Yes and in NI Kore. Working very well

Sorry, Core2Duo has a memory bottleneck, just like my 2007 Mac Pro. It doesn't do much there...

It's possible. Halion uses it, Harmor does it and now Diva, and maybe a few more that I do not know.

(actually, disc streaming samplers would need to be multi threaded too)

You're going to be lucky with the 12 core The older pre-Nehalem Mac Pros do not benefit much (2 voices perform slower, 10+ voices use up all cores)

Please send those patches to us. Clemens needs to look into that. We're not aware of any changes, but this rises a red flag. We do need to check this out before releasing anything.

I'm sure there's some voodoo Multi-threading is like a factory designed by MC Escher - it makes practically infinite things and it looks like one can use a thousand conveyer belts in all sorts of combinations to do so, but it turns out it's only one conveyer belt with a lot of twists and tricks ... multiple cores is like multiple conveyer belts but they're in different buildings and have to be managed by the guy in the main building. Gets a little chaotic

I wouldn't swear on anything, but - generally there's sort of a global threading environment managed by the OS, communicating in relatively direct way with hardware. An OS will have a 'make thread' procedure that's called by a program; that should be the same inside a VST as a DAW in terms of, what sorts of requests it can make of that environment's resources.

Here you are, sir.
http://www.3amnoise.net/diva_filterdrones.zip
These were made with build 478. In that build, C3 is perfectly in tune and 4 octaves up is 6 cents below C7. Tuning is roughly the same across all filters.
In Diva 1.0.0, C3 is perfectly in tune and C7 is ~13 below. Tuning is roughly the same across all filters.
In Diva 1.0.1, C3 is ~1 cent below and C7 is ~17 below. Tuning is roughly the same across all filters.
In build 585, C3 is ~1 cent below and C7 is ~37 cents below. Tuning is still roughly the same across all filters.

i5-430M Windows 7 here with Ableton Live.
On poly patches I have up to +20% cpu load, all other cases zero difference.
But I had no crashes when switching accuary / OfflineAcc / patches.

using diva 64 bit, a quad core q6700 / win7 64bit / reaper 64 bit - 256 samples ASIO buffers, it works fantastic in the new multithreading mode! i can now play patches fine that without multithreading just dissolve into crackling! no crashes here.

Urs mentioned there is an improvement in the sound of the transients in the new version, anyone picking up on that?

I'm getting more than a 50% performance increase with the new version on my i7. Nice job Urs, Clemons, and co.

Does this multi-threading not apply to Ableton Live? As far as I understood, Live assigns each track to a core and I didn't think there was a way around that for any plug-in. I've tested the new beta and haven't seen any different in cpu on the Live cpu meter but I can't say I've tested thoroughly the system cpu meter. If anything Live may be showing slightly more cpu from DIVA. I'm on a macbook pro i7 from last year. I am seeing however that it loads now the quality setting correctly as is stored in an Ableton rack so that's cool.

Try it on a pad poly type patch, like the wash in the polys area. With 5 notes on the initial hit in divine I can get close to 50% in ableton. With multithreaded that drops to 25, you can flick it on and off and see the difference immediately.

Then again this is also windows x64, not sure if that makes a difference, but its still ableton

I'm also using Live on a 2011 i7 MBP, I'm using the Audio Unit. The difference is most notable for me on patches with stacked voices or when I play chords. Particularly when I play patches with lots of FM and high resonance.

On my MacPro, Diva in Divine mode is performing better than Bazille in some cases!

I did use a polyphonic pad to test - I used "Deep Space DIVA" (beautiful sound!)

But I'm using the VST in Ableton on OSX Lion. Maybe the optimization was AU only? Or maybe Ableton's cpu meter wasn't accurate? I opened the AU in Logic and that patch seemed higher than what was normal in the VST.

I'm still on 10.6.8.
The CPU meter in Live is not accurate. It's as if it's just polling a single core or something. I was watching the Activity Monitor utility built into OSX.

Just a point that comes up a lot when Live's CPU usage does - the Live CPU meter only measures the busiest core. It's more truthful than an averaged overall meter, in a sense, because only one core has to max out for drop outs to occur. So overall CPU load is at most that number times the number of cores in use.



Speculation - I'd imagine Ableton can run a single track as a thread or something like that, but VSTs are sort of a black box - Ableton can manage in and out but not the internals. I'm assuming VSTs can spawn threads by talking directly with the OS, so that's outside of a DAW's control.