It doesn't sound to me like MIR is an elite product for the few but a necessary solution for what is now impossible: correct placement in space of each and every instrument. Gigapulse is designed to do something like this but you're out of processing after you place a couple instruments. So if there are (say) presets to chose from where all the VSL instruments are placed right where they belong and with processor issues, that would be great.

Dave Connor

And I agree with him.
Far from being an elite product, it seems VSL is trying to work with all of us, by configuring programming in such a way that mkaes it possible to get the best out of the samples, and provide a workaround the limitations in soft and hardware, so we don't have to.
I see that as something for everybody.

Our collective continuing problem is the inability of the hardware to keep up.
The common sense, non elistist approach, is to bolt on another machine, with the advantage of additional processor(s) and ram. Even i understand that.
And what's an additional Pc cost these days?
500 pounds for something reasonable that's capable of running MIR? Say 750 pounds, for a dual processor, multigig ram spec, for a program that Sir Dave of Langeles rightly says will actually do what we've been trying to muddle through with in stereo, no matter how 'enhanced' that stereo may be?
I see it as a simple solution to a challenge that's been going on for years, and as Sir Dave also said the solution gets even easier with presets.
Turn on, select preset, play orchestra with everyone sitting in the right place spacially, not in one long aural line across the front of the stage.
As as i wrote earlier, i'm keen to finish the project by finding out if it's possible to place samplers across the 'boxes', sharing the load, and increase the available track count.
Far from being elitist, i'm getting the impression we're all closer than ever before to realising the dream of everything working at once, without having to workaround.
The price of VSL is a serious but sensible figure for a serious and sensible investment. I would think it sensible to purchase a setup that allows VSL to be the Rolls it is, and not handle like a Ford.
3 Boxes, shared load, and the potential for no freezing or bouncing?

I can live with that.

Regards to you all,

Alex.

... to keep expectations from going up through the clouds ... [:)]

The MIR will be able to do a healthy amount of real-time processing, with carefully chosen curtailment (is this the proper word?) within the actually necessary amount if IR-data to give the user a meaningfull and convincing auditive impression. The _full_ load of several hundreds of IRs will still have to be rendered off-line, until computers are powerfull enough for our needs.

Sharing the processor-load between several machines is of course a tempting scenario, but I can't promise that this will already be possible in the first incarnation of the MIR. We'll see ... [;)]

Dietz,
So would it possible (leaving MIR aside for a moment, so you don't feel I keep pressing the point) to do this with a work box (sequencer), and a server?

Regards,

Alex.

Uh - I'm not sure if I really understand the question ....? Sorry, Alex [[:|]]

Dietz,
In the discussion, the point's been raised about multiple boxes. This may be speculation, but i wondering if a computer as a workhorse, housing the sequencer could be wired to a server containing samples, 'Mir' etc. Would there be a benefit to this in terms of latency, playback etc. Do any of you use a server to house samples and play them back from an attached computer? The reason i ask is because i don't know if a server would be (a) more or less stable, and (b) whether the server's on board processors could do a better job of playback with less latency. If so, i'm thinking and wondering if MIR could also sit on the server and be useable.

Regards,

Alex.

That's beyond my technical knowledge as well as my real-life experience. Maybe Chris Marin (CM) can chime in here, with some educated opinions ...?

well, to house samples on a server (eg. for shared usage from several giga-slaves) is a realistic scenario - given the storage is fast enough and the network is gigabit ... no problem at all (the inconveniance with giga 2.5 was that the quicksound database didn't accept net-shares to list the instruments, exs accepts links to shares). the term *server* should be understood in a functional way here, no need for special motherboards, memory or processor (filesharing is a boring job for a computer).

latency originates only marginally from a processor - it comes from (physical) midi devices, soundcard buffers, drivers, ect.
each machine with reliable (must not mean expensive) hardware, a carefully installed (and a little bit downsized to the needs) operating system and not running too many applications (which in most cases have never been seriously tested to work side by side) runs stable. some for years.

