Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

How to "strap" 5 or 6 G5's together?

M

MaxMacHac

macrumors newbie
Original poster
Dec 5, 2003
10
0
North Carolina
I evidently posted this question in the wrong forum before... if this is the case here, please let me know where would be an appropriate forum. TIA.

I am interested in "strapping" say... 5 or 6 dual G5's (not a hard number... if it takes 5, 10 or 20 that's what it takes.) together to form a "mini-supercomputer." This will be for a recording studio operation.

Here's the concept... By having multiple processing, timing can be achieved to allow signal paths which will be able to leave the digital realm and be processed though analog gear and back into the digital realm and finally back out to the analog world for final composition.

I have about 8 months to research and purchase everything... which ain't all that much time... Sooooo, does anyone have any ideas as to developers to consult with?

Thanx for your time.

Max
 
Re: How to "strap" 5 or 6 G5's together?

Originally posted by MaxMacHac

I am interested in "strapping" say... 5 or 6 dual G5's (not a hard number... if it takes 5, 10 or 20 that's what it takes.)
Click to expand...

This is simple go to Your local Hardware store and buy some straps and strap Your G5's together and Boom You are all set
 
Re: Re: How to "strap" 5 or 6 G5's together?

Originally posted by Spock
This is simple go to Your local Hardware store and buy some straps and strap Your G5's together and Boom You are all set
Click to expand...
Rent some of the old Red Green shows, he can do anything with duct tape. ;)

Duct Tape Virtuoso Deluxe
 
Re: Re: Re: How to "strap" 5 or 6 G5's together?

Originally posted by Sun Baked
Rent some of the old Red Green shows, he can do anything with duct tape. ;)

Duct Tape Virtuoso Deluxe
Click to expand...

Don't use duct tape - use scotch tape or something - duct tape has a monopoly in the market and is trying to bully all of its competitors out of its space (including masking tape)
 
i dont think you'd gain as much of the "mini-supercomputer" effect in the way that I think you're going to string them together.. all you would really need is to have one station act as the primary (time-master), and have the rest act as secondary's using the timing provided by the master... that way you could have your secondary's store samples, control your devices, etc..
 
I think the word you're looking for is "cluster", not "strapping" ;)
Go and take a look around the Virgina Tech website/forums, they have done what you want to do (all be it on a larger scale) but the principles will be the same.
 
I tried "duck" tape on my G3 and a coupla dual g4's and the tape wouldn't quite hold... darn feathers... I thought about those straps like they use on the air conditioner duck-work... I dunno...

The Virgina Tech site is what prompted my idea. I was wondering if a cluster arrangement would suffice for the core processing, but then have a couple other boxes available as "real-time" processing for D/A and a couple more for A/D back in... i.e. insert points. With the core cluster handling all of the timing... then utilizing the core to handle the final D/A conversion down to say 32 final automated tracks for summing to the 2buss... thus a "modified" cluster? or a cluster with "thingys" and "hangers"??? awww shucks, I'd just call em' $^%&^ strap-on's... maybe... maybe I don't go there.:D

The real issue is whether the OS can handle all of the timing... which I don't see the OS as the bottleneck for implementation as much as finding code warriors to do the hack.

In regards to the comment about a little audio... 64 to 128 tracks including analog insert points... yeah, just a little-bit of audio... then synching it to images and video... cool stuff.

I checked out the http://www.daugerresearch.com/ and http://www.platform.com/products/virtualization/ sites. Anyone here at all familiar with either of these folks? Is this the right direction? hmmmmm...

Stay tuned for further details....
 
A cluster is not really meant to do what you're trying to do. A cluster is best for batch type processing. you have to have software specifically made to distribute tasks between the machines.

If you want to build a cluster using MPICH, i can help you there. I dont know if theres software available to do what you're thinking though.
 
Re: Re: Re: Re: How to "strap" 5 or 6 G5's together?

Originally posted by benixau
Don't use duct tape - use scotch tape or something - duct tape has a monopoly in the market and is trying to bully all of its competitors out of its space (including masking tape)
Click to expand...

Sounds like a software company I have heard of.
 
Here's an Idea, buy 3 or 4, AMD Opteron 4-way servers, they exist and will **** all over a cluster of G5's, + you dont need to waist money several graphics cards.
If your after a mac solution however a G5 server may be released sometime next year, but don't hold your breath.
Also Fiber channel cards are really expencive $1000AUD.
 
check out Digidesign and look at the ProTools HD Sysytems. Thier top of the line system supports 196 tracks and 96 simultanious analog import tracks. This should be more than enough for what you need and a lot easier to setup than clustering multiple macs. Also, I do not think there is an audio application that supports clustering.
 
