GS 3.x (32 bit) and 64 bit Windows XP

[fizbin]

Has anyone been brave enough to try this yet? I'm wondering if the memory management limitations will have changed, even when the consuming application is still 32 bit. It would be great if we could make use of the better part of 4 gigs of RAM, or even more of the approximately 1.1GB of 2GB RAM that most users can eke out of GS at this time. Imagine being able to have just one Gigastudio machine for your entire orchestra template.

PM me if you have 64 bit (Athlon 64) hardware and are interested in trying this.

fizbin

P.S. It's just now dawning on me that since GS installs itself as a kind of driver that it will have to be ported to 64 bit first, before this will work at all. Any news on that?

[griels]

Sorry, this couldn't possibly work as far as I know, due to the 32 bit kernel code in Gigastudio... Only 32 bit user space code can run under XP64.

I imagine they will eventually code a 64 bit version, but not before they've ironed out the problems with the 32 bit one.

[JonFairhurst]

I spoke with Tascam at NAB. They're actively working with the latest 64-bit goodies in the lab. (As is every other major Windows developer who is worth their salt. No sensitive news here.) They are taking 64-bits and multi-core/multi-processing very seriously. They know that the big win with 64-bits is that it blows through the current RAM limitations.

That said, GS3 is 32-bits, period. They've made no public announcements about any 64-bit products, thus far.

To get a 64-bit implementation, one needs a 64-bit processor, a 64-bit OS, 64-bit audio drivers and a 64-bit app.

Come to think of it, I would guess that they will need to spec GSIF-64 and get the hardware manufacturers to implement it. I would guess that this would be a fairly trivial change though.

-JF

[Joseph Burrell]

But won't the bottleneck still occur in the DFD streaming end of the application? I don't see the benefit really. Even though you aren't streaming huge chucks of data anymore (assuming the new RAM threshold means that much more loaded into memory) you're still sending a lot of information up the pipe.

Not bashing anything, just curious as to the real world benefit of this considering the current state of hard drive thruput.

[Bruce A. Richardson]

But won't the bottleneck still occur in the DFD streaming end of the application? I don't see the benefit really. Even though you aren't streaming huge chucks of data anymore (assuming the new RAM threshold means that much more loaded into memory) you're still sending a lot of information up the pipe.

Not bashing anything, just curious as to the real world benefit of this considering the current state of hard drive thruput.

I think it would probably be a matter of being able to load much larger pre-buffers, therefore decreasing the chance that a sound would need to "stream" before its pre-buffer had played out.

Of course this all depends on people being able to afford larger amounts of RAM, as well.

Bottom line, there is no free lunch, no panacea. One way or another, there will be a cost associated with pushing performance past the boundaries of the current state-of-the-art.

As much as everyone would love to believe that the world is going to change forever at the very next technological plateau, this is just not the case. I have been making music with computers since the early 1980's. Progress is almost always incremental, with leaps very few and far between.

And, at least in my experience, the computer you want will always cost $3500.

[Joseph Burrell]

And, the computer you want will always cost $3500.

You've got that right!

Putting it the way you did, I guess I can see how it could increase DFD performance, which in turn will allow greater polyphony in realtime. I just wish there were more focus on the architecture needed for DAW performance. It is markedly different from video production or gaming, which most PC design tends to be created for.

But given the sizes of the instruments necessary lately due to the design of the libraries on the market, how long do you think it will take before the benefit of such a setup will markedly decrease? As soon as the bar is raised, VSL or someone else will simply create 1 (or more) gigabyte patches. I mean, the legato instruments are already HUGE. Of course, you can mix them down a track at a time when you're done.

I'm just saying when raising the bar, there's always someone else who'll come along and require it just a bit higher.

[Bruce A. Richardson]

That's totally true.

The other half of the "computer you want" equation is this:

This year's software always need next year's computer to really kick a**.

[Bruce A. Richardson]

...at which time, Apple and Microsoft will release a new OS within months of each other and break everything.

This is how it works. We were "almost there" when I was using Apple IIe's. We were almost there when I bought a "Lisa" (worst money ever spent award).

We were almost there when I bought my MAC, almost there when I replaced it with a 286 that scorched it and ran Cakewalk, almost there when I got my 386-DX 20, almost there with 1 mb of RAM, almost there with 8 mb of RAM, almost there with 128 mb of RAM (surely), etc., etc.

I was SO almost there when I bought the ADATs.

Almost there, really. Almost there.