i'd like to see MIR as a complex sound processor - here we had to look carefully at short processor pipelines (to avoid dropping of too many miscalculations), a well sized processor cache, as much GHz one can get, fast memory and bridges (between PCI-bus, RAM and processor) and a performant technology to feed MIR with the dry audio streams. the machine itself and the OS can be a streamlined server or workstation (there are lots of services nobody really needs on an audio workhorse)

the output format and device is then a matter of taste ... probably the data has to be gathered as audio tracks/files on a disk. considering the various reported (and encountered) problems having a series of applications running on a single machine i feel it would be a good idea to let MIR run on a dedicated computer.

christian

Thanks Christian.

I've done a bit of looking on the Apple site, and particularly at their Xserve machine.
Dual Processor, 222GB of 10,000 rpm hard drive(just over 1 TB of 7,200 rpm), Gigabit ethernet (x2) and options for 2 Fibre Channel Pci-x cards.
Oh, and ram up to 16GB.
Apple seems to say on their website that plugging one of these into a Mac G5 will happen, with no extra fuss. Given the obvious need for wiring (Gigabit and Fibre Channel?), and file hierarchy (again Apple says this is standard), is this a viable option? I also hear what you say about Mir being more likely to perform best on a dedicated machine, and it's possible to have a server for samples at the processor and ram specs above, and then add another server in 'Cluster mode' that operates as a speed machine, based on power at the sacrifice of some hard drive space and the optical drive. Only thing is, Mir is going PC first, and i wonder if a PC instead of cluster server will actually be able to keep up with the rest of the setup.

Thanks again, i just learnt something,

Regards,

Alex.

p.s. This would be a strictly Music composition only set up, so no extra programs would be installed. Barebones.

I'm not sure if you both talk of the same thing. A sample server is not the same as a seperate sampler host.

It's possible to store the samples on one or more seperate PC, map the drive where they are located as network drive to such a GS3 PC. That doesn't need a lot of RAM for the server (256MB is enough) but for such a GS3 PC the usual amount (max 2GB). The PC hosting GS3 doesn't need the samples stored on itself though.

The other way is to have the sampler installed on every box so that. For the full amount of samples you need big enough harddrives in every and each of them to have everything availible everywhere. You'll need more sampler licenses, 2GB of RAM for each PC and so on etc...

Are we talking of the first or the latter?
PolarBear

Polar,
That's the problem. I'm not a techno type. I'm looking at the Apple site at Xserve, and thinking a server, or two, could store samples, connected to a G5 with gigabit ethernet and/or fibre channel pci-x cards, and the G5 (with logic) would see the servers as external hard drives. The servers would simply store samples, but 'play' them' when required. The Apple Xserver also has firewire ports.
I'm fishing in the dark here, trying to find an alternative to 3 or 4 Computers, or an external hard drive, and try and take the strain off the computer (G5) with sequencer.
With my extremely limited knowledge, i can see the servers are dual processor equipped, with a buss for each processor, and the option of 3 10,000 rpm HD's in each. (In addition to the capacity for 16GB of Ram in each server)
It seems an alternate option, but perhaps i'm missing something obvious, as to why servers can't be used for this purpose.

If you have info, i would appreciate it.
The ultimate aim is to run a full orchestral setup in Logic, without having to Freeze, or push the G5 to it's limits.

Regards,

Alex.

Alex... Well... I was typing some lame car analogy that wouldn't fit here. Please don't get me wrong on this, but I can't say it less harsh: I don't know why you want to settle for it right now, as I can't see any reasons for that. You talk about a PC product that has no official release date yet and not nearly anything like system requirements posted. You even go further and think of a Mac port of the software that is not yet even announced (and inofficially said to be only *considered*).

In any way, if you buy that thing now for let's say $10k you could as well buy the same or a better system for that money with the proper specs the day the software is out. Why keep fishing in the dark?