Maybe you're doing something I'm not comprehending but without you being A LOT more specific about exactly what it is you're trying to perform I'm going to take a stab at this and say you're way over complicating this dillema. Krimson is probably on the right track.

Through the magic of SMPTE and/or MIDI timecode you can probably sync any number of machines to share the burden of your enormous audio/video project. As for the analog ins and outs latency will be an issue when you start piling on the inserts no matter how many processors you have in your machine -- you're not going to increase your bandwidth before and after the processor. Start of with one computer as a test (using for example a MOTU firewire interface or multiple interfaces) and see how many signals you can input, process and output. You can also look at something like tc electronics powercore to help with effects processing.

Anyway I'm not being very succinct but you could use multiple machines synced together via SMPTE for different tasks and maximum i/o.

Look here:
http://www.tcelectronic.com/PowerCore
and here:
http://www.motu.com
midi and audio interfaces
 
Originally posted by manitoubalck
Here's an Idea, buy 3 or 4, AMD Opteron 4-way servers, they exist and will siht all over a cluster of G5's, + you dont need to waist money several graphics cards.
If your after a mac solution however a G5 server may be released sometime next year, but don't hold your breath.
Also Fiber channel cards are really expencive $1000AUD.
Click to expand...

As much as I am a mac fanboy, I've gotta agree here. Personally I see Apple making no inroads at all into the server market apart from strictly all Mac workplaces. Anyway, thats off topic ;).

What you need to do is get a few AMD Opertron CPUs, a nice mobo and get yourself a copy of some flavour of Linux.

I doubt you will really need more than 2 CPUs, and dual opertron mobos are realtivley cheap compared to Intel Xeon etc.

What I suggest you do is give that a go, and see what happens - if you need more CPUs, start look into clusturing. Clustering on linux is probably the best of all OSes, and Red Hat can provide you with a nice GUI to guide you through it for a few $$$.

However, I'm confused as what you want to do. If you want to just render your audio and video, this setup will work v. well as long as their is a linux server available, otherwise, you will have to use a different solution.

If you are wanting to do some sort of analog clustering in sync, i'd really forget it. It will be very, very, very hard to keep it in sync and if it falls out you will spend the rest of your life diagnoising tonnes of problems.
 
have you ever heard of pro tools? you should really look into proTools HD systems (as a previous post said). getting a "cluster" of macs to process audio in real-time is probably impossible.

ProTools HD systems are pretty expensive, but building a "cluster" of macs for a recording studio is probably more expensive. and pro tools will work.

if you need any more information on digital audio recording, or using macs in a recording studio environment, drop me an email. I am a professional audio engineer who has been using pro tools for 10 years....

hope this helps...
shake
dave@mountaindogmusic.ca
 
... and now for something COMPLETELY different...

Yup, already know about PT HD... looking at 96 tracks... the worksurface issue sux. Also looking at Nuendo, a couple of DMX-R100's and 4 RADAR Nyquists. Also in the running is a Studer 950... or the "new" 950 by the spin-off from Studer... ooops, is that OK to mention?!? I have a hard time with the co$t of AMEK and SSL... plus there's the i$$ue of digital storage.

The real problem is that the whole recording industry is based upon non-standard standardization...

midi, as wonderful as it is, has SERIOUS limitations when trying to run say 16 Appogee's going out from a digital console such as the R100's digital insert points... and forget about the propogation delays and synching another 16 Appogee's coming back in on the HD system... now there's the DSP jitter that will need to be going on when the Nuendo is feeding to/from HD and/or the RADAR's. AES works, but even MADI is limited to 24 tracks at a whack.

With all of the "better" (read pro-level) software and hardware supporting Mac, it just makes sense to me to look at a total Mac solution to manage and tie all of these different systems together into a coherent homogenous workflow.

In a sense I'm looking to hybrid most of the major digital platforms/formats into a fully sync'd bidirectional system, capable of real-time DSP of high quality analog AND digital gear. e.g. a big honkin' digital processing patchbay from h&!!

Currently I have real issues with much more than 28 tracks of simultanious tracking with my 2408 and 24i on my G3 B&W with 1Gb RAM. (I know... it's only 300MHz and would see a significant increase with a faster CPU. That's part of why I'm here!) I've managed to build 64 tracks with processing and 128 tracks with no processing... but man is it a drain on the brain.

When I think of tying in all of these formats and then getting them to sync to NTSC or HDTV for video/film/foley/post?... Arrrrrrrrrrrgh!?!?!? There's gotta be a better way than having 16 completely different systems that don't talk to each other and jumping through 64 hoops to bounce stuff back and forth.

Glad to know there IS an audio contegency here!

