KVR Audio

Ingonator · Post by **Ingonator** » Mon Sep 26, 2016 12:31 pm

chk071 wrote:Yes, but for a real life test, you can just compare how the plugins perform, when setting equal conditions for each plugin, and see how the host can handle them. I don't think the mathematical principle behind it is very relevant for the performance on the user's computer.

As i just mentioned above the biggest CPU improvent in The Legend is foudn with more tahn one voice used (e.g. the Poly or Unison mode). Richard had explained this with a "vectorized core" where mutiple voices are calculated (that term is even mentioend at the offcial website for The Legend and also taht thios saves CPU with multipl voices palyed at once).

So to do a fair test you would have comparae all 3 synth using 4 voice polyphony. I know how to use poylphony with Monark and 4 or 5 voices seem to be the maximum it is capable of before you get audio dropouts with a single instance (at least with my older Core 2 Quad Q6600 CPU).

zvenx · Post by **zvenx** » Mon Sep 26, 2016 12:32 pm

fluffy_little_something wrote:On the SA site it says for instance that The Legend is 8x oversampled. No idea about the other two...

ah answered.

rsp

Ingonator · Post by **Ingonator** » Mon Sep 26, 2016 12:34 pm

zvenx wrote:
Urs wrote:
chk071 wrote:To do a fair comparison, use the "divine" option, which is the default option on the other 2 plugins.
To do a fair comparison we need to know on what mathematical principle Diva, Legend and Monark compute their best quality setting. But we don't (I know for Diva, but that doesn't help here, does it?)
True.
Either way I just did and did 3 screenshots.

Diva 8, 10, 10
Legend 14, 14, 14
Reaktor 12, 15, 13

with Divine and Multicore on.
next will you ask me to take off multicore

Like I said this isn't a scientific test but for me clearly at the very least strongly suggests that compared to these two, Diva isn't a cpu hog... quite the opposite..
Btw how did you know the oversample rate of Legend?

rsp

actually you SHOULD take off multcore for the comparison as the others do not seem to use it too.

BTW with multcore the overall CPU use (for all cores) seems to stay comparable, just the load of a single core is reduced.

zvenx · Post by **zvenx** » Mon Sep 26, 2016 12:34 pm

Ingonator wrote:
chk071 wrote:Yes, but for a real life test, you can just compare how the plugins perform, when setting equal conditions for each plugin, and see how the host can handle them. I don't think the mathematical principle behind it is very relevant for the performance on the user's computer.
As i just mentioned above the biggest CPU improvent in The Legend is foudn with more tahn one voice used (e.g. the Poly or Unison mode). Richard had explained this with a "vectorized core" where mutiple voices are calculated (that term is even mentioend at the offcial website for The Legend and also taht thios saves CPU with multipl voices palyed at once).

So to do a fair test you would have comparae all 3 synth using 4 voice polyphony. I know how to use poylphony with Monark and 4 or 5 voices seem to be the maximum it is capable of before you get audio dropouts with a single instance (at least with my older Core 2 Quad Q6600 CPU).

How is that a fair test?
1. Monark does monopoly without the hack.
2. So does a minimoog...which we are comparing all too.
rsp

zvenx · Post by **zvenx** » Mon Sep 26, 2016 12:35 pm

Don't be ridiculous now.

That's like saying comparing a Mercedes Benz vs a Yugo in speed, you can't press the MB accelerator all the way down.

come on now.
let's not get ridiculous.
This is a comparison in normal use.
why not ask urs to then make a special Diva devoid of optimisations so we can compare them again.

rsp

Ingonator wrote:
zvenx wrote:
Urs wrote:
chk071 wrote:To do a fair comparison, use the "divine" option, which is the default option on the other 2 plugins.
To do a fair comparison we need to know on what mathematical principle Diva, Legend and Monark compute their best quality setting. But we don't (I know for Diva, but that doesn't help here, does it?)
True.
Either way I just did and did 3 screenshots.

Diva 8, 10, 10
Legend 14, 14, 14
Reaktor 12, 15, 13

with Divine and Multicore on.
next will you ask me to take off multicore

Like I said this isn't a scientific test but for me clearly at the very least strongly suggests that compared to these two, Diva isn't a cpu hog... quite the opposite..
Btw how did you know the oversample rate of Legend?

rsp
actually you SHOULD take off multcore for the comparison as the others do not seem to use it too.

