just want to share my latest experience as I have some dense LIDAR scan data here that I was testing on the latest WIP.
the rhino file is 10 GB (the original e57 was 9GB), total about 500.000.000 points. 500 million, yes
Rhino 5 was opening it and i could navigate a separate 28.000.000pt cloud ok when all others were hidden, but it was tough to work with.
Rhino 6 is in a totally different class for this, I do run a TITAN X (maxwell) and the file uses about 65% of the 12 GB Ram, when I rotate the model GPU core is on 100%, and even with all point clouds turned on, it rotates almost fluently.
I noticed that there is some automatic degradation happening which culls points as soon as the model is rotated, which seems totally fine visually.
It does however cause the graphic pipeline to stall for a few seconds after stopping the rotation, as the culled points are displayed again. During that time it is not possible to start rotating again.
That speed improvement is fantastic, however I am wondering if there could be more control over the culling of points when displaying pointclouds. eg. culling them “permanently” on object properties level so that there is no need for redrawing the culled points.
btw. when I switched to “round points” performance was really bad. only square points draw fast.