Max
 
You're making this way too complicated. Everything you're looking to do is an everyday occurence in all the audio post studios here in LA. Your demands (as written) aren't even on the high side. I suggest you call one of the major proaudio equipment dealers and ask to have a sales rep outline a solution for you. It'd take no more than a few hours to outline exactly what you need. Just make up a list of your current equipment you have that you _want_ to tie into this system (don't list all the stuff that's useless) and fax/email it to the sales rep.

(First step, your current system is woefully out of date for current software... so chuck that. The 2408 could come in handy, but I'd even put that in a closet as backup.)

For PT, ProController 2 is due at Winter NAMM. I've seen the prototype and it's cool.

I also think you need to figure out _what you want_. Sit down and outline what steps you want to process in analogue and which in digital. Also examine what kind of resolutions you are looking for.

Also, are you doing Audio and Video? If you're not doing Video, it'll be more important to speak to who will be supplying the source material and that'll determine your needs. To work with HDTV, you don't need a specific HDTV setup-- you don't work with uncompressed HDTV to do audio work, just a compressed frame accurate QT file (given to you by the editors.) Same goes for any other video format; the demands on audio post in the video realm are small, so you don't have to spend too much there.

My 2cents
 
hey,

you are over-complicating things.

what u need is a protools HD system with Sync I/O.
configure it with appropriate number of I/O's (analog or digital).
you dont need a radar system. protools acts as your recording medium (tape deck) and handles all of your dsp. if u need a console, i suggest the Control24.
SSL's are great, but the Control24 is a fraction of the price.

what u are asking for is not uncommon. syncing protools to LTC (or whatever you are using) is pretty solid. I've been syncing protools to many sources (LTC, 29.97df/ndf) for commercials, films etc.

You Said:
midi, as wonderful as it is, has SERIOUS limitations when trying to run say 16 Appogee's going out from a digital console such as the R100's digital insert points... and forget about the propogation delays and synching another 16 Appogee's coming back in on the HD system

WHY ARE U RUNNING MIDI FOR DIGITAL AUDIO? midi has nothing to do with digital audio. if u are using it to sync two digital systems together (like cheap time code), then i feel for you. Midi's SERIOUS limitation is that it's not digital! LOL

YOU ALSO SAID:
now there's the DSP jitter that will need to be going on when the Nuendo is feeding to/from HD and/or the RADAR's. AES works, but even MADI is limited to 24 tracks at a whack.

DSP jitter does not "need to be going on" at any time. jitter is an artifact from digital clocks that are not synced properly. jitter is a timing error. why are you feeding tracks to/from nuendo/HD/radar ?

DO YOU KNOW WHICH DIGITAL DEVICE IS THE CLOCK MASTER IN YOUR SYSTEM? IF YOU DONT, THEN MAYBE YOU SHOULD INVESTIGATE THAT.


have you used protools before ? have you had your first session yet?
 
Originally posted by shake
Midi's SERIOUS limitation is that it's not digital! LOL
Click to expand...

Alright since I think I first mentioned MIDI here let me clarify this:

MIDI certainly is digital AND distinctly a different beast than digital audio.

What I was picturing is one machine running a bunch of soft synths while the other handles audio and sequencing, because that's what I've got going on using the MOTU midi timepiece. It's a flawless sync. Yout may want to look at the midi timepiece AV http://motu.com/english/hardware/mtpav_usb/body.html
argh, frames

But certainly if he's working in video post he's going to want a SMPTE master.

I'm not as familiar with Pro Tools as I should be but, yeah it seems everyone else who's posted here basically has the right idea. MaxMacHac...... what's the confusion. Is everyone not getting what you're trying to do, or are you not following this? You've got my curiosity at least.
 
another possible alternative

though i'm not 100% it is supported by the major sequencers (ex. logic, pro tools, etc), you may also want to investigate open sound control, which is basically a protocol for networking computers via ethernet, and has many advantages over midi (OSC can transmit data at 10 mbps, versus MIDI's 31.25 kbps). some info can be found here:

http://www.ixi-software.net/content/info/osc.html

i'm not sure if this is a solution you're looking for, but you could theoretically use it to sync together one master computer running sequences transmitted to slave computers processing audio or running soft synthesizers, which would be a whole lot easier than actually clustering them.

i personally use OSC within a program called max/msp -- audio running from a powerbook g3 is fed into another powerbook g3, which processes the audio and spits out audio analysis data (volume, pitch, etc) over OSC to a powerbook g4 for video generation. it works great once you've gone through the hassle of setting up the internal network (not that hard).

hope this provides some help,

best,
np
 
What you need is one of these babies, + a whole load of other stuff that Mackie make:)
 

Attachments

  • hdr2496_main.jpg
    hdr2496_main.jpg
    8.1 KB · Views: 508
Thanx for all the interest in the topic folks! Maybe I just ain't quite expressing myself too clearly here. Lemme' see if I can 'splain it a bit more better.

