Then you’ll be happy to hear Steve is currently working on block display speed.
Is the wireframe result a typo? Seems backward from the other results.
Argh! A copy-paste-edit user-error! I fixed it now.
Tested now on the GTX 1070 on dual Xeon, 50x50 heads, maximized viewport 2560x1440 monitor.
(7.500 meshes (head + 2 eyes) + 22.500 polysurfaces and 10.000 extrusions)
T OFF: 139.11 sec
T ON: 6.45 sec
That’s a whopping 21.5 X speed increase!
And the same goes for shaded:
T OFF: 142.34
T ON: 8.42 sec
Blocks should be faster in next week’s WIP. I’ll make an announcement of it, but just in case I forget…
Does it affects both blocks and nested blocks?
And do you want reports on speed comparisons on different systems?
It should affect both.
The improvements aren’t really vendor/driver specific so it shouldn’t matter a whole lot. GPU tessellation will make a difference and is obviously driver specific, but we have been drawing all polysurfaces and extrusions (even in blocks) when that feature was turned on.
Hmm… I think that you could also consider distance based tesselation it would be awesome feature for both viewing model and rendering purposes. Wouldn’t that be insane to have auto LOD system ? Rhino mesher should be able to do this without any problem. And of course simple check box to on and off when needed
It would be great for egde softened stuff smooth in front of cam and jaggy far behind and maximizes vram efficency for gpu rendering.
Also adaptive subd could be done in future when subd will be released.
This bug should be fixed in the next WIP. It is part of a larger bug that I haven’t completed so there may not be an announcement about this being fixed next week. I have only partially fixed the bug and handled the case that you found. Thanks
The “tessellation” referred to here isn’t for meshing surfaces, it’s for drawing curves and isocurves.
Sure, but i said that in case of srfs and solids sth like this could/should be concerned also i think - it would be really beneficial feature for all users
4 posts were split to a new topic: Improve shaded display performance
I’m still focused on the wireframe display since we are seeing poor display speeds on AMD cards and this feature still is not implemented for curves. I believe I have found the issue with the AMD drivers today and hope to have code in place so we can see improvements in performance for most GPUs very soon. There are also edge cases that need to be fixed where the wires look choppy and need to be further tessellated.
Just tested it on the laptop where I made a nested block of the head with headphones and made a 20x20 array of it. Then made a script that makes a 1600x900 px floating viewport in shaded and redraws 100 frames.
Your improvement gave it a 3X gain and went from 33 seconds to 10.9 seconds.
Will test it on the 1070 before I upgrade tomorrow.
@stevebaer, just tested with a current built and can see some speedup with polysurfaces. Tested in shaded mode (with wireframe, curves and 8x AA) on a lighter scene it is less noticable, eg. with 456 polysurfaces:
Time to regen viewport 100 times = 1.09 seconds. ( 91.58 FPS) (GPU Tesselation OFF) Time to regen viewport 100 times = 0.39 seconds. (256.41 FPS) (GPU Tesselation ON)
Larger scene, having 194 points, 10797 curves, 394 surfaces, 786 heavy polysurfaces:
Time to regen viewport 100 times = 9.88 seconds. (10.13 FPS) (GPU Tesselation OFF) Time to regen viewport 100 times = 3.21 seconds. (31.11 FPS) (GPU Tesselation ON)
Thanks a lot
You must have special privileges to builds😀. I was holding off til the next WIP to announce improvements in this area on AMD.
Yes, sorry. Btw. when i hide curves the speed goes up more noticable in the larger scene:
Time to regen viewport 100 times = 7.24 seconds. (13.82 FPS) (GPU Tesselation OFF) Time to regen viewport 100 times = 1.20 seconds. (83.26 FPS) (GPU Tesselation ON)
Super awesome !
I’m working on curves this week. You will most likely not see as drastic of an improvement at first with curves as were seen with polysurfaces and extrusions, but there should be at least another small bump in improvement on your model soon. I don’t know if I’ll have this in place by tomorrow’s WIP, but that doesn’t seem to be a problem for you
Thanks @stevebaer, i am looking forward to it. Really cool that Rhino WIP now runs much faster while still using this old hardware.
The GPU Tessellation feature should work well with AMD GPUs in the latest WIP. Previously the feature was only working well with Nvidia and Intel GPUs.