[Joseph Burrell]

Hey, look at it like this, you're keeping the economy healthy. :D

[newmewzikboy]

I think you should hold out until there are some good mobo's for the INTEL 64bit dual core processor$$$

[fizbin]

...at which time, Apple and Microsoft will release a new OS within months of each other and break everything.

This is how it works. We were "almost there" when I was using Apple IIe's. We were almost there when I bought a "Lisa" (worst money ever spent award).

We were almost there when I bought my MAC, almost there when I replaced it with a 286 that scorched it and ran Cakewalk, almost there when I got my 386-DX 20, almost there with 1 mb of RAM, almost there with 8 mb of RAM, almost there with 128 mb of RAM (surely), etc., etc.

I was SO almost there when I bought the ADATs.

Almost there, really. Almost there.

I think there will come a day when a computer will be able to do all the (less than 96kHz) DAW-related tasks we ask of it. I don't think its more than 10 years away. I may eat my words, but I don't really think that anyone will remember this post in 10 years :-)

fizbin

[JonFairhurst]

I think you should hold out until there are some good mobo's for the INTEL 64bit dual core processor$$$ Good point. A lot of the current 64-bit mobos don't support all that much RAM. AMD, Intel, whatever. But if you don't have the RAM slots, there's no point.

-JF

[Chadwick]

The thing I'm looking forward to with removal of the ram limit isn't so much increased polyphony as an increased number of simultaneously loaded patches.

Right now, even with streaming and very small buffer sizes, your patch count is limited by ram size. At the same time, we're being provided wonderful tools like VSL's legato setup, which needs LOTS of buffer space.

I like the idea of being able to load the whole orchestra on a single PC.

I don't need to have 120 stereo musicians playing simultaneously, but I would like to have them all loaded up on one machine and ready to play when I point to them ;)

[fizbin]

To get a 64-bit implementation, one needs a 64-bit processor, a 64-bit OS, 64-bit audio drivers and a 64-bit app.


Yup - we have all those except the 64 bit app (gigastudio). M-Audio has had 64 bit drivers available for awhile. I'm unsure if that includes gsif though.

http://www.m-audio.com/index.php?do=support.drivers

Choose: Delta card, 64 bit Windows, and check "Show Beta Drivers"

fizbin

[fizbin]

I don't need to have 120 stereo musicians playing simultaneously, but I would like to have them all loaded up on one machine and ready to play when I point to them ;)

I run into this very problem pretty quickly even with my modest Opus 1 sampleset.

fizbin

[Joseph Burrell]

Yes, but how can you ever create large orchestrations if they can't all play at the same time? Unless you want to create a few tracks, create wave, import, make another 6 tracks, mix them to wave, realize you don't like something in the first wave track, start all over...

And the RAM limit isn't going away, it will just be replaced by the fortitude of your wallet. How much can you afford? How much is the motherboard that will hold the amount you desire? Is that much RAM even going to be available on the current architecture? There's a lot that's going to have to happen before you see any type of jump in this. And even then you'll just have to hope you can afford it.

But, I know, this is a fantasy. Even with GPO, you get to the point where the performance outstrips the CPU. And with everything else I own, its DFD bottlenecking, or RAM, or polyphony. And so on and so forth. I guess I'm dreaming that technology will finally present us with a computer that will last longer than 6 months and will outstrip the needs of the apps we're running.

[newmewzikboy]

How about LinuxSampler as an option? Doesnt it allow you to access more memory or create multiple instances? Doesnt Linux allow you to seemlessly access multiple processors?

[fizbin]

Not sure - you tell me. Can LinuxSampler load my complex giga patches and play them back properly?(ignoring convolution for the moment) Can it load more patches than Windows XP on my PC with 2GB RAM (loads about 1.1GB)?

fizbin

[sbenno]

about LinuxSampler:

the current state of memory management of Linux is :

on a 32bit machine (P4, AMD Athlon) every process can normally address 3GB of RAM. (there are tricks to get it up to 3.5GB per process)
This means if you install 3GB of RAM you can use almost all the memory for samples and streaming buffers, not a meager 1GB out of 3GB.
The Linux OS consumes some RAM too, it cannot do miracles, but keep in mind that the graphical interface of the operating system is the part that sucks up the biggest part of the RAM.
Since you cannot run Windows in text mode this is why you always lose some RAM.
On Linux if you want you can run in console mode and have the applications remotely controlled. LinuxSampler supports it, and it's GUI (qsampler) can run on another machine (which can be Linux, Windows, OS X).
That way you can use almost all the memory for the sampler.

