[deadalvs]

hi Dieter !

and welcome to the forum !

... I can't help but thinking there is a lot of redundancy in the images.

i am not sure what You want to say with this sentence. as far as i know, redundancy is a «feature» that is for savety, let's say if You have no more current, there's a battery (redundancy !) that instantly gives current to the computer, so it just does not shut down.

thus, i am not sure what You want to say with this...

* * *

next point. distributing a rendering over the network can take some time. You have a lot of bottlenecks like GLAN, slow harddisks, priorities, (under mac no virus scans... ) and so on. this all takes time and like this, You will never get (in Your case) the factor 2. but, the longer the rendering takes, the better the ratio will be, based on the fading significance of the distribution time. it might be that half an hour may be not enough time to get decent results on that factor. (plus, You certainly know that SLs grow exponentially, thus can not be doubled by rendering with two computers...)
me for myself, i'd «calculate» with the factor 1.8 (linearly) when rendering with two identical computers.

i have not tried again network rendering on macs since i have given away my second mac to my dad. but as i remember it did not really work properly.

especially when working on macs, i'd wait for the update to maxwell 1.2 which should be out «soon» (one to two months) because stability is an issue. if You can not get it running in a day and it works totally stable, it's unusable for professional work... my opinion. just start the scenes on each mac and merge manually in the end. that's not too much work.

but NL's working on it. let's hope for the best.

[deadalvs]

ah, i see.

-->

1) yes, there is 100% differentiation between two renderings, this is what the seed value is for. behind the seed value is an algorithm that distributes rays for each .mxi. it's the same approach as with even/odd frames, just with pixels. but since rays interconnect with geometry on the sub-pixel level, it just «looks» as if there was no differentiation!
but that's a problem of resolution.

2) it is the great advantage of maxwell that always the whole image is rendered, and only the quality increases. like this, renderings can be aborted when You're satisfied! this is a different methodology of maxwell. i like it !

yeah, do that test.

so it on one machine with 100 minutes, check the SL
do it over the network with 55 minutes, automerging, check the SL
do it on each machine individual 50 minutes, manually merge the image, check the SL.

please post these values...

[Maxer]

Maxwell can't do distributed rendering which is what you’re describing, it's an unfortunate result of the way the engine works which requires the entire scene to generate the correct result. The reason that you don't see a 100% increase in quality when you merge two MXI's together is because of the diminishing returns you experience as each SL is reached. Cooperative rendering is supposed to speed up this process, however if you have two images with SL's of 10 the combined image wouldn’t produce an SL of 20 it would be more like and SL of 11.

[b-kandor]

The easyist way to think of it is: 4 machines rendering co-op for 10 hours will produce (approx.) the same image as 1 machine rendering for 40 hours. Not counting the merging time.

Kandor

[deadalvs]

1.) Ok, I got it. So when you say that there is a difference on the subpixel-level, it means these differences might become visible when using higher output-resolutions?

to 1... no, You will not see it. but i'm not really sure how it works...

* * *

As far as I understand it, when submitting a job to a renderserver, the number of machines available is determined and the seed-value parameters are automatically set up during distributing the mxs to the nodes.

correct.

Only problem are the bugs in the manager Software and the bugs in the mac implementation.

correct.

* * *

just also export .tifs so You have an additional backup file if in any case, the .mxi were garbled. this might be very seldom, but possible: when the computer crashed when writing the .mxi file.