Hi,
We’re throwing some progressive test renders on the farm for a render job with V-Ray in Maya and we’re limiting the frame time using the ‘Max Render Time’ to keep it efficient but obviously our machines are very different spec so we get a wide variety of qualities. Is there a way that we could use the slave’s render time multiplier to change this per machine or have it as an option on the job properties? I know we could change this to use a noise threshold instead but then you run the risk of ramping up the render times on difficult shots and it’s quite useful for a rough render to be able to guarantee the speed.
Thanks
Nick
I think this is more of a Vray Question… I think they need a ‘passes’ parameter… as you can tie this to a rough time/quality normally…
Hi Dave,
They have the variable for time, we set it to a minute which works, the issue is some of our machines are twice as quick so the quality varies drastically from frame to frame. I was just wondering if we could use the slave time multiplier within Deadline to adjust this by machine type. It’s a bit of a hack and maybe we should just explore the noise option more but it’s easy for the render times to creep up with that kind of approach.
Nick
Yeah but you don’t know how much a multiplier will actually work… even VRay can’t predict the difference… hardware, drivers, software, hard disk space all come into the mix.
The passes parameter would be like a number of itterations to resolve to, rather than trying to hit a target quality as such…
We disable time-limited progressive renders because we have so many different specs of machine.
This is where the multiplier would come in. I know the relative time differences between the speeds and this is in place for each slave on Deadline so that we can calculate tasks timeouts accurately. If Deadline fed this information in for each slave as it picked up the job then it could adjust the time to spend per slave based on the expected performance. Basically normalising the render time across the different spec.
Again I think given that VRay have developed their profiling tools for benchmarking hardware it might be better for them to do it. I just think it’d take a long time to work out what the coefficient is for every machine you have,
Maintenance job~
Actually, have you guys seen the new Cinebench plugin we added in 9?
