Largo is here!

[MarcTale]

@SolarRainUK:

Thats right but still not really in sync.
I don't want to search for the correct syncing delay even thogh I already set the timing on 1/4. The closest is 36 or 37 but still...

[SolarRainUK]

A problem is that he set the voices to 48. At 3 voices unison this means you can play 16 notes. If you decrease the voices to 12 you can still play 4 notes and the CPU load is much less. I tried this myself.

Even if you set unison to 6 and lower voices to 24 the CPU load is lower than the original patch. The reverb also increases the load heavily so in the worst case an external reverb should be used.

I set the voices to 42, anyway I only made that pad as an experiment, it's still only a single layer and only 1 FX (reverb, which btw when I turn off seems to make very little difference on CPU for me), just proves the point that Largo is quite CPU hungry, at least when it comes to epic pads like this.

Sure, you can reduce the voices. But in the case of my example pad, if you reduce to 12, when you change chords you'll hear the notes cut off abruptly. Though 24 voices seems a good compromise in this case, but that still puts me at 50% CPU.

Anyway, that's not so good..but I'm going to experiment on this pad a little more, by using extra layers but without Unison, to see if I can get a more efficient version without sacrificing the sound too much.

[Ingonator]

A problem is that he set the voices to 48. At 3 voices unison this means you can play 16 notes. If you decrease the voices to 12 you can still play 4 notes and the CPU load is much less. I tried this myself.

Even if you set unison to 6 and lower voices to 24 the CPU load is lower than the original patch. The reverb also increases the load heavily so in the worst case an external reverb should be used.

I set the voices to 42, anyway I only made that pad as an experiment, it's still only a single layer and only 1 FX (reverb, which btw when I turn off seems to make very little difference on CPU for me), just proves the point that Largo is quite CPU hungry, at least when it comes to epic pads like this.

Sure, you can reduce the voices. But in the case of my example pad, if you reduce to 12, when you change chords you'll hear the notes cut off abruptly. Though 24 voices seems a good compromise in this case, but that still puts me at 50% CPU.

Anyway, that's not so good..but I'm going to experiment on this pad a little more, by using extra layers but without Unison, to see if I can get a more efficient version without sacrificing the sound too much.

I made a patch using 4 layers with all 5 OSCs (3 + SUB) per layer. This simulates a hyperasw/unison. the reverb seems to be quite CPU consuming but i left it in.

link: http://www.kvraudio.com/banks.php?s=lis ... order=date ; the files is called "20 OSC Strings 1 IW"

UPDATE: Strange thing: I noticed that the CPU usage in Live stays at a certain level even if i play no notes. When i switch of the reverb in the layers one by one the CPU load goes down until it's normal. Means that the reverb gives a permanent CPU load which increases with the amount of layers you put it in.

Here's the message i just wrote to the Waldorf forum:

"Hello,

today i made a patch in Largo which contains a layer of 4 sounds. I noticed when i switch effects on Ableton Live shows a permanent CPU load which increases with each layer in which an effect is activated. This happens without playing any notes and disappears when switching the effects of the layers off.

With 4 layers using 1 effect (e.g. Reverb) for each layer Largo shows appr. 17-20% CPU usage on a Core 2 Quad Q6600 CPU.

Ingo"

[koalaboy]

I made a patch using 4 layers with all 5 OSCs (3 + SUB) per layer. This simulates a hyperasw/unison. the reverb seems to be quite CPU consuming but i left it in.

link: http://www.kvraudio.com/banks.php?s=lis ... order=date ; the files is called "20 OSC Strings 1 IW"

UPDATE: Strange thing: I noticed that the CPU usage in Live stays at a certain level even if i play no notes. When i switch of the reverb in the layers one by one the CPU load goes down until it's normal. Means that the reverb gives a permanent CPU load which increases with the amount of layers you put it in.

I get about 15% usage on a 7 note chord with this one, but a permanent 5% with nothing being played, as you noted. I also tried turning the chorus on and flicking to the phaser at one point sent everything into self-oscillation/resonation ?!?! Odd whistling sounds, which I quickly stopped, and then couldn't reproduce

A really nice sound though. It hits 70% if I up the voices to 144, putting every part into unison-6, but doesn't really improve anything.

[Ingonator]

I made a patch using 4 layers with all 5 OSCs (3 + SUB) per layer. This simulates a hyperasw/unison. the reverb seems to be quite CPU consuming but i left it in.

link: http://www.kvraudio.com/banks.php?s=lis ... order=date ; the files is called "20 OSC Strings 1 IW"

UPDATE: Strange thing: I noticed that the CPU usage in Live stays at a certain level even if i play no notes. When i switch of the reverb in the layers one by one the CPU load goes down until it's normal. Means that the reverb gives a permanent CPU load which increases with the amount of layers you put it in.

