renderfarm rendertimes

[super gnu]

hi all. i have a scene here that takes approx 40 mins a frame on my i7 920 ht (3.6 ghz oc) with 24 gb ram.

ive sent it to rendernation ( lovely guys)

but im surprised at the rendertimes they have... around 35 mins a frame on their dual six core ht machines (2.93 ghz) with 16 gb ram.


since time in this case is literally money, im trying to figure out why they are not squashing my rendertimes like a bug.

the scene uses an average of 8 gig to render and a max of nearly 10 (according to their diagnostics) on my machine rendering straight out of max, it seems to use more - around 15 gb, but thats with the whole of max running too.

they have the same versions of max and vray as me...


any suggestions what i can tweak?

[robdeja]

When we have rendered at rendernation I have noticed that their dual 6-core machines render 30% quicker than our dual 4-core in house machines so it's strange that you're not seeing much difference in your case.
Could it be that it's getting hung up on presampling displacement? Have you changed the dynamic memory limit to suit the 16gb of ram available on their machines? Maybe it could be a single bucket taking a long time to complete which due to your core speed would complete faster on yours?

The other thing it could be is that when the frames start rendering a lot of machines request the texture maps at the same time and slow the network transfer down. They use deadline and there is an option to limit the number of machines starting jobs at once which helps in this situation.

[JOAN_LORD]

I find it crazy how quick you can get something to render if its optimized, and if you have the time, running from the system drive can cut render times by almost 40%! obviously there might be other things like there not increasing the setting to tell Max how many threads the machine has, I can't remember what it's called now damn-namit! It's in the LC settings...

Dynamic Mem has a part in it too, if it's set to low max won't use the full potential, I have it always wacked up to about 20gig, it never get's higer than 15-16 but it's handy. Apprantly using Glossy rays in LC helps speed up render times too.

[super gnu]

well i pretty much discounted memory issues as i have dyn. limit set to 0. and it only uses 10 gig max anyway. regarding the threading, ive precalced the lightcache here.. its just loading files. also i dont have any displacement or much fancy apart from a crapload of maps. so the obvious culprit would be network speed for transferring the maps.. however id then assume that after the first frame on each machine it would speed up drastically, as the stuff is al held in ram.. however this doesnt seem to be the case.. very mysterious.