Is there any point in a G4 render farm?

Question

Hi all.
I've been getting quite intrested in the idea of setting up a very small render farm laterly as I have the option of getting some old G4s for free.
I currently have a 2.0Ghz DP G5 as my main mac running FCP2 and Shake and was already grabbing a G4 466 digital audio to set up as a file/print and back up server running tiger server, a postscript RIP and retrospect. I have the possibility of getting atleast one more G4 466 DA and maybe a couple of G4 400 AGPs with Gigbit ethernet cards. As the main use would be farming out DV to mpg2 compression via compressor and shake renders (though not large shake stuff to begin with as very new to this) is there any point in looking into this or am I going to need to lots of G4s to make a real improvement in render times.
Also is RAM a major consideration. For example if I got 4 G4s with say 512Mb ram would it be better runing all four or using three but nicking the fours ram to give more ram in the three machine set up?
Any advice would be appreciated
Cheers
Steve

Answer 1

I vote no. G4 isn't the problem so much as 400mhz is. The ram shouldn't be a problem.
But you could test if you set up Compressor on one G4. Compress a job. Compress the same job with the same settings on the G5.
If the G5 is 4 times (or even 3, since time is spent sending the render data over the network between the G4's slow system busses) as fast the single G4, a render farm will not help you.
Using the G4 and G5 together in the same cluster does not work because Compressor looks at all processors and divides the job into twice as many segments. In this case, the G5 finishes it's two segments years before the G4's and then sits around waiting. You could try splitting the G5 into 4 instances, but then you have 8 slow processors attacking your job. Not very efficient.
Good luck though. You should test the one G4 just so you know. Report back if you do.