I get about 15% usage on a 7 note chord with this one, but a permanent 5% with nothing being played, as you noted. I also tried turning the chorus on and flicking to the phaser at one point sent everything into self-oscillation/resonation ?!?! Odd whistling sounds, which I quickly stopped, and then couldn't reproduce

A really nice sound though. It hits 70% if I up the voices to 144, putting every part into unison-6, but doesn't really improve anything.

It doesn't improve much with unison because the OSCs in all 4 layers were detuned to simulate a hypersaw without using unison. With 24 overall voices you could play 6 notes with this. Much of the CPU load corresponds to the reverb effect in the 4 layers.

[koalaboy]

It doesn't improve much with unison because the OSCs in all 4 layers were detuned to simulate a hypersaw without using unison. With 24 overall voices you could play 6 notes with this. Much of the CPU load corresponds to the reverb effect in the 4 layers.

I'm only just starting to learn about actual sound design, so I generally like to turn up everything

(and I've just realised you're the man responsible for a load of patches for Tassman 4. I just bought that yesterday - this month has been stunning for my sound capabilities )

[Ingonator]

It doesn't improve much with unison because the OSCs in all 4 layers were detuned to simulate a hypersaw without using unison. With 24 overall voices you could play 6 notes with this. Much of the CPU load corresponds to the reverb effect in the 4 layers.

I'm only just starting to learn about actual sound design, so I generally like to turn up everything

(and I've just realised you're the man responsible for a load of patches for Tassman 4. I just bought that yesterday - this month has been stunning for my sound capabilities )

AFIK i was the last one to publish a synth for the Tassman 4 database, called "Evolving Lead V1.0". that was in march 2008... Nothing new so far, except himalayas commercial soundbank. Maybe you found my HyperOSC synths for Tassman 4 which recreate hypersaw sounds. Fits well to this thread now.

[koalaboy]

It doesn't improve much with unison because the OSCs in all 4 layers were detuned to simulate a hypersaw without using unison. With 24 overall voices you could play 6 notes with this. Much of the CPU load corresponds to the reverb effect in the 4 layers.

I'm only just starting to learn about actual sound design, so I generally like to turn up everything

(and I've just realised you're the man responsible for a load of patches for Tassman 4. I just bought that yesterday - this month has been stunning for my sound capabilities )

AFIK i was the last one to do a synth for Tassman 4 called "Evolving Lead V1.0". that was in march 2008... Maybe you found my HyperOSC synths for Tassman 4 which recreate hypersaw sounds. Fits well to this thread now.

The HyperOSC synths were exactly what I just found after searching from this thread I'd been wanting Tassman for ages, as I love the sound (and it's cross-platform unlike Vaz) - What with that and Komplete, Largo, Minimonsta, and some great UAD plugs, this month has been one of the most virtually analog in a long time.

[Ingonator]

I made a patch using 4 layers with all 5 OSCs (3 + SUB) per layer. This simulates a hyperasw/unison. the reverb seems to be quite CPU consuming but i left it in.

link: http://www.kvraudio.com/banks.php?s=lis ... order=date ; the files is called "20 OSC Strings 1 IW"

UPDATE: Strange thing: I noticed that the CPU usage in Live stays at a certain level even if i play no notes. When i switch of the reverb in the layers one by one the CPU load goes down until it's normal. Means that the reverb gives a permanent CPU load which increases with the amount of layers you put it in.

I get about 15% usage on a 7 note chord with this one, but a permanent 5% with nothing being played, as you noted. I also tried turning the chorus on and flicking to the phaser at one point sent everything into self-oscillation/resonation ?!?! Odd whistling sounds, which I quickly stopped, and then couldn't reproduce

A really nice sound though. It hits 70% if I up the voices to 144, putting every part into unison-6, but doesn't really improve anything.

I found out that those hypersaw type sounds of Largo sound awesome when using external effects. I use the free Kjaerhus Classic effects bundle, especially the Classic Chorus (preset "String Pad Chorus"), Classic Phaser (preset "Synth Pad #1") and/or Classic Reverb. Quite low CPU usage when not using Largos own effects and the Kjaerhus effects just sound better.

links:
http://www.kjaerhusaudio.com/download.php (-> UNDER FREE PLUG-INS)
http://www.kvraudio.com/get/648.html
http://www.kvraudio.com/get/926.html
http://www.kvraudio.com/get/717.html

[Igro]

I made a patch using 4 layers with all 5 OSCs (3 + SUB) per layer. This simulates a hyperasw/unison. the reverb seems to be quite CPU consuming but i left it in.

link: http://www.kvraudio.com/banks.php?s=lis ... order=date ; the files is called "20 OSC Strings 1 IW"

UPDATE: Strange thing: I noticed that the CPU usage in Live stays at a certain level even if i play no notes. When i switch of the reverb in the layers one by one the CPU load goes down until it's normal. Means that the reverb gives a permanent CPU load which increases with the amount of layers you put it in.

