Slow Network Rendering

[Moulf69]

Hi all,

After setting up Keyshot Network Rendering (1 Master, 6 Slaves, 32 cores in total), I have an annoying problem :

I made some tests generating a VR with about 1000 images, and the results were quite good, in terms of rendering time.

But after rebooting the Master (and one or two slaves maybe), I made some new tests and the estimated rendering time is 4 times longer. And I noticed that now, on all my Slaves, the cores barely work at more than 10-15% of their capacity (which was not the case before the reboot).

Any ideas ?

Moulf.

[guest84672]

Brandon will take a look at this. It could be based on the image that you are rendering.

[Moulf69]

All the tests were made with a Mobius Ring I created for the occasion, with a glass material applied on it.

When creating a simple render (one single image I mean), the cores are working well, reaching 100% of their capacity.

[guest84672]

Same size image for single rendering and VR?

[Moulf69]

No. The single image was 6000px wide. The first VR with 1280px wide images, the second (the slow one) with 1024px wide images.

[guest84672]

Ah- that could be the issue. Image renders too fast so the slaves don't even kick in.

[Moulf69]

Haha so to gain time I have to increase my rendering settings ? Great news 

I'll make some more tests later !

[Brandon Davis]

When you mention "the second (the slow one) with 1024px wide images", do you mean the overall render time was slow or the CPU usage is low? In your case when rendering the VR, the images are being rendered fast, which is why you are seeing the 10%-20% CPU usage. 

[Moulf69]

I meant the overall render time was slow. I just presumed the CPU usage to be responsible for that..

With the 1280px images, for 1000 images, the overall render was made in about 45min.
With the 1024px image, with slightly lower quality settings, still 1000 images, the overall render time was predicted to be about 3 hours. I stopped it after one hour. In comparison, my old Dell Precision M6400 with two cores was a little bit faster than the 32 cores in network...