Rhino 8 performances are really bad

Hi,
I was modifying a project cloning some 2d symbols and I noticed that Rhinoceros 8 was getting slower and slower…

Unfortunately Rhino 8 drawing performances are really bad compared to the ones available in the old Rhino 7.
It doesn’t matter what specs has your graphic card/system. My old AMD + Rhino 7 (96GB Xeon 12 cores) beats the performances of an RTX 4090 + Rhino 8 (128GB Threadripper 16 cores)
Please developers, watch this simple videos. I don’t understand why you don’t measure the graphic performances with the old Rhino code.
Autocad LT is 1000 times faster to handle this type of graphics.



3 Likes

Hi -
As answered in the thread you linked from, curve display optimization in on the list of things to look into. If you post a 3dm file, we can add that to the YT report.
-wim

1 Like

Hi,
this is the example file.

trees.rar (15.0 MB)

1 Like

The difference even just trying to pan around is ridiculous.

2 Likes

But I’m not seeing that sort of difference here at all. Panning, deleting, it’s basically the same. People are making some very simplistic assumptions and conclusions about this stuff like “hur durr you idiots why can’t you see that?” Well they can’t.

I’ve certainly seen models with large object counts become weirdly slow, but that’s like a different thing, 8 seems to be better at that.

@ftzuk I certainly cannot see that here with panning.

This is a potentially daft suggestion, but can you use Intel PresentMon to monitor where the work is physically happening?

If there are few inroads to this, perhaps you can try the gaming approach and see if GPU Busy within PresentMon is high or low, and where exactly the load is happening. Perhaps that may be a useful tool to see what is happening at the viewport processing?

You can even overlay the graphs so at least someone here can better understand what the hardware is doing at the stuttering.

1 Like

I can understand the AMD Radeon being slower, but not seeing anything remotely compared to these times on a lesser RTX setup.

1 Like

Thanks, Gabriele.
I’ve created a new tracker for this so that it is visible to the public → RH-82891 Display: Lagging with Many Curves
-wim

1 Like

@gabrielefx it looks like the slowdown to focus on for your case has to do with when new geometry is created and when geometry is selected. Once everything is in place the display seems fast. On my older Intel GPU system I do see lags for this (not nearly as slow as what you are reporting though).

If I just load the model you posted and run TestMaxSpeed I get results around 50FPS. Do you also see similar results with TestMaxSpeed?

1 Like

I’m wondering why the selection of elements is taking so long.
This is with an ‘older’ Rhino version:

Zoom and pan is slow of course - no wonder, VM, XP, no graphics acceleration.

It is, even on low end Intel HD graphics.

I get 85.32 fps within the Radeon Pro 580X.
The issue is related to selection / unselection then drag.

FYI:

I opened it, zoomed to extent (_-Zoom _Extents) and Testmaxspeed gave me 60.39 FPS.

Selecting all (_SelAll) takes roughly 2-3 seconds, unselecting (Esc) about 3-6 seconds. Strange, why takes unselecting longer?

Deleting all but one of the tree groups took 10 or 11 seconds.

Then I made the array with this _Array 10 10 1 30 -30 _Enter SelNone. That took 17-18 seconds.

My system is a 5950X, 32GB, RTX-3080-10GB, Win11Pro and Rhino8. Why is an old 580X dishing out 85 frames per second compared to the 60 of my RTX3080? A long time ago I owned two 580X (Normal ones from Asus for Windows PC, not the Apple specific variant) but they have long been sold. I always felt like the Radeon cards had great viewport performance and better viewport quality but they were bad at rendering/raytracing. What’s going on here, why is that old 580X dishing out more frames per seconds than my 3080?

If anyone could confirm this, then my finding was valid. AMD graphics tend to handle wireframe viewport faster than Nvidia’s at the same category/level of competing cards. I tested them with Rhino 6 and 7.

This is not graphics card related and is due to several different parts of the UI trying to update state based on changes in what is selected. I am working on improving this and will hopefully have something that feels faster in the first 8.10 release candidate that should be available later this week.

@stevebaer
I can’t know if the following observation has to do with performance:

In the shade mode options set the dynamic display for objects to bounding box.
Load the trees file with the 10x10 arrayed copies.

Pan the view, you’ll see a big green bounding box.

Delete all the green loafs but one, or isolate one group.
Now pan the view, and you’ll see the same bounding box as before, much larger than it should be.

I can’t provide a file - once you save it and then revert, the bounding box is ok.

Just an observation.

RH-82891 is fixed in Rhino 8 Service Release 10 Release Candidate

1 Like