These are exactly the issues I was wondering about.
eg: if you need very low latency, you probably will have to use one single computer eg: if you need to tightly sync your output(s), you will probably have to use one single computer.
So you've answered some of my questions despite yourself:
your question can't be answered seriously without an idea of the project you are trying to do:
I can't give an accurate portrait of the project because it will be a workshopped piece that won't have a final form (the performance itself may not be 'final'). A few things I do know is that the same system will have to be used with a traditional performance, (eg. dancers triggering sensors in a more or less predictable manner) and also for an installation piece right afterward in which the audience will be triggering reactions via sensors in a more unpredictable way.
Extremely low latency will not be an issue. I won't be doing stereoscopic imagery or anything like that, and I assume that the speed of a (wireless?) ethernet network will synchronize actions sufficiently.
I'd really be interested just to hear about things people have done in the past, and what kind of computer power they found they needed.
in other cases it might be sufficient to use several pcs, which could be more fun, because for *most* things you will not need full calculation power of all the computers - so you can split your setting and run e.g. 3 different installations/measurements/whatever at the same time.
Does having several computers give any actual power benefit over one behemoth. Here it sounds like the main advantage would be organization; getting different elements into discrete places for easier management.
Performance wise, how would it compare to run GEM doing (for example) two 3D animations, one using video capture from a webcam on an AMD Barton 2.2GHz (XP3200+) with a gig of ram and an ATI 9800 to running the animations on two separate Intel Celeron 2.0 GHz w/ 256 MB Ram and an onboard video controller? (These systems are quick looks at the kind of stuff I might end up with, 6 months from now I'm sure my money will get me a lot further, but GEM will also do fancier stuff and need more power).
all of them. as for noise: not only the fans are important, but you should also consider a sound-absorbing interior and air-flow optimization.
the unfortunate thing is, that silent machines tend to be big and heavy.
What situations have you found noise to be a big problem in, and in what situations has it been more important to have small form factor (as the two seem to be mutually exclusive). Are there any unobvious advantages to SFF and low-noise other than the obvious portability, hide-ability and, well, low noise? eg. have you ever had problems with loud machines interfering with microphones?
Of course, that's a lot of nitpicky stuff, and you may have better things to do than answer it all, so don't sweat it; you've already been a great help.
Thanks, Ian Smith-Heisters
-------- Original Message -------- Subject: Re: [PD] Buying PCs for PD: Quantity or Quality From: zmoelnig@iem.at Date: Tue, March 30, 2004 11:09 pm To: "Ian Smith-Heisters" heisters@0x09.com Cc: pd-list@iem.at
Zitiere Ian Smith-Heisters heisters@0x09.com:
hi. your question can't be answered seriously without an idea of the project you are trying to do: eg: if you need very low latency, you probably will have to use one single computer eg: if you need to tightly sync your output(s), you will probably have to use one single computer.
in other cases it might be sufficient to use several pcs, which could be more fun, because for *most* things you will not need full calculation power of all the computers - so you can split your setting and run e.g. 3 different installations/measurements/whatever at the same time.
as for Gem: putting several gfx-cards (e.g: 1xAGP, 1xPCI) into one single computer *should* work (at least nvidia says it does) - but i have never tried it : i've never needed more than 2 parallel screens. again: if you need your outputs tightly in sync (eg: you want to do one big contiguous image or some stereo-image with polarization) than i don't think that using several computers that do the actual rendering is a good idea.
Are there any other things y'all have found useful? Extra quiet
fans?
Small form factor?
all of them. as for noise: not only the fans are important, but you should also consider a sound-absorbing interior and air-flow optimization.
the unfortunate thing is, that silent machines tend to be big and heavy.
mfg.ar IOhannes
PD-list mailing list PD-list@iem.at http://iem.at/cgi-bin/mailman/listinfo/pd-list
I'd really be interested just to hear about things people have done in the past, and what kind of computer power they found they needed.
Does having several computers give any actual power benefit over one behemoth.
Obviously one fast computer can make more complex sounds and then distibute those over different speakers, as opposed to a bunch of slow computers making a lot of simple sounds independly. It all depends on what kinds of sound you produce. I regularly overload my dual 2 GHz athlon box during performances making only sounds. But those are all synthesized on the fly. For sample playback you need a lot less CPU so if you want a lot of simple voices more or less out of sync a couple a slow computers will do the trick.
What situations have you found noise to be a big problem in, and in what situations has it been more important to have small form factor (as the two seem to be mutually exclusive). Are there any unobvious advantages to SFF and low-noise other than the obvious portability, hide-ability and, well, low noise? eg. have you ever had p roblems with loud machines interfering with microphones?
In a studio environment yes. A reasonably quiet of-the-shelf computer (not the cheapest allready) I have found not to interfere in a theater setting, with all the lights and aircondition etc. But in a recording studio with nice mics you really notice them. It all depends again on the kinds of sound manipulation you do and in what kind of environment. A noisy PA for instance will mask more than a high end one. I you have the good mic and you want to have it take samples, so you are gating the input, you will notice noisy input a lot sooner than if you only amplify it since noise becomes more noticeable when it is interupted. That can be nice or not depending on your esthetics.
For GEM I would definitly get a seperate card but not the most expensive/newest one.Onboard graphics are not meant for 3d stuff. But for pdp they might be ok, I don't know
Gerard
On Wed, 2004-03-31 at 07:13, Ian Smith-Heisters wrote:
Does having several computers give any actual power benefit over one behemoth.
In general I would say yes (assuming the work can be distributed evenly). Multiple systems means you have duplication of more system components such as L1/L2 cache, FSB, drive controllers. However, you pay for the extra hardware.
Resource contention for L1/L2 cache can be a significant bottleneck. Unfortunately its not easy to analyze... Proper use of SSE can mitigate the problem with streaming media applications by explicitly handling transient data. Not sure how well PD/Gem/etc will handle this.
Performance wise, how would it compare to run GEM doing (for example) two 3D animations, one using video capture from a webcam on an AMD Barton 2.2GHz (XP3200+) with a gig of ram and an ATI 9800...
Also I would consider the relevance of graphics acceleration to the task. Some things benefit greatly from the GPU, others do not, such as particle systems.
What situations have you found noise to be a big problem in, and in what situations has it been more important to have small form factor (as the two seem to be mutually exclusive).
Here is a system which I have used which is small, reasonably quiet and reasonably priced:
http://www.emaxasp.com/CBS/details.asp?item=SAMBA1845
I don't think its any louder than my laptop and is small enough to fit in a backpack.
-
Andrew (Andy) W. Schmeder -
gml＠xs4all.nl -
Ian Smith-Heisters