I find it weird but I'm getting okay performance with stack oscillators on my 2007 duo core 2.4 Ghz Imac.

So it's not only that the voices are stacked. Some patches that don't use stacking make the the system chock too. So it's more complex than just stacking.

I'm probably going to go with the iMac that comes out next, and wait to do the mac mini networking until their individual core speeds can be bumped up to 2.4 Ghz. That might not be this next update. The problem with the mac minis are the internal PSU puts limitations on the processor speed. They wont be going with an external as that would break with the concept..

Wade

Apart from quality setting Diva needs more cpu the higher the filter resonance, the higher the filter frequency and the more random the input. And the louder the input.

So a filter sweep can go from very cpu draining to piece of cake. Hence one can't always pinpoint patches or settings as one or the other.

We've been pondering about these issues and I think we've found some ways to tame that stuff. We might start seeding betas soon.

Diva is about 50/50 right now on a i7 2Ghz quad core. Sorta like what Urs said, it's "okay-ish". If you're using Logic, yeah, you could setup a network, attach computers fast enough to run Diva on a core, then connect those as nodes in Logic. I've not done it, but it sounds reasonable.

I don't do music full time. It's a hobby, and the Mac Mini Server is great, but as a server, or for apps that multithread like Photoshop etc. For music, I think it'll run pretty much everything no problem, but the fringe stuff like Diva, ACE, and other potentially intense plugins that'll only run on a single core, you really need the more heavy hitters in the over 2.5GHz range.

I don't really mind that Diva is capable of more than what my system can handle. Just as long as it sounds good and can do most things I want to with it (which I'm hoping it will with the forthcoming update). Eventually I'll get a new system and new doors will be opened. It's kinda like buying a simple VCO/VCF/VCA modular synth then adding to it later.
Last edited by quixotic on Sat Mar 03, 2012 2:47 pm; edited 1 time in total

Personally and for basses only, I didn't notice much difference between the draft and divine modes.

Great! The best gets better!


Interesting!

Btw. we've been experimenting with multicore support. It's possible. It's a less than suboptimal technology for plugins, but we might be getting it far enough to improve live usage. Needs more work though.

i find the default 'great' a little CPU heavy - i can drop to 'fine' - still having 0-delay filters and my CPU is ok, but i use 'draft' for real-time because i am not playing live..

You're absolutely right, I A/B'd the quality settings on several patches and found the same thing... But try a patch like HS Kit 1 acoustic, play C3 and above. The difference is night and day.

In general you'll hear the biggest difference on presets with high filter resonance/emphasis and on patches with quick modulation of the filter cutoff. Although more subtle it's also there with respect to osc feedback too.

Since a lot of the bass patches use almost no filter resonance you won't see much of a cpu hit.

Yes, agreed, I was only referring to basses. My love for DIVA is for its fuzzy, brown MS-20 sounds. I still prefer using Pro-12 for sounds with lots of high frequency content, both because it has a sweeter Prophet 5/Jupiter 8 air on top and because it doesn't cripple my system.

I'm thinking about these issues some more!

I'm ready to upgrade my system. I don't think it's unreasonable that High End plugs like Diva, ACE, Bazille should require a fast up to date computer to run at full quality level.

I Repeat, it is natural that advanced tools require more CPU than more basic tools. So if you want to play these synths, upgrade, upgrade, upgrade! Personally if the Developer is struggling with the limitations of a architecture created by the chip makers, (The multithreading technology was never meant for real time farming out of ONE plug) then I don't know how productive it is trying to bend the envelope of the underlying technology. At some point in the near future this multithreading issue with realtime has to reach a critical mass and a hardware solution will be found, problem solved at the hardware level is what we need.

For now what would be more conductive to the creative flow would be auto CPU load adjust where the system config is read by the plug and the Quality setting is setup AUTOMATICALLY, with a suggested/recommended max number of voices. Further the Number of active instances and their Quality Setting would somehow need to be polled. This would be complicated to setup, time consuming no doubt. But I feel AUTO Quality adjust would be far more useful in the long run, because lests face it CPU resources coming up short is always going to be a problem.

End section on Auto adjust.

Real world workaround! I patch played from four instances on four cores.

Use a host or rack that can place several instances of the Plug in on different cores. Here you could use midi filtering to direct your playing to two three or four Divas playing the same patch, but the notes would be interleaved. My old Korg Wavestation had this ability if you wanted increased polyphony to interlace to wavestations.

This is a practical workaround that can be set up in Numerology for one but likely could be done in other hosts or using midi filtering plugins.

Now you need to play a CPU heavy patch you want 8 note polyphony you set 4 Divas to 2 note polyphony and now your CPU load is spread between four cores. For that monster sound.

Regards

Wade

Agreed. I'm really glad that u-he went no holds barred with DIVA and it's evident in its results.

In my own case, most of my plug-ins are handled by my Scope, UAD, and Powercore cards, so my 4 year old Q6600 processor rarely went above 25% prior to DIVA. It's averaging at 50-75% after DIVA.

I don't like the 75% figure at all, it's to close to break up point on my system. I'd like to invest what it takes to get 25 voices of high quality playback at around 50% loading. Just to keep a buffer for when you want to expand on an idea. My thinking is to use 3 or 4 up todate standard versions of iMac, mac mini top version without upgrade. and Mac book pro 15" top version with non glare screen. The stock versions because the Apple upgrades are very expensive for a small inprovemnet in performance. I'd rather have 4 off the shelf versions than 3 BTO versions which is about the same price level..

I'm figuring this all out carefully and will implement it within the next 18 months or so. It'll be my last great setup before I change gears and just start playing the Frame drum and the penny whistle and ankle bells, on a pan eurasian walk-about.

Wade

I got started in making my monophonic and polyphonic interlace filter in Numerology this morning. This will mean I can take the output from a midi data stream and farm it out to two or four Divas instances playing the same patch.

The first setup for a monophonic synths I setup binary operator and Tenary Operaters that split the midi stream into Odd and Even notes then routed to one of two midi output modules.

Now I could setup Two Divas with the same patch in a Main Stage rack to get my multithreaded patch.

The other setup is polyphonic with two versions one that send to every other note, using To Divas and the other would be if you needed Four cores to manage an especielly knarly patch.Then each Diva gets assigned every fourth note.

I can use the four Divas in 4 voice polyphony mode for 16 voices total.

This has taken about an hour to do here this morning.. I've done a similiar setup for my own Delay Monster Project. So this was not something I had to experiment with to setup.

Now I just need the new iMac when it comes..


Wade
Last edited by TwoToneshuzz on Sun Mar 11, 2012 6:14 am; edited 1 time in total

Unfortunately we've been hilariously busy preparing stuff for Musikmesse so we didn't actually get everything done that we wanted to show. Clemens will give multicore support another shot, but some other stuff came up, so we can't spend the whole week.

That said, we will demonstrate a new method that does tame the overly expensive situations. The speed increase is however not seen at patches where Draft or Fast are good enough. She should however exhibit a more consistent cpu footprint, i.e. performance fluctuations while holding a note should be lesser.