BTW with multcore the overall CPU use (for all cores) seems to stay comparable, just the load of a single core is reduced.

zvenx · Post by **zvenx** » Mon Sep 26, 2016 12:38 pm

I know you fell out of Love with Diva for whatever reason, (Saurus thread), but loving the Legend doesn't mean you have to be so biased against Diva.
I have all three and will use all three. But surely in a minimoog test Diva is NOT a cpu hog compared to the others in everyday use.

rsp

Ingonator · Post by **Ingonator** » Mon Sep 26, 2016 12:43 pm

zvenx wrote:I know you fell out of Love with Diva for whatever reason, (Saurus thread), but loving the Legend doesn't mean you have to be so biased against Diva.
I have all three and will use all three. But surely in a minimoog test Diva is NOT a cpu hog compared to the others in everyday use.

rsp

I am not biased against Diva in any way or "fell out of love" but as a proper Minimoog emulation and in terms of CPU use i prefer The Legend.

zvenx · Post by **zvenx** » Mon Sep 26, 2016 12:52 pm

Ingonator wrote:
zvenx wrote:I know you fell out of Love with Diva for whatever reason, (Saurus thread), but loving the Legend doesn't mean you have to be so biased against Diva.
I have all three and will use all three. But surely in a minimoog test Diva is NOT a cpu hog compared to the others in everyday use.

rsp
I am not biased against Diva in any way or "fell out of love" but as a proper Minimoog emulation and in terms of CPU use i prefer The Legend.

That's fine and I agree with you as a minimoog emulation specifically, I too prefer Legend over Diva. But do you honestly think logically and reasonably your comments about taking off the multicore were sound?

I know you did sound design for Diva, why not fire it up and do a cpu comparison yourself.
rsp

Ingonator · Post by **Ingonator** » Mon Sep 26, 2016 12:57 pm

I just compared the same patch ("3-Octave Lead") in The Legend and Diva (using "Divine" quality and multecore set off).
With a single note (at Mono mode) The Legend shows 13% in the display of Live 9 while Diva shows 15%.

When switching to 4 voice poly with holding 4 notes The Legend uses 21% while Diva uses 55%.

When adding longer release and playing faster notes (with overlapping notes) The Legend reaches a maximum of around 23% while Diva gets unplayable this way. In the other tests Release was off and for this test i switched the envelopes in Diva to ADSR envelopes (from Jupiter

. With a short Release Diva gets playbale again.

As mentioned earlier i was able to play 32 voices at once (in 8 instances) with The Legend without having audio dropouts (even wit husing a longer Release).

That's due to the advantage of the "vectorized core" i just mentioned and taht is also mentioend at teh official website for The Legend.

chk071 · Post by **chk071** » Mon Sep 26, 2016 1:10 pm

This is how it looks for me, with a fairly simple two sawtooth oscillator patch for Legend and Monark (both set to mono):

Bit difficult to compare though, as Reaktor seems to always process the sound, even when it's idling, and both vary + - 4 % of CPU in the Studio One meter.

zvenx · Post by **zvenx** » Mon Sep 26, 2016 1:14 pm

Using Ingo's 3 osc lead test on Divine but yes with Multicore on cause its there to reduce cpu hits.

in Mono I did three screen shots

Diva 3, 16, 7
Legend 11,16, 12

with chords:

Diva 3, 6, 16
Legend 12, 13, 19

I didn't do the increase sustain/release test you did.

rsp

fluffy_little_something · Mon Sep 26, 2016 1:23 pm

Finally finished work and downloaded the demo. Didn't like the fact that I had to install the 32-bit version as well. Why not a checkbox?

Anyway, the sound is excellent, even on my headphones, which tend to analyze the shit out of most synths, there is still that direct, solid bass foundation, nice

It runs without any issues on my old AMD processor, even in unison mode and with reverb on.

Ingonator · Post by **Ingonator** » Mon Sep 26, 2016 1:27 pm

To avoid using Polyphony for the test i just duplicated the same monophonic patch in The Legend 23 times to have 24 instances of the same patch. On my old Core 2 Quad Q6600 CPU i get a CPU display of 75% inside Live 9 64-bit. So this correspnds to playing 24 voices at once.

With the Poly mode i was able to get a comparable CPU use with 8 instances and 4 voices each resulting in 32 voices at once.

For me those are more than enough voices and with a more modern CPU you should get the double amount or more.

chk071 · Post by **chk071** » Mon Sep 26, 2016 1:40 pm

zvenx wrote:Using Ingo's 3 osc lead test on Divine but yes with Multicore on cause its there to reduce cpu hits.

in Mono I did three screen shots

Diva 3, 16, 7
Legend 11,16, 12

with chords:

Diva 3, 6, 16
Legend 12, 13, 19

I didn't do the increase sustain/release test you did.

rsp

Does your PC have 4 cores actually?

zvenx · Post by **zvenx** » Mon Sep 26, 2016 1:46 pm

For me Both Legend and Diva sound fantastic, with Legend indeed to my ears sounding more minimoog-like which is to be expected from their Design and goal.
I just believe that this Diva is a unusual cpu hog is a fallacy
Rsp

Synapse Audio Minimoog emulation "The Legend" for VST/AU and RE released!