@Charles, what chair? Do you mean you used a chair from the file posted by @dmoyes?
When you do that many copies from the same object you should use block instances. You probably just have to be very patient when using hard copies of those chairs. That is about 16M triangles that need to be uploaded to the GPU. I haven’t done the math really, but it’ll be quite a huge scene to upload (several GB if not more). And if you happen to have already lots of the GPU RAM in use you may even see that part of the data gets uploaded to the host RAM (your desktop/laptop RAM). That will be another factor of slowdown once the data is uploaded to Cycles.
Doing block instances saves you a huge amount of memory, and once the data is at the GPU rendering is OK, and you can even manipulate the view with some degree of ease. And that is even without joining up meshes where possible.
I have so much to complain about rhino6, but viewport performance is something I’m actually happy with. I use a desktop PC (Core i5, 32GB-RAM, GTX1060, SSD only, Monitor 2560x1440). I guess it also heavily depends on the size and complexity of your model. I can pan, zoom, rotate or do whatever in shaded mode with the model you uploaded - it’s always smooth and no degradation occurs on my system. Btw, when comparing with testmaxspeed, everyone should use a floating viewport with the same resolution.
For other reasons I’m so fed up with rhino, that I’m currently trying out Autodesk Fusion 360. I realized, that rhino isn’t the best fit for the stuff that I do at the moment anyway. Onshape is similar to fusion and I might try it as well later.
I know, it’s hard to give up on a program that you have used for a long time, but if you take your time and explore what other solutions have come up with over the last years, you will be surprised.
This is such a great idea, waiting in the dark is absolutely the most annoying and stress full thing with computers, and so easy to avoid with a few lines of code
Oh, and it makes absolutely NO sense nor is logical to a “normal” user that Raytraced mode needs to upload geometry to the GPU when that exact same GPU is allready showing the objects in all other display modes. So communicating that something timeconsuming is going on makes a lot of sense.
@nathanletwory
Something I’d like to add regarding viewport performance. I did a comparison with testmaxspeed between my office PC and the one at home, same scene of course, floating viewport of same size, both AA8x and tes on.
Strange is that the RX460 in the home PC (which is overall a much weaker system) gives me higher fps than GTX1060 in my office PC. Both “feel” fast, but I’m just a bit surprised here. When switching to raytraced, the GTX1060 blows the RX460 away though, raytraced on the RX460 is barely usable.
These are the results with _TestMaxSpeed:
RX460 (Tesselation ON, 8x AA)
Floating Viewport @ 1200x800
Shaded:
Time to regen viewport 100 times = 2.33 seconds. (42.96 FPS)
Arctic:
Time to regen viewport 100 times = 15.06 seconds. (6.64 FPS)
GTX1060 (Tesselation ON, 8x AA)
Floating Viewport @ 1200x800
Shaded:
Time to regen viewport 100 times = 17.56 seconds. (5.69 FPS)
Arctic:
Time to regen viewport 100 times = 34.14 seconds. (2.93 FPS)
I always make sure to have the scene set up the same way. I also do this before running _Testmaxspeed:
_SetView _World _Perspective
_Zoom _Extents
OK, the system with the GTX was set up a long time ago. Hope to finde the time to make a fresh install along some hardware upgrades and then I’ll test the GTX again.
There should be messages about data being handled, but I now notice that not all ends have been properly bound together. I’ll make some fixes that should be then available in the second next service release (6.9 I believe).
I would describe this as an encoding error occurring in de-de no commands.xml, where open[[37690]] has been localized as offenen[37690]], with a single bracket [ on the left side of the word. Looks like it has been there for years but only recently discovered. Thank you @Charles for the hint, I will get this fixed with the next strings project coming up.
Interesting, I did tweak a lot in the display modes indeed. But here Shaded mode was on its defaults, hit “Restore Defaults” anyway, just to be sure. Arctic was not, just restored the defaults there. Will have to check the home PC later today.
But anyway, I just updated to driver 398.82, still the near exact same results with the GTX1060.
Nvidia wants to install lots of crap. I only let it install what’s necessary:
Graphic Driver [YES]
3D Vision Controller Driver [NO]
3D Vision Driver [NO]
HD Audio Driver [YES]
Nvidia GeForce Experience [NO]
PhysX [NO]
Hope I don’t need that other bloatware crap, right?
There is also an option to make a new install - guess that could help eliminate potential problems, I always tick that box.
v-sync:
was decided by application, is turned off now. makes a 0.3 difference in fps - within margin of error I’d say.
suggestion:
a holomark for rhino6 would be great to set up a database.would help to compare hardware. can’t @mcneel just pay @Holo so he can make it? Holomark2 is for rhino5 only, right? is also free, right? mcneel and the community in this forum benefits from it and rewarding him for an rh6 version would be justified imo.
v-sync only kicks inn when you are exceeding the frame rate that your monitor supports
(usually 60 hz and thus the GPU will calculate and hand out max 60 fps, no need to consume energy to calculate 500fps if only 90% of those are shown on screen)
Oh… yeah Holomark for V6… I’ll get there one day, just finishing off the last days of summervaccation for the kids these days… I can’t prioritize it over paid work right now.
