1.6 DR

[tolgahan]

I think the pass numbers are wrong.The result and the pass number do not correct.

[tolgahan]

Doesn't start the other backburner job ? And 3dsmax.exe  is still running.This version is more problematic than the April 3 version.

None of these problems existed in the April 3 daily build version.

[tolgahan]

1400 pass :)

With this pass, this result can be verified ? Can I be doing something wrong ? I use hdri on my scene

[ihabkal]

This is just one pass made by your main pc, all the other passes are black blank passes made by your render nodes. I have been talking about this issue for 2 months. I think Ondra should hire a high caliber artist and task him with testing Corona on very large scenes on a network of 10 to 20 render nodes.

[Freakaz]

Can confirm that number of passes done by the slave shows random numbers. Currently rendering a scene on 2x2670 as a master + 4770k as a slave, and total amount of passes is 75 (DR:61) highly doubt that my slave is 4 times faster than the master.

[tolgahan]

Can confirm that number of passes done by the slave shows random numbers. Currently rendering a scene on 2x2670 as a master + 4770k as a slave, and total amount of passes is 75 (DR:61) highly doubt that my slave is 4 times faster than the master.

I've disabled the pass section.I get all the visuals time adjusted.

[ihabkal]

Can confirm that number of passes done by the slave shows random numbers. Currently rendering a scene on 2x2670 as a master + 4770k as a slave, and total amount of passes is 75 (DR:61) highly doubt that my slave is 4 times faster than the master.

I've disabled the pass section.I get all the visuals time adjusted.

can you please elaborate what you did and if it fixed the issue? this bug plagued me for weeks! if you solved it then we need to get you a prize :)

[Ryuu]

Anyone still having the DR empty/black passes problem - could you please describe your setup/problem in as much detail as possible? Unfortunately we weren't able to reproduce this issue yet, so this is currently a severe case of shotgun debugging.

• Do you use just the Corona DR or Backburner or something else?
• Are there any errors/warnings displayed on master/slaves during the rendering?
• Please double check that you have a Corona license activated on the slaves.
• Please check that both master and all the slaves are running the same version of Corona
• Please send us all logs from master and the slaves.
• If possible, please save the EXRs generated on slaves and send them to us.
• Do you use any firewall?
• Are the slaves on the same local network as the master?

Also please let us know if this was happening to you in the past and has been fixed.

[ts_berlin]

Hi,

no Region-Rendering with DR, is it true?

EDIT:
OOps...now it works, I have to restart the slaves!

beste, Thomas

[sevecek]

This is just one pass made by your main pc, all the other passes are black blank passes made by your render nodes. I have been talking about this issue for 2 months. I think Ondra should hire a high caliber artist and task him with testing Corona on very large scenes on a network of 10 to 20 render nodes.

What kind of lighting do you have in your scene? Are you using some native 3ds Max lights?

[Frood]

From changelog:

"DR reverted back to rendering whole passes on slaves"

Somehow... I like that very much! Will that option (full frame by default + option for stripes) be present in the release?

Finally we can have the situation that 1DR pass = 1 local pass, period. No banding, even low pass DR rendering is consistent this way no matter which nodes (slower/faster) contribute.

Not sure about the sampling focus mask though. It always looks much noisier and less blurred compared to single node rendering? To my understanding the mask is generated by the master and periodically distributed to the slaves so there should be no such big difference?


Good Luck

[ihabkal]

From changelog:

"DR reverted back to rendering whole passes on slaves"

Somehow... I like that very much! Will that option (full frame by default + option for stripes) be present in the release?

Finally we can have the situation that 1DR pass = 1 local pass, period. No banding, even low pass DR rendering is consistent this way no matter which nodes (slower/faster) contribute.

I agree,
if I knew this was the case I would've guessed it was the reason for black passes quickly, as I tried backburner strips a long time ago and they came out black,
There is one major problem with strips I think! when I render a 6000 pixels image it takes 2 hrs, when I render a quarter of it in a region it takes a third of the time not quarter. when I render a selected item in an animation the frame takes 5 minutes for the small very small element, while the whole scene took 10 minutes originally. I think Corona is optimised to calculate the whole scene it calculates everything, but when you slice and dice you pay a speed penalty because it has to calculate some hidden items to do it's lighting and reflections etc... so my opinion is strips is overall going to be slower! please correect me on this if my guesses are wrong.

[Juraj]

From changelog:

"DR reverted back to rendering whole passes on slaves"

Somehow... I like that very much! Will that option (full frame by default + option for stripes) be present in the release?

Finally we can have the situation that 1DR pass = 1 local pass, period. No banding, even low pass DR rendering is consistent this way no matter which nodes (slower/faster) contribute.

I agree,
if I knew this was the case I would've guessed it was the reason for black passes quickly, as I tried backburner strips a long time ago and they came out black,
There is one major problem with strips I think! when I render a 6000 pixels image it takes 2 hrs, when I render a quarter of it in a region it takes a third of the time not quarter. when I render a selected item in an animation the frame takes 5 minutes for the small very small element, while the whole scene took 10 minutes originally. I think Corona is optimised to calculate the whole scene it calculates everything, but when you slice and dice you pay a speed penalty because it has to calculate some hidden items to do it's lighting and reflections etc... so my opinion is strips is overall going to be slower! please correect me on this if my guesses are wrong.

I thought it was always rendering true full passes just sending them out in strips to save bandwidth, so slicing them up somewhere in process ?

I always cranked up the pixel amount to get full passes, I rather value immediate visual feedback.

My nodes always report passes correctly to Master:Node performance ratio. But I always wanted to check if 100 Master slaves passes equaled 100 (Master+Nodes) passes in visual clarity. I just don't have patience or time to test this stuff :- )

[Frood]

it calculates everything

Hmm, with some view dependend exceptions like displacement (or forest objects with visibility check) for example maybe. But today I learned that in fact it currently it works like this:

always rendering true full passes just sending them out in strips to save bandwidth

Problem now is the sampling focus mask generated when using DR: It seems strange to me, it's very noisy and thus can cause additional noise apparently. In the images you can see the mask (1:1) and a comparison of the result (4:1) with and without adaptivity. It should be exactly the other way round :)

But I always wanted to check if 100 Master slaves passes equaled 100 (Master+Nodes) passes in visual clarity.

This is what me and others here would love to have since the time adaptivity was implemented and switched off later on slaves. Maybe slaves should just calculate their own mask, they render fully anyway now :)


Good Luck

[Juraj]

ha, that's a big difference.

In theory, shouldn't the current solution eventually provide superior result further in rendering ? Initial render mask will provide noisier result, but eventually master will send up more precise mask than render slaves would have generated by themselves.
Or maybe that's wrong...