vray DR vs backburner

[Fat Tony]

Does somebody have exp on speed in both backburner and DR. i understand the differences, but i would like to know if there is any speed advantage in either case.
Thanks a lot

[NicoC]

Hello Fat Tony,

We use both quite intensively here (I'd rather use Deadline, but that's another story). You have to see BackBurner system (Manager) as a jobs queue managing tool, nothing more. It often gets the term "distributed rendering", but not in the same way -in my opinion- as VRay Distributed Rendering which actually gathers the power of several machines to render one image. I find this double-appellation pretty confusing.

Well, if you can use DR and BB, then you have several possibilities:

1) submit all your jobs to only one machine (one BB Server, VRay DR master machine) that will "invoke" the others (VRay Spawners) -> images are rendered in queue, one by one, but each one uses all your machines
2) submitt all you jobs to all the available machines (BB Servers), with no VRay DR involved -> N images are rendered simultaneously by N available BB Servers

I'd say the way you'll choose depends on your needs. Method 1 is cool when you need images as quickly as possible, to do some post work on each image as it's rendered. Method 2 will globally need the same global time to render N images ... you'll wait longer to get an image, but you'll get them all. But I guess it depends on how homogeneous your rendering machines are.

Side note: a pity that VRay settings aren't exposed in BB Monitor ... I guess I have to switch to Deadline for this, right ?

[Fat Tony]

yes i know about about deadline but for us it would b overkill i guess, it seems to technical for the load we produce. To what you said above, that was also pretty much my conclusion. i will be using BB probably for final anim, as iti s safer if some machine will go crazy, but DR is just more handy and quicker to just hit it, since i anyways dont use local host.
Thanks a lot

[anthonyh]

hi,
DR i find useful for producing big-res slow renders as it can get the job done faster eg. a bunch of still images. we use backburner for animation on 1 machine per frame because in practice it`s quicker than using all machines DR on 1 frame. that`s on our network with our hardware of course. maybe some mega networks don`t have some of the overheads that seem to bottleneck DR for animation. if you have any issues with 1 machine stuck with long rendering buckets it means all of your other DR machines are now dormant until that one machine finishes where backburner all other machines may continue to render while only 1 may be stuck. backburner is a little crap but it gets the job done although when we moved to 2016 it wouldn`t run unless it was installed as a service which means you don`t get local access to the manager interface. ie. a bit crap.
anthonyh

[Fat Tony]

hi,
DR i find useful for producing big-res slow renders as it can get the job done faster eg. a bunch of still images. we use backburner for animation on 1 machine per frame because in practice it`s quicker than using all machines DR on 1 frame. that`s on our network with our hardware of course. maybe some mega networks don`t have some of the overheads that seem to bottleneck DR for animation. if you have any issues with 1 machine stuck with long rendering buckets it means all of your other DR machines are now dormant until that one machine finishes where backburner all other machines may continue to render while only 1 may be stuck. backburner is a little crap but it gets the job done although when we moved to 2016 it wouldn`t run unless it was installed as a service which means you don`t get local access to the manager interface. ie. a bit crap.
anthonyh

that actually is one thing i didnt thought about, the issue with stuck buckets. This seems legit for this case actually. I am just nervous to see only so fewer buckets on frame so you know, i assume it is slower somehow. But that is a very good point. We also have quite equally setup nodes so shouldn't be a problem.
Thanks a lot !

[dean_dmoo]

I've stopped using DR and just render per node with BB. For high res I too often ran in to issues with a stuck bucket with DR.For testing I just use progressive. DR with progressive just seems like a waste of time if you have multiple images to do.

Recently I've started to render the same image twice, once with progressive with a max time (just to make sure I have a good enough image the next morning) and once with bucket (if needed) just so I can get a clean renderid pass for photoshop....and a slightly cleaner shot as a bonus.

I have 3 dual xeon's for the renderfarm so most of the images are fine on a single node.