Hi, I’m am using an Amazon cloud instance with 60Gb memory and 4GPU’s.
I am constructing a model with 1910 images of 4112 * 3088 pixels in high detail.
During processing I can see RC utilising the 4 GPU’s.
After about 8 hours of processing it starts with the “creating the model” step.
In this step I never see any utilisation of the GPU’s anymore.
In addition, at some point the memory usage shoots to 60Gb (I have 60Gb) and then the machine starts swapping.
Main CPU utilisation drops to 4-6% for lengths of time.
See attach for several diagnostics screens.
Naamloos4.png
My two questions:
a) Is it normal that during “creating model” step no GPU’s are used?
b) What can I do to prevent RC from using that much memory? 60Gb seems a whole lot to me! :shock:
Greetings,
Steven Verver
Hi steven verver
I am constructing a model with 1910 images of 4112 * 3088 pixels in high detail.
During processing I can see RC utilising the 4 GPU’s.
After about 8 hours of processing it starts with the “creating the model” step.
yes its is normal as it first calculate Depth maps on GPUs and then the next step is to calculate the actual model (3D mesh ) that take its time…
Hi Steven,
Thanks for posting. Glad to hear you could run it on Amazon 
Regarding your questions:
a) Is it normal that during “creating model” step no GPU’s are used?
Yes, creating model phase uses just CPU cores, but as many cores/cpus as you have
b) What can I do to prevent RC from using that much memory? 60Gb seems a whole lot to me! :shock:
This is not normal, we observed this already and bug fix is in progress. Try rotating the ground plane or changing a reconstruction box. Depth maps calculation is cached so they will not be calculated again.
Ok thanks for your fast response I’ll try and let you know the results.