Man, I love technology. As soon as you think you pretty much know everything there is to know about something, boom, you find out that you were wrong. Very humbling, I must say [:)]

Anyway, I had to try this thing, so I did a couple of tests here, on a Dual 2Ghz G5 with 8GB of Ram.

Here are my findings...

First, I tried maxing out a single stand alone instance of VI. I went up to 3.25GB of used memory (VI indicated "3004 MB" of used memory). I couldn't load more samples than that - after that, the app simply crashes (I tried it twice!).

With two Vienna Instrument instances loaded, I was able to load 2.71GB on one, and 1.10 on the other. At 1.11 GB, the second instance crashed. Total = 3.81GB.

http://vsl.co.at/upload/users/22294/VI_Duo.png">

Conclusion: it seems more stable to have two, smaller instances of VI loaded, than a single, fat one.

So... I decided to try something even crazier.

I launched Plogue Bidule, and loaded a huge preset in VI, with about 3GB of samples.

Then I launched one stand alone instance of VI, with 1.5GB of samples.
Then I launched *another* stand alone instance of VI, with 1.5GB of samples.
And then, I launched *a third* stand alone instance of VI, with 1.5GB of samples.

Everything worked quite ok for a few minutes, but the third instance then crashed.

So, I launched another third instance, and, this time loaded only 1 GB of samples. Everything worked fine, and is still working fine right now.



Total : 3 + 1.5 + 1.5 + 1 = 7 GB of samples.

That's quite nice. And a *huge* find.

I gotta say thanks to you all for challenging us with that idea. That's a huge discovery for me. In theory, and if the CPU is powerful enough, you could have one Mac Pro instead of four or five Mac Minis!

WOW!!!

Jerome

jerome, thanks for posting this numbers - although i had to take the liberty and crop your images to ~600 px because for users reading the forum within the website they had exceeded the max.available width
christian

Hey Christian,

no problem. I'll make sure they're not more than 600px next time.

Jerome

Nick, JBM, Jerome - - many, many thanks to all of you. Following your collective advice I've managed to get the thing working - using Soundflower - in cooperation with Logic. Jack OSX looks interesting and I'll try and figure it out later.

Previously, I'd gotten as far as the horrendous feedback loop (didn't blow my speakers luckily!), but Nick's solution works perfectly. The one thing I'm not absolutely clear about is running more than one instance of the standalone. Nonetheless, in my primitive way, I seem to have accomplished it by simply making a copy of the application and running the copy along with the original. Certainly glad I asked for help on this. Pretty nifty to be able to run a lot more samples on one computer.

Again I'm extremely grateful.
Stephen

Hey Steve -

yes, to run more than one Stand-Alone app, you just need to copy it. If you want to keep both in the same folder, they just need a different name.

If you tried it with more than one stant-alone app, can you let us know how things are going ? Any problem regarding MIDI ? Is everything going to the right stand-alone app? There's nothing like the MIDI signal mixing-up or the like?

Jerome

Jerome, using different IAC busses ensures there are no mix-ups - you just select a different "MIDI interface."

What I haven't experiemented with is how to separate the audio outputs from each instance. The Vienna Instruments player allows you to select one device, e.g. RME Fireface or MOTU PCI-424 or built-in; there doesn't seem to be any way to be any more specific than that. I assume Soundflower is just another device, and it'll have the same problem.

So the first instance is no problem, because it's going to go to the default output pair for the audio driver. And you can of course route the outputs inside any host to a separate pair; so far so good. But when you add the second insetance outside the host, that's when I think you might have to do some thinking.

Yes, I did realize this problem. In VI Stand Alone, you can only choose the audio interface - you can't choose a specific channel. So, let's say you have an interface with ADAT, you can't tell "Stand Alone 1" to go out ADAT 1-2, and then "Stand Alone 2" to go out ADAT 3-4.

The solution could be:
- For the VI Stand Alone 1, to go out the default outputs of your audio interface (let's say, ADAT 1-2)
- For the VI Stand Alone 2, to go out SoundFlower, back into Plogue, and then to route the signal to ADAT 3-4.
- For the VI Stand Alone 3, to go out Jack, back into Plogue, and then to route the signal to ADAT 5-6.

What do you think?

Just brainstorming here...

Jerome

My first reaction was that it sounds a little kludgey. But then OS X has no problem with multiple audio drivers, which is what those are, so why not.

Another thought - which I haven't investigated - is to create two aggregate devices, one Soundflower-audio interface and the other one audio interface-Soundflower.

I don't know if that would work...

It works!

That eliminates one audio driver.

The silly thing is that I actually get excited about stuff like this. [:)] [:O]ops: [:D]

I don't follow you here... what do you mean it elimitaes one audio driver?

Jerome

I mean you don't need to use both Jack OS X and Sunflower - you can just use one of the two and your regular audio driver. That gives you four output choices: Sunflower; audio driver; Sunflower/audio driver aggregate device; audio driver/Sunflower aggregate device (or substitute Jack for Sunflower).