Yes I've done a PT session, albeit LE. Mixing with a mouse is rediculous. Not enough tracks of CPU horsepower is the downfall of almost all of the "prosumer" level audio systems out there. DSP is all reliant upon the overall CPU strength... correct?

If, in building a "high-end" studio, one is to be compatable with: PT LE, PT HD, Nuendo, Cakewalk, DP, etc., each of these systems' files have to be converted into a common medium. e.g. broadcast wave. But even converting to broadcast wave, all of the session detail is lost as each system is proprietary in nature.

So..... why not build a system comprised of each of these systems? If in building this system, you use the additive power of the multiple processors, in theory, you overcome the CPU power issue. Correct? That's part of what I'm investigating.

Using MIDI to sync everything was suggested... not quick enough to overcome the inherent A/D lag as data is created and transferred... e.g. jitter. That was the point I was making.

As far as the actual transfer of digital audio signal... 10base-T is way to slow to transfer the data in real time. 1000Base-T might come close, but 40Gb should definitely do the job. Those switches are available... (Va Tech). Cisco announced yesterday that they are entering the foray of the 40Gb switching arena with upgradable devices.

The RADAR Nyquist system is hands down far superior in sonic quality to any PT system... yet, I will still need to be able to handle PT HD files in native format. To be able to integrate HD w/RADAR I can use the Sync I/O, but if the data can be placed on a common highspeed data buss, then I don't have the external issue with artifacts from the sync function to compound issues that will occassionally occur within the data buss from the other systems being tied in and integrated... just one less thing to buy or worry about... maybe.

Now we get to the really interesting point of it all... control surfaces and analog inserts. In my 40 years of music career and the last 5 of those as a mobile tracking studio owner, I've yet to hear anyone with even a year of tracking or mixing experience say that there is any DSP card or plug-in to have nearly the same sound, as say a 160 on a kik. Let's face it, there are some analog devices that will NEVER LIKELY be duplicated in the digital realm. So how do you integrate an analog insert device into the digital domain and not run into artifacts? OK, 4 or 5 consoles are capable... but they aren't compatable with PT HD. Arrrrrrgh! So, there's the key issue for arguing for a system which is capable of real-time D/A-A/D conversion, but still compatable w/PT HD.

Not to bash Mackie, but their control surface is junk IMO. I'm a hands-on engineer. Three buttons to get a knob to turn what I want is worthless compared to a true professional console that I can reach for the right knob and just turn it. The same applies to the Control24 and ProControl from what I've seen in the lit and from talking to a couple of other engineers. That's also a concern I have about the R-100, although not as much. I've talked to two friends who I respect that have used both, and said that sonically, the R-100 smokes the PM1-D. Bang for the buck, either of these beats the Studer, SSL API or AMEK.

I got a reply from an email I sent to Dean Dauger of http:\\www.daugerresearch.com
-----------
Hello,

Thank you very much for your interest! Yes, I believe it is also
applicable to digital audio. As with any application of parallel
computing, it's a matter of partitioning the problem into pieces that
can be divided among the processors in the cluster. There are usually
application-specific issues that need to be dealt with, but none has
yet been insurmountable.
------------

While I'm not convinced that I have to build the cluster, I just can't see why it won't work, and work well. A side benefit of a cluster would be the lack of need for a storage server... that and the backups should be fast as lightning.

Do I 'splain it more better now?

Thanx,
Max
 
LE vs. TDM

I haven't followed this post until today, and here's what I think:

You have worked with Pro Tools LE. That is a weak beast and I'm not surprised that you want something more powerful. When dealing with an LE system, your DSP is based on your CPU power. A TDM system, like the Pro Tools Mix and HD systems has DSP cards with the system that take the load off of your CPU. That is what allows for you to run over a hundred tracks with plugins vs. 32 in LE.

A G5 dual 2.0GHz with and HD-3 (Pro Tools HD system with 3 PCI DSP cards) and the interfaces to connect all your gear will be more than adequate and work with your other DAW software. While it might be great to have such a clustered system in the future, it's overkill for what you want.

Add a control surface (Pro Control) to this so you don't have to mix with a mouse and you'll be set. Basically, one rack of Pro Tools gear (AD/DA, MIDI I/O, Sync I/O) and it will all work out of the box. But if you want to spend at least twice the cash (and Pro Tools HD systems don't come cheap) on your G5 cluster and the time researching it and setting it up, then that's your call...

In any case, good luck with whatever your choice is. If you do get it working, let's have some benchmarks on that puppy!
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back