Why didn't my render end?

rich_4318641

Hi all

This may well be because my expectations are wrong, but when I ran a render earlier today it seemed to go into an endless loop of iterations.

As I understand it there are 4 things that will end a render. Time, max iterations, convergence, and quality. (I know the quality setting is a little nebulous for a newbie). And my understanding is that the first criteria met stops the render.

I wanted a reasonable render so I left convergence at 95%, time was 7200, quality was 1, and I ramped up the iterations to over 100K. My expectation based on previous aborted renders was what it would hit 95% convergence easily within an hour.

After 12 minutes the convergence was at over 91%, so I expected it to hit 95% after maybe another 20 miutes to half an hour and stop. But it didn't, the logs stopped reporting the convergence value at all and it run until I stopped it after 76 minutes.

So after 779 seconds and 2147 iterations it was at 91.6% convergence, but then stopped reporting convergence, and run until 4640 seconds and 12617 iterations. The logs never gave the final convergence figure.

Are my expections wrong that it couldnt get another 3.4% convergence in over 12,000 iterations?

Thanks in adavnce!

Rich.

Richard Haseltine

Quality isn't a stop condition, it determiens how fussy Iray is about counting a pixel as converged. Other than that, you are correct. It is certainly possible for soem parts of the image to converge more quickly than others (if not it would simply jump from 0% to 100%), but the lack of any progress reporting does suggest that something had gone wrong with the render.

rich_4318641

Thank you Richard. And that's a very useful definition of quality!

Ok, I'll chalk that one up as a error then.

I guess it's possible for a particular element of a render to be more troublesome, so the rest of the image is converged but one part needs far more work. I did have some shiny surfaces in the render so perhaps they were the issue?

Richard Haseltine

Glossy surfaces can certainly slow things down, both because the paths go on for longer and because the angles of the paths affectwhere they go after hitting the shiny surface and so they take samples from a wider area and need more iterations to settle on a final value.