[deboy]

Hi,

How does Co-op rendering compare to each node rendering its own project?

Whats faster;
- two machines working on two jobs sequentially as co-op
- two machines each working on their own project

It always "feels" like its inefficient when both nodes render to 18.03 & 16.30 just to get to a sum of 19.00

Any comments welcome

[Bubbaloo]

Take note of the time it takes to render a scene to s.l. 18.
Then measure the time it takes to render from s.l. 18 to s.l. 19.
You will see that with each new s.l. it takes an increasingly long time to reach the next s.l.
So when you have 2 nodes and one reaches 18 and one reaches 16, and it combines to "only" 19, you feel like it is not right. But it is!
So the thing to remember is that the more processing power you can give to Maxwell, the faster it will reach higher s.l.
The only "time" you will lose is when merging the mxi's at the end of the process which usually amounts to a small fraction of the total render time.

Personally I try not to pay to much attention to the s.l. number.
There are scenes that clear up at s.l. 10 and some that do not clear up until s.l. 19 or higher (SSS).
Also, render resolution plays a big part.
I have a scene right now that clears up at around s.l. 18 at 1280x720 resolution, but gets clear at s.l. 10 at 15,000 px.

[deboy]

I agree that the SL takes an exponential amount of time.

The render SL is important when you have a big job queue and need to set finishing criteria though.

My still remains is there a more efficient option?? Certainly as the SL increases the computer resources are used less aggressively in my experience also.

Cheers

[Mihai]

It's still using 100% of the CPU no matter what SL the image is at, just that it won't write an mxi file as often as in the beginning, it will write it every 10 minutes while at the beginning it will write to disk with every new SL, if it takes less than 10 minutes to reach it.

As Brian said, it's a small difference between doing co-op or separate nodes for each project, it's a bit more efficient rendering on separate nodes.

[Bubbaloo]

Nevermind, Mihai got you covered.