There are even mainboards that allow more than 4GB but this requires a special addressing mode which needs to be supported by the OS.
Linux supports it and the limitation is that you experience a slowdown of a few % (6-7% if I remember correctly) and that a single process can see only 3GB).
But even with that limitation you could for example have a 32bit machine with 8GB of RAM and run 3 LinuxSampler instances and let each instance allocate up to 3GB per instance.

But those special 32bit mainboards are not cheap so it's better to buy a real 64bit mainboard/cpu ( eg Athlon 64).

Linux is 64bit capable since a few years, thanks to it's crossplatform nature an that 64bit CPUs existed long before PCs went 64bit. (Alpha (Digital Equipement) , Sparc (Sun) , MIPS (Silicon Graphics) processors).

So Linux runs perfectly on 64bit machines like the Athlon 64, including the audio drivers (which exist in source code form, so they were made 64bit clean long ago).

When you run a Linux application in 64bit mode the memory limits fall and you can use as much RAM as you have installed.
AFAIK current Athlon 64 mainboards are mostly limited to 16GB but I think this limit will be shifted upwards with the next generation of mainboards / RAM modules.

The big advantage is that the software does not need to be changed to address more than 16GB.

I know LinuxSampler is not yet ready for the joe average user (many are scared by the Linux OS which is not fully point and click as most win/mac are used) but the situation will improve in not too distant future like OS X and Windows port (which will have to live with window's memory limitations) and what's perhaps even more interesting,
clustering capabilities (something similar to fx-teleport).
You have N LinuxSampler slave machines that need no monitor and no audio/midi card (samples streamed over network) and you control them all from within a VST/AU plugin on your main DAW.

This will certainly ease it's use since you can continue to use your favorite OS , sequencer software but at the same time have a cluster of machines running stable and powerful OS stream your sample libs from the disk.


cheers,
Benno
http://www.linuxsampler.org

[sbenno]

Hi fizbin,

while GS 2.5 support should be more or less complete not all GS3 features are implemented yet (no convolution, midi rules).
You mileage may vary.
I think the best is to try it out. If you are not familiar with Linux you can ask a friend or someone if it compiles and installs it for you. The GUI is easy to use and for now if you want to drive a LinuxSampler machine from your DAW you have to use a normal MIDI interface and pipe out the sound using a soundcard. (eg if you use a Delta 1010 you can use it's midi inteface to send MIDI events to LS and pipe out the audio via it's audio outs.


see here for a list of soundcards / midi intefaces supported:

http://www.alsa-project.org/alsa-doc/index.php?vendor=All#matrix

when choosing an easy to use, audio friendly distribution, I suggest Fedora Core 3 with Planet CCRMA audio packages:

http://ccrma.stanford.edu/planetccrma/software/

LinuxSampler is a work in progress and I think for people not having programming skills the best way to contribute to it is to use it and give feedback, report problems if certain sound libraries don't play well etc.

cheers,
Benno
http://www.linuxsampler.org

[newmewzikboy]

Thats the most complete spec I have seen to date on LinuxSamplers Capability. I wish you guys would put this information in your project page.

Question: When you speak of clustering, are you talking about native clustering schemas, or something like Suns Grid Engine, etc?

Question: LxS doesn't necessarily need a windowUI to run...right? You can use batch scripting to preload a configuration of instruments etc.

[Nick Batzdorf]

And the RAM limit isn't going away, it will just be replaced by the fortitude of your wallet. How much can you afford? How much is the motherboard that will hold the amount you desire? Is that much RAM even going to be available on the current architecture? There's a lot that's going to have to happen before you see any type of jump in this. And even then you'll just have to hope you can afford it.

I just bought an extra 4GB of RAM (total 5GB) for the dual 2.5 G5 I just upgraded to. It cost $350. With that amount of RAM you can safely load about 3GB inside Logic, leaving room for other programs as well. That's 2-1/2 Gigas or Kontakts on one machine, which is getting there.

When Logic is able to access the 8GB you can install now, that'll be about 6-1/2 Gigas or Kontakts - and the machines around when that happens will probably hold more RAM. 6-1/2 Giga/Kontakts is enough RAM access to load an orchestra, I think, and at today's prices it's not outrageous at all.

[peter269]

....And, at least in my experience, the computer you want will always cost $3500.

