I’ve noticed for some time my slaves are super slow to kick-up, which is odd given they’re all dual-xeon monsters. It can’t be the scene given it starts in seconds on master workstation.
Monitoring task manager showed me the whole precomputation is happening in single-thread only, i.e transforming vertices, parsing the scene and seemingly also large part of secondary GI ?
Is this intended behavior or some bug along the path.
Also, the slaves keep rendering the initial pass over and over. The passes do stack up on master-node, but is the resulting quality identical to if the slaves keep sending more refined result ?
Or is this affected by the time period over which updates are sent ?
As to 1 - I am not really sure, but I don’t see any reason why it should be single-threaded. Ondra should have the final word. You are not using any string options, right?
As to 2 - You got it right. It just resets after the data is sent. So it is by design.
https://coronarenderer.freshdesk.com/support/solutions/articles/12000015664
I don’t even use that feature. Every node ( 1.4.1 corona, both public and daily), exhibit this. Will try to make screenshot.
Can any big DR user chime in on his experience ?
I’m not rendering very heavy scenes, so for me it’s not that a big difference.
Transforming vertices seems to be single-threaded- yes. But not secondary GI- what do you mean ‘large parts’ of it?
Maybe my secGi-calculation is too short and I don’t see it switching from single to multi-threading.
I’m using 1.3
Transforming takes the longest, but parsing is the second after. These two together, take almost 5-10 minutes, opposed to 20 seconds on master pc. My scenes are around 1GB big. (although they can have a slew of proxies and xrefs attached so that true, virtual size could be almost 2GB).
Is the any chance we’ll see new DR in 1.5 release ? I’ve seen around here that at least 3dsMax will be pre-launched.
it wont make it into the release, sorry. We will try to release it ASAP in 1.6 daily builds though
Is this issue resolved in the new DR? I’m not seeing a big difference in parsing times on slaves.
I will check tonight. But I think it’s quite some time ( recent dailies ) that this is indeed fast now.
hi, is this confirmed as resolved?
no longer happening