I just wanted to share some interesting info with you, in regards to the Render Farm I put together..

Here is the hardware setup

Master PC: Compaq Persario - 2.0Ghz / 1GB Ram - Windows XP Pro

PC#1. AMD Barton 2800+ OC'ed to 2.0GHZ (512MB Dual Channel Memory) - Windows XP Pro
PC#2. PIII 1000    - 1 Ghz / 768 MB Ram both @ 133 Bus Speed - Windows 2000 Pro
PC#3. PIII 800EB - 800Mhz / 768 MB Ram both @ 133 Bus Speed - Windows 2000 Pro
PC#4. PIII 800EB - 800Mhz / 768 MB Ram both @ 133 Bus Speed - Windows 2000 Pro
PC#5. PIII 700EB - 700Mhz / 768 MB Ram both @ 133 Bus Speed - Windows 2000 Pro
PC#6. PIII 600EB - 600Mhz / 768 MB Ram both @ 133 Bus Speed - Windows 2000 Pro

I used Omac2's Bryce 6 Benchmark www.alexpensotti.com/bench2.html

Here is the results I got using Bryce 6.1

First I rendered the scene using PC #1 standalone and got 18:56
I rendered the scene again using PC#1 but through Bryce Lighting and got time of 20:00 a difference of 1:04

I than proceeded to add the next pc and rendering the same scene recording the results as I went along.  The final test utilizing all PC's with a render time of 8:37 a difference of 11:23 or 54.49%.

Whats interesting about the final Render Time of 8:37, is the same as  an Athalon X2 4800 which is  a 2 Core CPU.

Now unfortunately I did not have an Animation benchmark file to test with...would be interesting to see how that would turn out.

Now while I was doing these tests...  I decided to throw Carrara into the mix and see what kind of render times I would get.

I used a default scene included in Carrara and did the some render tests...one thing I'd like to note is, network rendering in Carrara is different than Bryce.  You have control over Tile Size 32-128 and how to distribute the job load over the render farm... No Load, Odd, Even.

Using the default scene I did a bunch of render tests and the difference between a single PC and all the PC's in the render farm only yielded a savings of 4 Seconds!!! Yup 4 Seconds.  Not much.

So I decided to test one of the included Animations which had something like 768 Frames.  I decided this time I would skip the single PC Render, I was pretty sick of doing all these tests.  So I concentrated on all PC's rendering and played with Tile Size and Render Load settings.  And the end result was very interesting.  The Longest Render time was over 1 hour and the shortest render time was 36 minutes.  That is quite a big difference in render time by simply tuning settings on how the network render should be carried out.

So whats the conclusion? Well I think first of all what I've build is not economical, when you consider you can get the same results with a Single High End PC, with a Dual Core processor.  I would like to do some Animation tests in Bryce and see if it does any better.

Still almost 55% savings in rendering time is not bad from PC's that I built from Junk Parts.

I think it may be possible to get better results with the same software, but maybe with better Network Rendering engine for Bryce.  What do you think?

If it works and cuts your times, it works!

This business of farm rendering is not all that straight forwards....  Like you said just settings can effect the render time significantly....   You should have gotten much better time breaking the bulk by 1/6th....    Yet it does not work that way...  Perhaps it is what you suggested.. better network and distribution software...

 Why don't you get in touch with one of those pro render farm places and ask what they use..  you never know what they are willing to tell you...   If I had the money I would venture to organize a search party for the better render farm with you....  I feel the I/O hardware is also a key....  these computers are made to work with thses huge OSs...  Perhaps special boards, more compact and less clutter between the   input - Ram - PU - output

Thanks for that great info Analog-X.

I do think Bryce needs a re-write for its network code, but Im no expert here. I didn't get very good results with L2.0b and B6 when I tried. It also seemed that the larger the scene, the slower the network would go. Maybe clients are waiting to much for other clients on network(im guessing). My other 3D app gets a 40% boost, from just one other similar client. I wish Bryce would get that...

I did look at professional render farms and one in particular was boasting, 3.0 Tera Hz Computing Capacity with 1000 CPUs of Xeon 3.0 GHz and Minimum 2 GB Memory for each of those CPUs.

Also they use customized software, and mainly deal with 3SMax and Maya.

I was hoping to not only speed up my own renders but also help out others in the community.  If I had the budget I would probably build a few Dual Core Processor CPU's, but were looking at around $10,000+ to build a decent set.

I think Bryce Lightning was added as a nice feature to help a little, but not optimized to its full potential.  Although Pentium III's are now oudated and slow compared to whats available now... they do far exceed the minimum requirements from the Bryce 6 page, which are Pentium Processor or Compatible,700 MHz (1 GHz or faster recommended),512 MB RAM min (1 GB recommended)

I think Bryce and Bryce Lightning could be optimized even more to get better render times with the hardware setup I have.