Well, we costed out the system Sonar 4/64 is using in demos - try $6500 today (http://www..com/roaringlambs/editions/april2005.shtml).

32-bit programs will ONLY work on XP64 in the emulation mode, and they don't benefit from 64-bit at all. They run at the same 32-bit rate. Audio drivers for XP64 are coming out (M-Audio, E-MU, etc), but realistically on the PC, you won't see any real 64-bit action until Q1/2006 realistically.

Also, motherboards are only up to 8 RAM slots right now.

The thing is, I feel that the best way to approach XP64 is with a server mentality. Steinberg is supposed to already have a 64-bit program. So the key issues for most of us will be how quickly NI, Steinberg, and Tascam plus audio card drivers, get to the new 64-bit standard. Then let's see how much can really be done on one system.

[sbenno]

Hi newmewzikboy,

the information is in our project web page but it's quite synthetic :)

http://www.linuxsampler.org/features.html

Regarding clustering: the initial plan is to provide non load balanced clustering which means it's the responsibility of the user to assign samples to the single machines. (eg like you do with fx-teleport setups).

Automatically figuring out the best distribution of samples is a difficult task and it mostly conditioned by the midi sequence plus you need to keep the data on the disk on the machine that is going to access it fro sample streamig since streaming files over network is slower than having them on a fast ATA bus.

Regarding scriptability:
the GUI is not necessarily needed, LS can be controlled by using LSCP (LinuxSampler Control Protocol) which is a human readable, text based control protocol.

the full documentation is here, it's actually not hard to understand.
http://www.linuxsampler.org/api/draft-linuxsampler-protocol.html

here you can find a few examples along with comments:
http://www.linuxsampler.org/documentation.html

the GUI (qsampler) can load and save such scripts but you can use your own tools too if you want (eg a text editor and an utility to send text over the network eg "netcat") and then place those commands in batch scripts to load large setups with a single command etc ...

I hope that I did not bore you guys with too much technical stuff :)

cheers,
Benno

[aplanchard]

I just bought an extra 4GB of RAM (total 5GB) for the dual 2.5 G5 I just upgraded to. It cost $350.

Hey Nick,

Where did you find it so inexpensive? Presumably you bought 1 GB chips, which I can only find at about $230 a pair.

Allan

[belbin]

Although this is a fascinating and exiting thread, I just want to point out that it's also horribly depressing. I've spent my last 12 months' earnings just getting to the point where I can Run GS 3 decently on one machine.
The longer the world waits to have to upgrade everything again, the better.
I know many of you can't wait, and I'm sorry, but progress be damned! At least for a couple of years.

Belbin

[newmewzikboy]

I think you can safely wait to upgrade just about everything.

I think you should also seriously look at the LINUX alternatives, which I believe is the cheapest and most powerful way to expand the studio...and with old hardware to boot!

[Maraxalamanta]

How come these new toys (like X-box 360 and PS3) are getting 4x the CPU power of todays computers at a fraction of the cost? This annoys the living h*ll out of me! On a side note, does anyone have any idea how long it till take before the PCs and software out there all run in a true 64-bit environment? Upgrading to 64-bit comps right now seems pointless because Giga doesn't even support more than 1.5gb ram (hah, the irony of it all! Giga marketing itself as a sampler that is virtually limitless and as it turns out their major selling point ends up being the true bottleneck... too ironic! :)

Is Logic on a Mac 64-bit already?

[fizbin]

How come these new toys (like X-box 360 and PS3) are getting 4x the CPU power of todays computers at a fraction of the cost? This annoys the living h*ll out of me!

I'm not sure exactly, but my understanding is that a large part of the cost of the console is subsidized perhaps by the game licensing and sales. That is, they sell the hardware to you as cheaply as possible and hope to make the real money in game sales.

fizbin

[Nick Batzdorf]

Is there a royalty for developing for a platform? That would surprise me.

[newmewzikboy]

...Upgrading to 64-bit comps right now seems pointless because Giga doesn't even support more than 1.5gb ram (hah, the irony of it all! Giga marketing itself as a sampler that is virtually limitless and as it turns out their major selling point ends up being the true bottleneck...

When I first purchased Giga, I called nemesys software tech support to verify memory configuration sizes because I wanted to MAX my memory..and the RAMBUS memory I was purchasing was expensive. They assured me that I could in fact use as much of the 2GB to load samples as I wanted. I went ahead and purchased the memory, and installed XP. To my frustration, I could not get it to work, and their tech support suddenly didn't have any answers. Through the forum boards, and much experimentation, we concluded that they were totally misleading us. In the end, through research via INSIDE WIn 2000, a couple of config settings were found to help get a bit more than 1GB, but we were screwed.