If Jack OS X and Sunflower - two software drivers - will work in tandem, my idea gets you nothing. My fear is that using two audio drivers is risky.

But as I said, I could be wrong.

Ok, I see your point now... but where would the signal go if you use the aggregate device? Let's say you use "Soundflower+Audio Driver", where does the audio go?

Jerome

To an input pair inside Plogue, most likely, via Soundflower (since the first two Soundflower inputs are likely to be where the Vienna Instruments player sends its output).

Except that I'm stuck in RME Fireface 400 land, having been working with it for a review. Audio Driver + Sunflower is only going to go to an audio interface output, not into Plogue, so that doesn't work. The RME box has internal mixing, so software outputs can be routed to any DAW input.

Maybe I'm not as clever as I thought I was. You still have Soundflower and Soundflower + Audio Driver; I think that would give you two different routing options with Soundflower. But that may be another hole in my thinking.

To me, it would seem that the same issue would happen with "SoundFlower+AudioDriver", because it would simply go out SoundFlower 1-2... which is already being used by "SoundFlower"...

J.

Probably, which is wny I don't think I'm as clever as I think I am.

But it might go to another pair if Soundflower decides that pair is already being used - in which case I'd be clever all over again.

[[[:)]]]

I'd love for you to be right, but I don't think SoundFlower would act that way [[[:)]]] But there's nothing wrong in trying! [[[:)]]]

J.

Jerome:

In answer to your question- MIDI works perfectly with two standalones as each is on a different IAC bus - - so there's no mixing of signals.

Nick:

I'm not sure how having two aggregate devices would work with Logic as, unless I'm mistaken, it appears that one has to choose only one default driver in both Audio MIDI Setup and in Logic's Audio Hardware and Drivers dialog box. However it seems possible that one could make an aggregate device with both Soundflower and Jac OSX. Haven't tried it yet, however.

In answer to your earlier questions I have a MOTU 2408 mk3 audio interface, my computer is a dual 2.5GHz PowerMac, equipped with 7GB of RAM running OS 10.4.8. Logic is v. 7.2.3.

My usual working procedure is to write in Finale which is set up to play back Logic's Audio Instruments through the IAC drivers. When the score is completed, I make a MIDI file and transfer it to Logic for detailed editing. The Vienna Instruments matrices have made this a much less laborious enterprise - previously I had 64 channels of sampled instruments - - so writing each instrument on one staff and one MIDI channel has made life a lot easier.

Finally, I hope the folks at VSL are listening to our discussion here as it would be great if an upcoming revision of the VI software would include giving the standalone the ability to choose MIDI channels and, more importantly, the ability to choose audio output channels. Using the standalone is clearly a way around the 32bit limit.

Again many thanks to all of you. I will continue testing later today.

Stephen

I'm with you on that, Stephen. I think main big point that comes out of this discussion/investigation is that it may be worth VSL bulking-up the standalone a little - different midi channels, access to channels on audio devices, and maybe even multiple instances of the VI -- like a host, but not a host for plugins, just a host for the basic VI. I don't know exactly what could be done in this regard, but it certainly seems worthwhile to investigate what's possible with multiple processes.
I had the idea a while ago about the vsl-server actually spawning new instances of itself in response to memory demand, and based on available memory. Though I suppose this could be a bit of a nightmare with the server. Mind you, since the standalone seems totally self-contained, it certainly is a good candidate for such "spawning"...

J.

Given the revelations in this thread, I think that simply allowing the player to access different audio interface outputs would solve the problem - and from my armchair it doesn't seem like it would be a huge deal to implement.

I'm really curious to hear from someone who has more than 8GB installed in a Mac Pro. Can you keep loading instances? Or are there other limitations, such as CPU?

I've had no success so far in eliciting separate audio outs from different standalone instances. Also I've run into a problem when trying to record (a MIDI recording) of a perf legato cello loaded into a VI plugin after recording a sus-vib violin in the standalone. THe cello's notes keep getting cut off when I'm trying to play along with violin. However if I record the perf-legato cello first using the plugin, I can then record the violin (in the standalone) without anomalies in either instrument. Clearly this is, at least in my implementation, far from a perfect solution. (I don't think the problems I encountered result from my computer's deficiencies as, using the plugin version in the normal fashion, I'm able to play a good number of very complex matrices simultaneously - - without anomalies).

I don't think that the standalone was designed with aim of using it simultaneously with the plugin , so it would be great if the usage we've discussed in this thread - - which is essentially to find a way to overcome the RAM limits imposed by a 32bit system - - were included in the design goals of the standalone. Of course a comprehensive solution would be a 64bit version of OSX (which supposedly OS 10.5 will be), a 64 bit version of Logic and a 64bit version of the Vienna Instruments software. This desirable goal does not seem entirely likely in the immediate future, so refining the standalone, especially in light of the current licensing policy restricting each library segment to one machine, would be a significant interim improvement. If one computer can do the job of several machines even with present day technology, why not do it?