Final Gather Map on a Network

[JakCarver]

Funny you should bring this up, I had the same question floating in my head a couple days ago. So, I went ahead and tried it. It worked pretty good. I thought that there would be a conflict when 30 machines would be modifying the FGmap file on the network, but appearently it works. Heres what I did: (Keep in mind I used XSI to do this)

I first made a prepass so to speak, of my FG by setting my antialiasing settings extremely low, I believe I used something like -2/0 or something like that. I set 30 machines to render different frames of the archviz animation, i.e. PC 1 rendered frames 1-30, PC 2 rendered frames 31-60, etc, etc. Once all computers finished rendering, I had a nearly 40MB FGmap file on my network drive. I then froze the FGmap file in my render settings and proceeded to render out the animation with full AA settings. This morning I came in to the lab and found that everything seems to have rendered correctly except for some strange things. I found that some frames has inconsistent FG points and it seemed to be patterned to certain frames. I came to the conclusion that yes, you can infact network render one FGmap file using many rendering stations, which is awesome. The small issues I enctountered I believe were user error when I set my FGmap pass, I probably missed one computer or something so that when the final render occured, it couldnt find missing FGmap points hence my strange FG flicker/jump issue on a certain number of sequential frames. Go ahead and try it, should work with Maya as well. Just make sure you quadruple check everything-you dont want to come tomarrow morning only to find that you left late last night and forgot to hit the Render button

[wizzackr]

I’m posting here to see if anyone has any experience with an idea I’m planning on implementing.

To save time with Mental Ray & Final Gather for an architectural visualization, I thought of turning rebuild to ‘off’ to use previous final gather points to speed up render times. Nothing special with that idea.

The question is this: Has anyone tried this while rendering with multiple machines across a network; where all the render nodes are accessing and writing to the same final gather map simultaneously?

I’m not sure how this will work out, and I’m afraid that that they will conflict and I’ll get crazy render results. Should each node work off their own final father map instead?

Any input is appreciated!

Cheers,
Bill

Funny that you ask this question. I've been pondering a lot of times, but it somehow always occured to me when a deadline was looming and I didn't/couldn't risk it.

I figure it will go ok as long as two machines will not accidently try to write into the map at the same time - then, I assume, things will go caddy whompus. No idea though. If you do a test make it decently complicated, as FG maps can get pretty big, writes longer and the chance of screwing things up higher...

Definitely keep us updated!

[JakCarver]

Yes I tried this over the last couple of days and it works, at least with XSI.

[wizzackr]

Yes I tried this over the last couple of days and it works, at least with XSI.

Sorry, buddy, your answer didn't show at the time when I posted (could be our network, as the forum is slow as molassis, too, today). I'll definitely try it out - thanks for the heads up!

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.