Which machine for DR speed?

[simmsimaging]

Hi all -
I have a small (3 machine) network for use with DR, which are very different in terms of perfomance. Each one is basically twice the speed of the next one.

My question is this: in terms of speed, when using DR does it matter which machine I run the render from, or will it be basically the same speed in the end? The reason I ask is that my fastest machine is not the one I am normally working on and running the renders from, but wondering if it should be?


Thanks!
b

[ruud3dv]

When using DR all machines will contribute.

If you're using lightcache however, (up to now) all slaves will have to calc the lightcache on their own, before they will join in IR or rendering buckets on the main computer. So it's best to work on the fastest system, lightcache will be done quickest there. So whenever the slaves are done with LC they can join in the rest of the calculations on the main computer. Otherwise you can end up with slaves waiting for the main computer to finish the lightcache. After the lightcache stage you'll have the benefit of the faster computers though.

EDIT:A possible future option has been mentioned where you could choose which node will do the lightcache, after which it'll be send to all nodes. Hope this makes it.

[simmsimaging]

Thanks for the reply, that helps a lot.

Just to make sure I understand though: if I am *not* using LC then it wouldn't matter which machine I initiated the render from?

Thanks again!

b

[ruud3dv]

That's correct, as far as i know all other GI-types can use DR. Not sure about caustics calculation though.

[simmsimaging]

Thanks again!

b

[Jynks]

so as of sp2... lightcache can not DR render.. it needs to be rendered on every system?

So using other GI solutions is the way to go for DR Rendering?

[vlado]

The light cache is not distributed yet in SP2. What we will most likely do, is simply take the light cache from the machine that calculates it first, and distribute it to the other machines, rather than distributing the light cache itself.

Best regards,
Vlado

[Jynks]

I see thanks for the tps...

so if you wish to use network rendering and use light cache it is better to use backbrunner and have each system render a complete seperate frame?

[stevesideas]

Caustics calcs cant be DR'd. Is this likely to change ?