Chaos Corona Forum
Chaos Corona for 3ds Max => [Max] General Discussion => Topic started by: jetcrow on 2014-04-21, 13:18:03
-
Hi all. I realy want to complete animation project via corona render. As far as i get experience in architectural visualization corona bring to me fantastic control on scene lighting. I tied from vray ))
I ask experienced corona users help me get close to good animation result. I attach some wip images.
So to the questions:
1. first try to make simple animation flyby byte my brain strong. Solution was to try distributive corona render. PT+HD (HD precalculated and loaded) I start render 12 hour ago, but if you can see (time.jpg) elapsed time is 3:16 how ?
2. Render is not canceling
3. What is the best way to save render result? (missframes.png) some frames missed.
4. Beauty channel is black. Was saved via max frame buffer to hdr.
Realy need help!
-
If you want to render animations, you really should use some kind of render manager like Backburner (which is included with 3ds Max).
-
its mean to render each frame at 1 node?
-
yes, that is the simplest and fastest solution, when available.
-
its mean to render each frame at 1 node?
Yes. This is much more efficient than trying to render everything with distributed still frame rendering. It's also more reliable. If you use a render farm manager and the renderer crashes during a frame, it will only crash on that one machine, and the controller will automatically restart it and try again.
-
1 frame per node via backburner is good strategy, but I don't see why it is necessary seen as 'superior'. I hope to see one day DR+BB work together as well, that is workflow I enjoy and prefer a lot.
-
Thank to all for the answers. Its clear now to render via backburner.
What can you say about image save procedure? To get result 1to1 as corona framebuffer the only way i found is to save to .hdr But animation sequence .hdrs are black. Render elements are normal in .hdr
.exr not accept exposure and color ballance manipulations after image save.
-
1 frame per node via backburner is good strategy, but I don't see why it is necessary seen as 'superior'. I hope to see one day DR+BB work together as well, that is workflow I enjoy and prefer a lot.
It's seen as superior because it is superior. :)
When rendering, there are some tasks that aren't even multi-threaded, let alone allow for spreading the load between many machines. For example, constructing the renderer's internal representation of the scene itself is typically a single-threaded task. Using distributed single frame rendering just means that all the machines load the same scene once, each. I guess in Corona that the HD Cache is calculated on a single machine as well, and isn't distributed among all the machines. So the rest of the machines will sit idle while the master machine generates the cache.
It's also much more reliable, because the farm controlling software is typically pretty stable. If the renderer crashes or fails for any reason, it will simply try again. More often than not, that is sufficient. Random crashes happen occasionally with 3D software. If you're using distributed still rendering and a slave goes down, it won't restart automatically. If the master goes down, the whole thing will fail and you will need to manually restart the process.
Last but not least, depending on which software you're using, distributed still frame rendering might tie up a single license because you need to run the GUI version. This might prevent you from working in the software at the same time. When using a render farm controller, it will typically just run on render slave licenses and not block you from using it.
I'm honestly having a hard time thinking of advantages with distributed still frame rendering for animations.
-
I've never had single crash of either hardware (slave or master), neither DR spawners or BB servers.
With that said about benefits, it lies with stills foremost. I can schedule 15-20 pictures, and they will complete hierarchically one after one at short periods of time fully automatically. Where one node would take 20 hours, now I can see each one after 2 hours sent to me to my phone.
The non-distributed parts are pretty minimal by now with most renderers (with Vray it is only LC now, and that is circumvent by setting BB server to be the strongest machine, i.e dual xeon for example, so it takes like 30 seconds, while frame will still take 5 minutes).
I wasn't advocating the use of DR on behalf of BB, but "DR+BB", so I don't see the confusement.
The way it works the work is still fully managed by farm controller (BackBurner Manager in this case), so your machine is free to use or quit, but the work is sent to single server only, which afterwards gather the performance for distributed to itself from other slaves, which are running spawners only, but not servers.
This feature was highly wanted back in 1.5Vray and not because everyone was idiot who couldn't do their math. It was so people didn't need to use idiotic "strips" feature.
So it's not superior or worse, but regular alternate workflow which its own benefits.
-
Unfortunately i switch corona to develop mode its start to save image to .tif normally. I dont know maby it was a damnation of last week?
So for present moment my questions on top close. Tnk you all.
Now about light with animation, for axample my camera moves from closeup on table to overall room view. Would it be correct to calculate HD from last frame and load it to whole animation? how much i loose in closeup view. Im doing tests but if some one have opinion on this?
-
Not a valid vimeo URL
test with HD cache calculated for each frame. 1024 \ 300 a bit Flickering