Hi.
I am currently using ContextCapture Desktop edition Update 19 v10.19.0.122 with the following hardware:- Intel Xeon Gold 6240R 2.4GHz, 24 cores- NVIDIA Quadro RTX 6000, 24GB GPU- 128GB (8*16GB) DDR4 RAM- 2TB PCIe Class 40 Solid State Drive- Windows 10 Enterprise 20H2
I am currently surprised that only 8% (10GB) of the RAM is used in the production of the point cloud or mesh, although CC reports that the required RAM exceeds 16 GB (without tiling). CPU utilization is around 100% (see screenshot). Vulkan API and multiGPU processing is enabled (benchmark ~300).
Are there any setting options to speed up the calculation?
I'm sorry to raise this question again:Are there still settings in the ContextCapture Settings or in the Nvidia Control Panel that can be used to make full use of the computer's performance and thus increase the speed of production in ContextCapture?In the event that programme-related settings can be made in the Nvidia Control Panel, should these be set for ContextCapture Master or Engine?
Yes, I disabled Vulkan API and multiGPU processing..
Hey,
sorry for the late feedback. I've been busy and haven't had much time to use ContextCapture.
Now I have evaluated a new drone flight, this time with two different engines. I have already described the configuration of the first engine (workstation) above. The benchmark of the graphics card NVIDIA Quadro RTX 6000, 24GB GPU with deactivated MultiGPU is 299 :-)
Now I have calculated the creation of a 3D point cloud with a Dell notebook with the following parameters:
- Intel Core i7-10875H CPU @ 2.30GHz- 64 GB RAM- NVIDIA Quadro RTX 3000 (6 GB GPU) - Benchmark 140.8- Intel(R) UHD Graphics (1 GB)
The calculation of the point cloud with dimensions of 250 m x 210 m with a resolution of 2.5 cm from a total of 973 photos took a total of 7:05 h:min with the notebook (89 tiles, 10 GB RAM usage).
With the workstation described in the posting above, the production of a point cloud with dimensions of 210 m x 175 m with a resolution of also 2.5 cm from a total of 930 photos took 5:44 h:min (6 tiles, 38 GB RAM usage). Although the project area was smaller...?!
Somehow I had hoped to need less computing time with the "big" machine. Are there any settings, for example in the NVIDIA control panel, that can be selected to make the computing process go faster? Somehow I can't imagine that the much better processor, 2x more RAM and 4x more GPU should do pretty much ... nothing!?
Yes multiGPU is only for multiGPU otherwise it has performance penalty. Also installing additional card won't improve speed 2x times but only 25%.
Disable VulkanAPI, do benchmark and post scores so I can update my CC benchmark results .
So that means that disabling MultiGPU might speed up the calculation? Or do you mean that it doesn't make sense to enable MultiGPU if there is only one graphics card?