Diversion - new VST synth released!

[bronxsound]

i've been watching this topic for a while. also came across on positive feedback about the synth. appreciating the fact it is packed with feature it seems to load CPU heavily. i'm wondering if the dev could consider trimmed down version as an option. i suspect it might be not feasable to do so due to existing architecture of the code (fx section has been already separated hence the idea). but if it's not a major issue it could be a nice addition to the range allowing wider range of customers to aquire it.

i'm not moaning it's just a thought

[johnnyvn]

Yes, also waiting for the Mac version. Can't wait!

[zorniko]

i've been watching this topic for a while. also came across on positive feedback about the synth. appreciating the fact it is packed with feature it seems to load CPU heavily. i'm wondering if the dev could consider trimmed down version as an option. i suspect it might be not feasable to do so due to existing architecture of the code (fx section has been already separated hence the idea). but if it's not a major issue it could be a nice addition to the range allowing wider range of customers to aquire it.

i'm not moaning it's just a thought

I agree.
Now, the KVR people like to say that sound is what matters, not the CPU use.
BUT
With some patches my CPU goes up to 60-70%!
And that's E8400 Core 2 Duo 2x3GHz.
Judging by all the tests, the third gen. (newest) i7, should be 2-2.5 times faster.
Hm... so you can use JUST 2 instances of Diversion with the NEWEST processor?
Insane.
Where is the space for the other instr/fx?
There is no real point to make such VST.
It IS GOOD, but FAR from practical.
Just MO, of course.

[S0lo]

I agree.
Now, the KVR people like to say that sound is what matters, not the CPU use.
BUT
With some patches my CPU goes up to 60-70%!
And that's E8400 Core 2 Duo 2x3GHz.
Judging by all the tests, the third gen. (newest) i7, should be 2-2.5 times faster.
Hm... so you can use JUST 2 instances of Diversion with the NEWEST processor?
Insane.
Where is the space for the other instr/fx?
There is no real point to make such VST.
It IS GOOD, but FAR from practical.
Just MO, of course.

Those CPU usage indicators in the DAWs can be misleading. Better watch the real CPU usage in the task manager.

I have an i5, And I'm able to use 7 Diversion instances playing simultaneously while pressing 4 keys together, CPU approaches 90%.

[bronxsound]

^actually it hardly could be considered as an advantage

[zorniko]

I agree.
Now, the KVR people like to say that sound is what matters, not the CPU use.
BUT
With some patches my CPU goes up to 60-70%!
And that's E8400 Core 2 Duo 2x3GHz.
Judging by all the tests, the third gen. (newest) i7, should be 2-2.5 times faster.
Hm... so you can use JUST 2 instances of Diversion with the NEWEST processor?
Insane.
Where is the space for the other instr/fx?
There is no real point to make such VST.
It IS GOOD, but FAR from practical.
Just MO, of course.

Those CPU usage indicators in the DAWs can be misleading. Better watch the real CPU usage in the task manager.

I have an i5, And I'm able to use 7 Diversion instances playing simultaneously while pressing 4 keys together, CPU approaches 90%.

Well, I said SOME patches... I belive what you say, of course, but it is still strange for me that on my comp it uses ~10% on a simple mono saw without ANY mod.

[sches]

i've been watching this topic for a while. also came across on positive feedback about the synth. appreciating the fact it is packed with feature it seems to load CPU heavily. i'm wondering if the dev could consider trimmed down version as an option. i suspect it might be not feasable to do so due to existing architecture of the code (fx section has been already separated hence the idea). but if it's not a major issue it could be a nice addition to the range allowing wider range of customers to aquire it.

i'm not moaning it's just a thought

I already thought about that and already tried to make tests. I'm afraid it is hard to make a trimmed version from this plugin which will have both great quality and low CPU usage. The reason is not an oversampling or a long processing chain. The major difficulty is an oscillator by itself. The osc uses almost all required CPU power. If I cut it down, it would be not Diversion in terms of quality. So I think better way here is to make a new synth, developed with low CPU usage in mind.

[bronxsound]

^ thanks for that. as stated it was just an idea of a coding ignorant

[S0lo]

I agree.
Now, the KVR people like to say that sound is what matters, not the CPU use.
BUT
With some patches my CPU goes up to 60-70%!
And that's E8400 Core 2 Duo 2x3GHz.
Judging by all the tests, the third gen. (newest) i7, should be 2-2.5 times faster.
Hm... so you can use JUST 2 instances of Diversion with the NEWEST processor?
Insane.
Where is the space for the other instr/fx?
There is no real point to make such VST.
It IS GOOD, but FAR from practical.
Just MO, of course.