I don't know much about Logic or Macs, but if you push the limits now you have to get a second G5 with an additional Logic key afaik. I'm sure (and that's a speculation of mine) that MIR will have connectivity options for both, PC and Mac once it is out, be it only an analogue audio cable (I hope not). See, again fishing in the dark. If you want to have your setup now and ready, go for Altiverb and buy a second G5 if yours is pushing the limits. Everything else is like winning the lottery next week.

Hope this helps,
PolarBear

Polar,
I wasn't just thinking about MIR, but a complete setup. And i'm fishing for the future not today. When the time comes, i'll be better armed with more knowledge.
MIR will happen when it happens, but as I've posted more than once, it's actually using the sample library in a reasonable, practical way without so many workarounds as a result of Soft and Hardware limitations that interests me.
A lot of questions in the forum relate to limitations and ways to get around them. I read of Jay Bacall using 350 tracks to program and conduct Romeo and Juliet, with workarounds a major part of his time, freezing tracks because the hardware can't keep up.Why not consider alternatives to the current 'accepted' way of doing things, and think about the future? Sample libraries will get bigger not smaller, so i think now's the time to speculate about new ways of taking advantage of that.
Rushing out to buy this gadget or that has never been my way, and more than once i made good decisions based on research.
As i said, i'm no technical genius, but it's an undeniable fact that herds of computers lashed together is in itself a workaround, and there's no 'compulsory' method that can't be questioned.
I guess i'm literally trying to think outside the 'boxes'.

Regards,
Alex.

@hermitage59 said:

[...] now's the time to speculate about new ways [...]
Alex.

Dear Alex,

rest assured that we do not only speculate, but work _hard_ on solutions for making virtual orchestration easier [:)] ... your guys on this forum will most likely be the first to know about each of our moves, as soon as there's something new to tell.

All the best

Thanks for the KVM info, cm. Yes, my experience has been what you say: once I power-cycle the switch, the keyboard is there.

This is a real USB keyboard, by the way - I use it to control two Macs and two PCs. (The third PC is on VNC.)

Alex, I think I understand your point now to a certain extent. However I think I'm too far armed with knowledge so that I only can handle with facts and data sheets to think of possibilities. There are too many things that are simply limited in a way, and there is no way to get more without a useful "workaround" which you want to avoid. Although I don't look at the things always as "workaround".

A few years ago when impulse libraries were introduced and used, you couldn't run your piece orchestrated with your AO copy and three impulses in realtime (on one machine). A recent machine could do that with the most complicated orchestrations. But we have more tools than a simple convolution engine and AO now. So we have to build clusters of processing power just like back then. If computer nerds are looking for the next unknown prime number, there is not just one computer calculating stuff, they also have clusters to spread the load. This is only a necessary thing to get the work done in an efficient way - it still takes months btw - , and however not a "workaround" for me, because there is simply no other possibility yet.

You buy another harddrive if you're running out of space in addition to your existing one and not replacing it by the new one. A "workaround" to have two HDs if one could be enough? Ah, the performance also increases.

Future developments are always bound to limitations of some kind. To circumvent them may directly lead you to others. It would e.g. be possible to develop (for high cost) a DSP card PCI-xpress card that is handling all convolution needed for MIR e.g., but what's the effort worth if it only can be used with Mach5 exclusively on PC (or Giga or EXS for that matter). Formats and compatibility are major concerns for every development because it has to be compatible with the most setups people are running or it would never sell. It has to be easy to use, too (imagine to installing a PCI card for some hardwarephobics). So that could be outruled again.

Even with the biggest computer you could buy for money today there will be task where one of them alone will not have enough processing power. After all, it isn't economical, and history shows that only the economical setups are usually making it through a long time. The thing with servers is, they are designed for a very different task. They usually have their own server OS'es, which isn't of any help for us composers because we need the standard media OS'es for our Kontakts and Logics (economical solution). Those applications couldn't even make use of the processing power offered by such a system perhaps.

Maybe that and not only that is why I like to handle facts and why I try to get a long with limitations. Maybe I got it all wrong, I'm sorry if that's the case. I can't go 350mph with a Ford Mustang and there is no way for me around that. I can fly at that speed though. Different beast again.

PolarBear