Page 1 of 1
Distributed rendering
Posted: Mon Apr 07, 2014 1:40 pm
by Igor Posavec
Hello,
i think Distributed Rendering is from the production aspect one of the most important things. We have here dozen of CPUs rendering production shots simultaneously with Vray, Mentalray and since latest even with Corona.
Since Maxwell is, buy the nature, very slow renderer (or at least much slower the those above), the only way to save the time is to engage multiple CPUs to help. Since we can buy everything - cpu, ram, man-power and licenses, and we can not buy only one thing - the time, it would be most valuable to compensate the time waste and the disadvantage with distributed render processes.
thanks in advance
best regards
Re: Distributed rendering
Posted: Mon Apr 07, 2014 2:30 pm
by Hervé
Re: Distributed rendering
Posted: Mon Apr 07, 2014 3:17 pm
by Igor Posavec
Subject: Distributed rendering
Hervé wrote:how about RTFM..?
Hello Herve,
hm, to be honest i was sure this is only for frame rendering. Since the description is very short and not illustrated - does this sentence from RTFM:
"..The contributions of the machines are merged to a single image."
means i don't see the feedback all the time, but only at the end when the frame rendering is over i get some merge? In other words - each render a segment and it gets "merged" at the end into an image (which is almost useless).
Or do i see really the speed improvement with support of other workstations instantly on the screen?
I have the standard floating license, and if the DR works like expected (realtime update during rendering) this means i should update it to the 5 pack bundle, correct?
thanks a lot for help!
best regards
Re: Distributed rendering
Posted: Mon Apr 07, 2014 4:50 pm
by AlexP
Igor Posavec wrote:Since Maxwell is, buy the nature, very slow renderer (or at least much slower the those above)
That's just not true, just make same scene (watch for white value in materials) in Maxwell and Vray3 in progressive mode and you will see yourself. And mental is just incredibly slow in all modes.
You have to click preview to see what's on all nodes together.
Re: Distributed rendering
Posted: Mon Apr 07, 2014 5:42 pm
by numerobis
it would be very useful if FIRE (or maxwell in total) could use distributed rendering... like v-ray progressive rendering or corona
so... +1

Re: Distributed rendering
Posted: Mon Apr 07, 2014 5:49 pm
by Igor Posavec
Hello AlexP,
thank you for the feedback! I actually didn't want to start discussion about render speed, it is not necessary. In our case, and it may be shaders and setup specifics, it is unbearable slow. This is the reason i am interested in DR solutions.
I just wanted to ask if it the Distributed Rendering in Maxwell really delivers the same/similar kind of visual representation (output) as we see it in Vray/MR (bucket rendering or similar parallel work on single image and not invisible rendering and merging tile pictures at the end, as the descriptions in the RTFM my be interpreted)
thanks again!
best regards
Re: Distributed rendering
Posted: Mon Apr 07, 2014 6:33 pm
by Mihai
Have a read here:
http://support.nextlimit.com/display/mx ... g+Workflow
.....and also you can at any time press the Preview button in the Monitor of the network rendering which will then immediately merge all the MXI files from all your computers and send the result to the Monitor. This way you don't have to wait until the final SL is reached.
If you've read about the concept of SL in the beginning of the docs (
http://support.nextlimit.com/display/mx ... in+Maxwell), then you realize that there is really no "final" render, and no buckets, you don't have to wait until each of the buckets in your render has finished until you can finally use that image in your compositing app, you can start at a low SL, render the sequence, then resume the render, while you can now start doing your compositing/post work on the low SL version which will just be more noisy. Then, when the network rendering reaches a higher SL, you simply re-import your images into your compositing app, and you get to keep all your edits, effects etc. In essence, start doing the comp/post work while Maxwell is still rendering. This is explained in more detail in the link. (
http://support.nextlimit.com/display/mx ... g+Workflow)
Re: Distributed rendering
Posted: Mon Apr 07, 2014 8:32 pm
by AlexP
Igor,
In Maxwell it doesn't matter if you render job using network tools or using separate jobs on separate nodes. You can always merge files and get better result. Merging by hand is done by "merge" command inside Maxwell window.
Re: Distributed rendering
Posted: Wed Aug 20, 2014 6:32 pm
by Igor Posavec
Hello
thank you all, i have purchased the 10 seats and the distributed rendering does really the excellent job concerning both the speed and quality. It is so far very stable DR, maybe one of the most stable i have had so far.
The only negative side is the Preview window, which is very clumsy (i have manually to clean the temp folder in order not to see by each new job previews from the older jobs. This "shadow" temp images should be somehow deleted automatically after the job is done..). Some kind of auto-preview all 30 sec or so should be actually comparable to what i imagine a good, reflective preview.
thanks
best regards
Re: Distributed rendering
Posted: Thu Aug 21, 2014 3:51 pm
by Mihai
The previous job folders SHOULD be automatically deleted if the merging has gone ok. Unless you've set in the preferences of Node/Manager/Monitor not to delete temp files after merging....