]

.. the AR lag module.. aka lpf gate.

[gentleclockdivider]

.. the AR lag module.. aka lpf gate.

Aren't (most ) low pass gates vactrol based compared to ar lag modules ?

[bmanic]

Ah yes probably.. but I just wanted to answer Dan's bonus points question as a pathetic little goody two shoes!

The AR, if done correctly, is probably another filter.

[danatkorg]

.. the AR lag module.. aka lpf gate.

The AR does act as a lag filter for the GATE signal (or whatever other signal you patch in there), but it's different from a lowpass gate, and with audio signals I find it to be hard to get good results. Still worth experimenting with, though - that's the fun of the 2600!

Instead, I was thinking of the Lag Filter in the voltage processors. That's a 1-pole lowpass with a max cutoff of 1.6 kHz or thereabouts, useful for both audio and control signals.