Hello,
I’m wondering if RC can do network processing.
If not, can I use Xeon Dual cpu for RealiyCapture? And does it fast?
The reason why I’m asking is that I am currently using i9-9980xe and dual 2080ti, but it’s too slow.
So, I’m gonna try network processing, if possible, or upgrade my gears.
Let me know. Thank you in advance!
Distributed processing is currently not an option for RC. With the Xeon, I am not sure whether this is a good pick. Depends on what the specs are. The process does not get too quicker when count of cores is increased. Rather the pure power of the CPU matters.
Thanks for the quick response.
Then, which CPU do you recommend to use for RC? Should I look at the cpu clock?
Yeah I would like to prevent mentioning specific CPUs, but the core clock is the biggest player here.
Its possible to do this yourself, you can script CLI version and use a render manger like deadline or even backburner.
I was hopeful that they would allow you to do it with PPI, but that’s been locked to one pc.
if you have enough ram you could still run 2 versions at once using PPI.
though you need to be careful with the model tile edges, they need overlap or you will get holes. So it’s quite a bit more work to clean up. it would best to do a small test first.
you can also share gpu cache between nodes, which saves a lot of processing time, when it doesn’t get corrupted.
I requested this about 4 years ago, hopefully they can make this a bit easier one day.
Did you get any further with this? I presume the problem of multi-threaded, multi-gpu or cloud tasking is whether parts/tasks of CapturingReality can be partitioned and then brought back together again.
It would be good to know if there are any tasks that can be split up?
Do you also know if I can distribute the GPU tasks on my network to other GPUs? This is proving to be a bottleneck. I see you say I can use 3 GPUs, but do they have to be on the same machine?
I am working Gigabyte Aero laptop with 32gb Ram and 16bg nvidia GTX1070. I cannot add a gpu internally, but I have other machines and GPUs on my local network.
Hi Graham3D3D, this question wil be answere throught our support system.
Hi,
i’m interested in distributed computing, also. To make reconstruction iterations shorter in time. Usually, i have only 2 days for reconstruction. And the goal is to always get maximum details and minimum artefacts, so i have to make multiple iterations to get optimal result. Is it possible now?
Thanks in advance!
Currently, RealityCapture doesn´t support clustering or distributed computing, but there are ways how to use a component workflow and CLI scripts to create your own customized automated solutions. You will need a license of RealityCapture for each of these machines.
Regarding distributed computation, it is considered to be a part of our roadmap, but we don´t have any ETA on that.