Moreover, the sampler sold itself as being able to stream and thus not having to load memory except for a small hook. This turned out to also be bogus.

The moral from this lesson is trust no one, and trust no published article. If you can test the config before buying, all the better. Long term, I am moving as much as I can to Linux and open source, and steering clear away from the marketing hype as I can. The engineering and feature set, surprising to my skepticizm, is better.

P.S. Yes I am still a p**ed off customer :)

[rob morsberger]

Nick,

I'm interested that you seem to be exploring the idea of one machine doing everything.

As a Mac user I've been reluctant to updgrade my Giga machine and/or add new ones. I'm really on the fence about Giga 3. That's partly because I mix everything in Logic, use Altiverb, and I'm not sure about the need for Gigapulse etc.

However, heavy loading the Mac seems a bit iffy. I get crashes and consequent corrupt files often enough to worry. (Kompakt and Intakt seeem to be the worst offenders.) So offloading to other machines seems good for that reason. Plus I still really like Giga a lot.

But the simplicity of one machine is very very appealing. Please share your thoughts some more about this subject if you can. I'm about to buy VSL and since I think you use that on EXS, I'd love to hear how that works for you too.

Sorry if this is diverting the thread.

Thanks,

Rob

[fizbin]

When I first purchased Giga, I called nemesys software tech support to verify memory configuration sizes because I wanted to MAX my memory..and the RAMBUS memory I was purchasing was expensive. They assured me that I could in fact use as much of the 2GB to load samples as I wanted. I went ahead and purchased the memory, and installed XP. To my frustration, I could not get it to work, and their tech support suddenly didn't have any answers. Through the forum boards, and much experimentation, we concluded that they were totally misleading us. In the end, through research via INSIDE WIn 2000, a couple of config settings were found to help get a bit more than 1GB, but we were screwed.

Moreover, the sampler sold itself as being able to stream and thus not having to load memory except for a small hook. This turned out to also be bogus.

The moral from this lesson is trust no one, and trust no published article. If you can test the config before buying, all the better. Long term, I am moving as much as I can to Linux and open source, and steering clear away from the marketing hype as I can. The engineering and feature set, surprising to my skepticizm, is better.

P.S. Yes I am still a p**ed off customer :)

I sympathize with your predicament. However, you'd likely be in the same boat with the other software samplers under the Windows platform. (does anyone have more luck loading more samples into RAM with Kontakt?) In fact, I think using the default Kontakt settings you are actually loading a little longer bit of each wav than Gigastudio requires, effectively decreasing the number of sample sets you can load. (I may be mistaken here - correct me if I'm wrong)

I really do believe that these limitations will melt away as we migrate to Windows 64. In this case, the immediate needs would be in favor of a Gigastudio-only machine, because it is the only audio application that would have to run on the 64 bit machine, where if you had Kontakt or HALion or one of the other samplers, you'd be trying to run it from one machine and the timeframe for all your plugins, host, sampler, and so forth being ported to 64 bit may be much further away, if ever.

fizbin

[newmewzikboy]

I sympathize with your predicament. However, you'd likely be in the same boat with the other software samplers under the Windows platform. (does anyone have more luck loading more samples into RAM with Kontakt?) In fact, I think using the default Kontakt settings you are actually loading a little longer bit of each wav than Gigastudio requires, effectively decreasing the number of sample sets you can load. (I may be mistaken here - correct me if I'm wrong)...

My understanding is that you have more flexibility with Kontakt, by having multiple instances running at the same time - perhaps I am wrong?

In any case, I have a lot of faith that LinuxSampler could become a great product if it gets more support. Perhaps we should open a forum for it here?

[sbenno]

Hi,
since you seem to "think different" , how about trying LinuxSampler (if you haven't already) and join mailing list to provide feedback ? (eg what .GIGs don't play correctly etc)

http://lists.sourceforge.net/lists/listinfo/linuxsampler-devel



I think you can safely wait to upgrade just about everything.

I think you should also seriously look at the LINUX alternatives, which I believe is the cheapest and most powerful way to expand the studio...and with old hardware to boot!

