Mesh generation w. Depth maps | Surface Estimating GPU stop

[Casper_H]

Hi,

I've got an issue at the moment while generating a mesh with Depth maps, during 'Estimating Surface'. For the first part (around 25%) both CPU and GPU perform tasks like this:
[GPU 1] 16 Parts 21-36/6770...
[GPU 1]   Initializing octree...
[GPU 1]   Initializing histograms...
[CPU 2]   Initializing...
[CPU 2]   Iterating...
[GPU 1]   Initializing...
[GPU 1]   Iterating...

After the 25%, the GPU is no longer in the picture and the time to complete keeps jumping:
[CPU 2] 2 Parts 682-683/6770...
[CPU 2]   Initializing octree...
[CPU 2]   Initializing histograms...
[CPU 2]   Initializing...
[CPU 2]   Iterating...
[CPU 2]   Updating result...
[CPU 2] Part done in 400.562 sec

 If i calculate by hand it would be around 10 days.. to fully complete.

I use 1.5.3. 412 aligned 24 Mpix photos. 11 GB Video Ram, 64 RAM and  around 200GB virtual mem on SSD's.

Other comparable scans in terms of amount of complexity and amount of photos did not have this issue.

Any idea what could cause this?

[Casper_H]

Update:

I've cancelled the proces and started it again. I did reuse the Depth maps.

The process is now over the part where my GPU stopped 'helping', and the GPU is still covering tasks, unlike the first try.

My semi-uneducated speculated guess is that my vRAM got full - i was doing some photo editing at the same time - and Metashape stopped distributing tasks to my GPU and did not return doing it, even when my vRAM was cleared.

[Alexey Pasumansky]

Hello Cie,

There should not be an issue with vRAM, however, if you face the similar problem again, please check the video memory consumption using NVIDIA-SMI utility.

Also please specify the OS version number and GPU driver version used.

[Casper_H]

Hi Alexey,

When i use the NVIDIA-SMI utility, for what kind of memory behavior should i watch out for? Any anomalies that i can detect?

The GPU driver version: 430.86
Windows 10 version: 1809

Thanks for the quick reply.

Greetings, Casper