Diva 1.1 Final (beta testing finished)

[Urs]

What's the likelihood that the multicore optimisations in this release (which RULE btw) will be back-ported to ACE or other synths of yours?

Highly likely for ACE, lesser so for synths with ModMatrices. Until we rewrite ModMatrices...

[Urs]

Also, is there a way to force all Diva instances to enable multi-threading by default?

Not yet, but once we gather more experience with real life usage of the multicore stuff, I think we might get there

[Urs]

I'm getting dropped notes and sometimes stuck notes with 1.1 in FLstudio 10.0.9. This did not happen using 1.01. Anyone else having this issue?

Can you send us a project file or detailed instructions to reproduce this?

[standalone]

Finally play as many voices as you like on selected systems! This needs multicore system with on-cpu-memory-interface like i5, i7, Nehalem.

And also works perfectly in my now humble Q6600, and now I can play chords in divine mode almost without issues. The four graphics (one for each core) in the Windows task manager are practically identical when I play Diva in MiniHost, it's the first time that I see something like this!

[snigelx]

Browsing the manual top to bottom again today. ran over this funny little description
Uhe team is...
"• Howard Scarr (sounds, words, unused ideas)"



Surely we were supposed to browser right over that but I'm so into the synthn I am micro-analyzing the entire contents of the manual this go.

[pdxindy]

1.1 is working really well here... no problems at all... rock solid! and sounding better than ever...

My latest preset...

http://draigathar.org/sounds/Diva47.mp3

[Ingonator]

Hi,

sorry i chime in quite late but i am working on sounds for other synths currently.
I just installed and checked the new Diva 1.1 and checked the CPU use.

My system specifications are:
Core 2 Quad Q6600 CPU (4 x 2.4 GHz), Windows 7 32-bit SP1, 4GB RAM, Foczsrite Saffire Pro 24 DSP, Ableton Live 8

Here are the result for a 3 notes chord (played slowly up to very fast) for my own "IW Poly6 Strings" preset:

1.) Diva 517 (v1.01):
- Great accuracy: 42-78%
- Divine accuracy: not possible (100+ %)

2.) Diva 657 (v1.1) with "multi-threaded" disabled:
- Great accuracy: 30-60%
- Divine accuracy: 80-95+ %, audio dropout all the time

3.) Diva 657 (v1.1) with "multi-threaded" enabled:
- Great accuracy: 23-40%, random audio-dropouts
- Divine accuracy: 34-58+ %, random audio dropouts, when playing fast audio dropouts most of the time

Conclusion: There are obvious improvements in the CPU use but i still got problems using Divine mode with polyphonic sounds. With "multi-threaded" enabled i get random audio-dropouts even with "Great" accuracy.
On my system (Core 2 Quad Q6600 CPU) the best combination currently seems to be using "Great" or "Fast" accuracy with "multi-threaded" turned off.


Ingo

[Howard]

So: Diva's CPU-hunger appears to vary considerably between superficially very similar computer systems. Assuming this is true, it would be great to find out exactly what is choking Diva performance in e.g. Ingo's system...
Any ideas?

[aMUSEd]

Also, is there a way to force all Diva instances to enable multi-threading by default?

If your host has the ability to save track templates save a track template with Diva inserted in the mode you want and it will retain that setting in future tracks made with the template.

[hakey]

Q6600, Win XP, Live 8, Diva 1.1 (rev 657)

Using Ingo's IW Poly6 Strings patch, polyphony set to 16 notes.

• Multi-threaded Off
  
  Great: 3 note chord held = 28%, audio dropouts when holding 9 notes
  Divine: 3 note chord held = 37 to 39%, audio dropouts when holding 7 notes
• Multi-threaded On
  
  Great: 3 note chord held = 19%, no audio dropouts
  Divine: 3 note chord held = 23%, audio dropouts when holding 16 notes

(IMO Ingo your 3 chord "playing slow to fast" isn't so useful. Holding a chord and then adding notes until audio dropouts occur gives a better idea of max polyphony for a given patch.)

Generally speaking, cpu is no longer a significant issue, even on an aged machine like man.

-

[hakey]

what is choking Diva performance in e.g. Ingo's system...
Any ideas?

Windows XP vs 7? Soundcard audio buffer settings?

[Ingonator]

Q6600, Live 8, Win XP

Using Ingo's IW Poly6 Strings patch, polyphony set to 16 notes.

• Multi-threaded Off
  
  Great: 3 note chord held = 28%, audio dropouts when holding 9 notes
  Divine: 3 note chord held = 37 to 39%, audio dropouts when holding 7 notes
• Multi-threaded On
  
  Great: 3 note chord held = 19%, no audio dropouts
  Divine: 3 note chord held = 23%, audio dropouts when holding 16 notes

(IMO Ingo your 3 chord "playing slow to fast" isn't so useful. Holding a chord and then adding notes until audio dropouts occur gives a better idea of max polyphony for a given patch.)

Generally speaking, cpu is no longer a significant issue, even on an aged machine like man.

Hi,

while playing in a "real" song the speed of playing notes IS a significant issue. I could not always play as slow a possible just because i have to fear that the CPU could max out. I guess that many people seem to miss this aspect when doing CPU tests.
The lowest values in my test are when just pressing keys and holding the current notes.

It also looks like with my "old" Core 2 Quad Q6600 CPU i could not benefit from all of the improvements of the new update. Like Urs mentioned the most benefits are at the newer i5 or i7 CPUs.

BTW while the CPU measurements with the "multi-threaded" switch on show a 50% improvement in CPU use i get a few or many audio dropouts depending on the playing speed of the notes. So theoretically it is better but as soon as i get audio dropouts it doesn't really matter in a practical sense.


Ingo

[Howard]

while playing in a "real" song the speed of playing notes IS a significant issue. I could not always play as slow a possible just because i have to fear that the CPU could max out. I guess that many people seem to miss this aspect when doing CPU tests.

I don't think they miss that point, Ingo. It's about overlapping notes. If necessary, change the max. polyphony for the "offending" patch and resave.

[hakey]

while playing in a "real" song the speed of playing notes IS a significant issue.

Of course. But all that's happening is that you're temporarily reaching max polyphony as voices overlap - max polyphony is the limiting factor.

And your test doesn't tell us what max polyphony is for your machine (we don't know how fast, or in what style you're playing) and so the results you've reported aren't all that useful for direct comparison.

OTOH, holding a chord and then adding notes until dropouts are heard is a test that can easily be exactly replicated and so allow for direct comparisons.

[Ingonator]

while playing in a "real" song the speed of playing notes IS a significant issue. I could not always play as slow a possible just because i have to fear that the CPU could max out. I guess that many people seem to miss this aspect when doing CPU tests.

I don't think they miss that point, Ingo. It's about overlapping notes. If necessary, change the max. polyphony for the "offending" patch and resave.

Hi Howard,

i just checked with the same preset in Divine mode and with "multi-threded" turned on. There is an obvious improvement as before i was not even able to use any polyhonic sounds in Divine mode. The polyhony of that preset originally was set to 6 and there is no voice stack.
Even if the CPU stays at a reasonable amount from time to time i get audio drop outs but i am not able to reproduce this at any time. If i play that preset for around 30-60 seconds with different speeds (speed with which the notes are hit) i could reproduce this at least once at that period.
When i reduce the polyphony from 6 to 5 and play 5 notes at once i also get some note stealing from time to time.

Anyway there is some serious improvement so i don't really complain and if i get a newer CPU (e.g. i7) there should not be a problem at all.


Ingo