Just to speak about consoles.
LinuxSampler can run on them too since Linux exist for several game consoles.
The only problem is that except the XBox no one comes with a HD.
The XBox is a P3 with an IDE HD so it's performance should be comparable to a similarly equipped PC. But by default AFAIK it comes with only 64MB of RAM so I guess you cannot preload that much sample heads in RAM :)

The other day I did read the specs of the new Playstation 3 and Xbox 360, kinda impressive (the PS3 seems to have like cell processors with lots of DSP muscle).
And it has a slot/interface for a HD. Imagine LinuxSampler (but someone would need to implement DSP acceleration code for the PS3 to squeeze out the maximum from the hardware) running on a bunch of those beasts, all networked in a gigabit LAN :)


When price will come down (eg $200 per unit) 10 PS3 for $2000 would not be much for a personal supercomputer :)

The problem of consoles as flzbin said is that they are often sold below cost to gain market share and then recoup costs with pricey games. This is whay console makers are putting all sorts of hardware based DRM stuff into the consoles to avoid (or at least make it hard) that people run unauthorized software (eg only official games).
Otherwise at $149 the XBox would be a kinda cheap general purpose PC.


Ok back to the present, sorry for the digression :)

cheers,
Benno
http://www.linuxsampler.org

[newmewzikboy]

Problem is...I have no free hardware lying around right now to dedicate to LINUX partition or server. I thought about commandeering some hw at my gfriends office hiding it under a DB2 database :) .

[fizbin]

Is there a royalty for developing for a platform? That would surprise me.

My understanding is that there is. I think the console manufacturer gets some percentage of the game sales. I could be wrong, but this is what I read awhile back somewhere :-)

fizbin

[Nick Batzdorf]

Where did you find it so inexpensive? Presumably you bought 1 GB chips, which I can only find at about $230 a pair.


http://www.18004memory.com/cart/addcart.asp?Try=Yes&itemID=502403&itemTYPE=cat&session=xDEALRAM&shippingmethod=0

Yes, 1GB chips. They're generic, of course, but I don't care.

[Nick Batzdorf]

Rob, I load the main Mac to the gills and also use Windows machines. For me one machine isn't enough.

What specifically are you wondering about VSL?

[rob morsberger]

Ah, hi Nick.
Your earlier post was provacative in its comparison of what you could load up on a Mac relative to, say Giga. It got me thinking. I'm ready to pull the trigger on VSL but unsure of my platform. I've been very happy with Giga but haven't upgraded yet...so also contemplating alternatives, partly since the upgrade will entail new machine(s) etc. Are you running VSL on exs. giga or both?
But really you already sort of answered my question.

Thanks,

Rob

[fizbin]

My understanding is that you have more flexibility with Kontakt, by having multiple instances running at the same time - perhaps I am wrong?


I don't know that it's more flexible. It is a little more convenient to use as it runs in the host process so the signal processing and routing can be done in the host.

GS 3 has pretty good routing now too.

fizbin

[newmewzikboy]

I don't know that it's more flexible. It is a little more convenient to use as it runs in the host process so the signal processing and routing can be done in the host.

GS 3 has pretty good routing now too.

fizbin

I should have been more clear: Can you use more than one gig of memory, by creating multiple instances of K running on the same machine?

[Nick Batzdorf]

Are you running VSL on exs. giga or both?

Both, and I use other libs too. I can't really say one works better than the other or is more cost-effective.

The main advantages to Giga are that you can expand beyond one machine and that you can keep it loaded when you switch sequences - although EXS can keep used samples in memory without reloading too.

The main advantage to using it in EXS is that it's integrated into Logic and that it's all in one machine.

[rob morsberger]

Thanks for the reply Nick.

Rob

[Nick Batzdorf]

De nada.

What I actually meant to say is that they both work very well; what I wrote sounded very unenthusiastic.

I should add that the G5 that loads 2 to 2-1/2 Gigas also costs about the same as about 2 to 2-1/2 Gigas. :)

[Nick Batzdorf]

Oh, and I can't take credit for being the only person concerned with getting as much as possible out of single machines! That's the name of the game for everyone working with large orchestral libraries.

[fizbin]

I should have been more clear: Can you use more than one gig of memory, by creating multiple instances of K running on the same machine?

I don't think so. Since all the instances run inside the host process, they are all equally limited by the single process ceiling on RAM usage.

fizbin

[newmewzikboy]

:(
I don't think so. Since all the instances run inside the host process, they are all equally limited by the single process ceiling on RAM usage.

fizbin
How depressing..