I just downloaded the WIP. Looks really nice so far, but there is a big performance hit with point clouds with Direct3D. OpenGL still works good.
EDIT: I have an rtx 4070 with latest nVidia drivers.
Hi Dan -
I put 20-some points in a new scene and turned those into a point cloud.
I’m switching between Direct3D and OpenGL but am not noticing any difference.
Is what you are describing file-specific or are you also seeing the performance hit in the scenario that I tested?
Please run the Rhino SystemInfo command and copy-paste its output here.
-wim
I would not expect performance issues with so few points. Make it 20-million points instead?
It looks like the OpenGL mode uses decimation depending on zoom level and point cloud size. Direct3D does not do that (yet). I have an RTX 4070 as well, and the difference is quite noticable (use _TurnTable to see it) for point clouds that are 10M+ points.
Hello Wim, thank you for the speedy reply.
My post was too vague, so I want to explain the process I use almost on a daily basis.
1. I receive scanned data of entire buildings captured with terrestrial laser scanners (TrimbleX9, Leica RTC360, etc.)
2. I process these clouds with CloudCompare, subsample, colorize, trim unwanted parts etc. I usually keep a heavily subsampled cloud with the entire building and more detailed clouds based on functions and project needs (building levels, main staircase, zones with detailed piping and electrical, architectural details etc.)
3. I import the clouds saved in E57 format to Rhino via Attach method.
4. I use clipping planes and custom CPlanes and rebuild the entire geometry in Rhino.
I know this may be a niche for now, but Rhino 8 is very capable at doing this already. I tried other CAD and BIM software and, even with dedicated tools for point clouds, I find them lacking in flexibility because “as built” data is never ideal, walls are not parallel or perpendicular, piping is never straight, and repeating elements are rarely actually the same. The modeling capabilities of Rhino make it the best approach for accuracy.
In my experience, point cloud data has gained a lot of popularity in the past few years, not only for existing buildings, but also for new ones where you want to document the different building stages to have the information for later.
I recommend looking at software like OpenScanTools (https://openscantools.com/) to see an example of efficiency in viewing and navigating point clouds (I could load up clouds of around 2Billion points spread over around 250 scans and it worked flawlessly).
I’ve attached the file I tried in WIP. For me, this is on the small side. Files are usually around 5-6GB (200M-300M points with color information) and Rhino 8 works well with them. I’ve also attached a PDF with a few examples from different projects, to better understand the complexity and scale of the work. If you wish I can get my hands on bigger files to provide for testing. I really hope that switching to Direct3D will not impact the workflows around scanned data.
I will attach the system information as soon as I get to the office on Monday.
Point cloud example: Transfer - Dropbox
Project examples: Example_work.pdf (16.4 MB)
Thanks, Dan!
→ RH-90644 Display: Direct3D: Point Cloud Performance Issue
-wim