Those CPU usage indicators in the DAWs can be misleading. Better watch the real CPU usage in the task manager.

I have an i5, And I'm able to use 7 Diversion instances playing simultaneously while pressing 4 keys together, CPU approaches 90%.

Well, I said SOME patches... I belive what you say, of course, but it is still strange for me that on my comp it uses ~10% on a simple mono saw without ANY mod.

Your right, it uses 9% on mine on a simple saw. But when I turn on all 4 oscillators (on pure saw) CPU only jumps to 13% (not 9x4). And when I check Task manager's CPU it's still 10%!!.

I think the real CPU eaters in Diversion are the OSCs, Unison and Oversampling.

[sches]

^ thanks for that. as stated it was just an idea of a coding ignorant

You are welcome, it's a nice idea, I tried realize it two times, but no luck

Your right, it uses 9% on mine on a simple saw. But when I turn on all 4 oscillators (on pure saw) CPU only jumps to 13% (not 9x4). And when I check Task manager's CPU it's still 10%!!.

I think the real CPU eaters in Diversion are the OSCs, Unison and Oversampling.

Absolutely true, those 9% is the requirement for starting the whole polyphonic engine (the first voice), once it started playing each new voice consumes much less CPU. Such behaviour can be observed for almost all CPU-heavy synths.

[zerocrossing]

I agree.
Now, the KVR people like to say that sound is what matters, not the CPU use.
BUT
With some patches my CPU goes up to 60-70%!
And that's E8400 Core 2 Duo 2x3GHz.
Judging by all the tests, the third gen. (newest) i7, should be 2-2.5 times faster.
Hm... so you can use JUST 2 instances of Diversion with the NEWEST processor?
Insane.
Where is the space for the other instr/fx?
There is no real point to make such VST.
It IS GOOD, but FAR from practical.
Just MO, of course.

Those CPU usage indicators in the DAWs can be misleading. Better watch the real CPU usage in the task manager.

I have an i5, And I'm able to use 7 Diversion instances playing simultaneously while pressing 4 keys together, CPU approaches 90%.

Well, I said SOME patches... I belive what you say, of course, but it is still strange for me that on my comp it uses ~10% on a simple mono saw without ANY mod.

I think the demo tracks sound really good and from what I've experienced with other synths, the CPU usually goes crazy when doing unison or high resonance sounds... both very "big" types of sound that you probably don't want a lot of anyway.

[sches]

Yes Zerocrossing, the same is here, unison and long release makes CPU crazy. Thanks, I'm glad you like the demo.

Hm... so you can use JUST 2 instances of Diversion with the NEWEST processor?
Insane.

Not true at all, up to 10-15 instances on i7. Of course if you won't load every instance with 4xOSC @ 8x Unison patches, not sure if there are full-unison synths that able to deal with such bunch of patches.

[zorniko]

Obviously, the main problem here is the osc. You've said that it gives quality to Diversion and I agree to that, but you must find a way to re-program this essential part to use much less CPU. For e.g. Synth1's osc sounds ~the same as Diversion's using 8x oversampling (to be honest Diversion will had to have 16x, and a little fatter osc to be compared to Synth1, sorry), at 0.5% CPU cost. So, there MUST be a way (SSE,2,3 or something...)to make this quality thing using less CPU. Then, multiplying osc wouldn't cost so much, etc... you know what I mean.

Keep the good work

[Crackbaby]

I may be wrong here, but i think that what is taking all the CPU is what lets you modify it with all those knobs and x/y. The distortion is polyphonic also and so is the bus as default (filter, distortion, bitcrusher etc).

My host (maschine) has a lousy cpu meter so i haven't been checking this out regularly - How much does the modulation of various parameters cost in terms of cpu? Will having a 1/32 lfo controlling Chip take a lot more cpu than just having the chip on without any lfo?

I hope it doesn't come off as if i am defending high cpu usage.. It's just that when most of the time i use Diversion, it will be monophonic (though three or more pages of modulation) and my host begs to sample and be resampled

Dmitry! Feature request: Chord memory

[sches]

you know what I mean

Zorniko, I really don't understand what you mean. I don't understand why you are trying to compare Diversion and Synth1, I don't understand your quality comparison, and what you want from me. Sorry man…

Björn, modulations don't require lots of CPU, even fully filled modmatrix doesn't eat too much. I'll think about Chord memory, interesting feature, thanks for it.