I get about 15% usage on a 7 note chord with this one, but a permanent 5% with nothing being played, as you noted. I also tried turning the chorus on and flicking to the phaser at one point sent everything into self-oscillation/resonation ?!?! Odd whistling sounds, which I quickly stopped, and then couldn't reproduce

A really nice sound though. It hits 70% if I up the voices to 144, putting every part into unison-6, but doesn't really improve anything.

I found out that those hypersaw type sounds of Largo sound awesome when using external effects. I use the free Kjaerhus Classic effects bundle, especially the Classic Chorus, Classic Phaser and/or Classic Reverb. Quite low CPU usage when not using Largos own effects and the Kjaerhus effects just sound better.

links:
http://www.kjaerhusaudio.com/download.php (-> UNDER FREE PLUG-INS)
http://www.kvraudio.com/get/648.html
http://www.kvraudio.com/get/926.html
http://www.kvraudio.com/get/717.html

So from your long discussion you came to conclusion, that effects give so much CPU usage. This is better news, than if it was something before effects.

[Ingonator]

I made a patch using 4 layers with all 5 OSCs (3 + SUB) per layer. This simulates a hyperasw/unison. the reverb seems to be quite CPU consuming but i left it in.

link: http://www.kvraudio.com/banks.php?s=lis ... order=date ; the files is called "20 OSC Strings 1 IW"

UPDATE: Strange thing: I noticed that the CPU usage in Live stays at a certain level even if i play no notes. When i switch of the reverb in the layers one by one the CPU load goes down until it's normal. Means that the reverb gives a permanent CPU load which increases with the amount of layers you put it in.

I get about 15% usage on a 7 note chord with this one, but a permanent 5% with nothing being played, as you noted. I also tried turning the chorus on and flicking to the phaser at one point sent everything into self-oscillation/resonation ?!?! Odd whistling sounds, which I quickly stopped, and then couldn't reproduce

A really nice sound though. It hits 70% if I up the voices to 144, putting every part into unison-6, but doesn't really improve anything.

I found out that those hypersaw type sounds of Largo sound awesome when using external effects. I use the free Kjaerhus Classic effects bundle, especially the Classic Chorus, Classic Phaser and/or Classic Reverb. Quite low CPU usage when not using Largos own effects and the Kjaerhus effects just sound better.

links:
http://www.kjaerhusaudio.com/download.php (-> UNDER FREE PLUG-INS)
http://www.kvraudio.com/get/648.html
http://www.kvraudio.com/get/926.html
http://www.kvraudio.com/get/717.html

So from your long discussion you came to conclusion, that effects give so much CPU usage. This is better news, than if it was something before effects.

Basically it's four things that max out the CPU:
1.) number of overall voices setting (1-256) -> standard setting: 16, should be as low as possible. With 6 voice unison you need 24 voices for 4 notes of polyphony.
2.) unison mode (2-6 voices) -> related to 1.) -> more unison voices lead to more overall voices needed !!
3.) the FX, especially the Reverb
4.) a combination of 1.) to 3.)

[SolarRainUK]

So from your long discussion you came to conclusion, that effects give so much CPU usage. This is better news, than if it was something before effects.

That's not my conclusion, I've tested this so many times today..It is NOT the effects causing such high CPU usage, I proved this with the single layer pad I created. Turning on/off reverb has negligible CPU impact, same as the other effects. I'd even go so far as to say (at least as far as the reverb is concerned), that it's actually very efficient given how good the new reverb sounds.

BUT..stacking multiple instances of the reverb on top of each other (or any effects for that matter) in a multi-layered patch WILL hit the CPU quite a bit. But that's no surprise.

What Largo really needs is a Main Effects output, so that you can route all layers through, say, a single reverb at the end of the chain. That would fix the multi-layered effects problem using too much CPU.

[MarcTale]

What Largo really needs is a Main Effects output, so that you can route all layers through, say, a single reverb at the end of the chain. That would fix the multi-layered effects problem using too much CPU.

That's what I just wrote here this afternoon...
I'm really surprised they didn't do it.

[SolarRainUK]

Me too. To be honest the effects section wasn't too well thought out for multi-layered patches. Delay and Reverb should both have their own main effects section, with serial and parallel routing, with the ability to bypass either/or.

[Igro]

What Largo really needs is a Main Effects output, so that you can route all layers through, say, a single reverb at the end of the chain. That would fix the multi-layered effects problem using too much CPU.

That's what I just wrote here this afternoon...
I'm really surprised they didn't do it.

So i think it definitely will in new version.