Benchmark for SketchUp

If you use the GPU render options in V-Ray the difference between GTX1060 and RTX4000 might have an impact on render time.
I think that your very detailed clothing components, too, might have potential for optimization, especially those that only make your background. I know nothing of the fashion industry, but I would use more generic entourage objects, especially as I understand that what you show is bound to be last season or older…

Test 1: Default settings = 34.3371 fps
Test 2: Edges, Profiles and Dashes off = 60.5476

Edit: Dell U3818DW 4K
SU v 20.2.172

Dell Precision 3630
Intel i7-8700 @ 3.20 GHz
32 GB ram
RTX 2080
512 GB M.2 PCIe NVMe

Also of note I ran Ambient Occlusion each way and the default view rendered at 1080 was 42s and then 40s when I turned the Edges, Profiles and Dashes off. Rendered images looked identical.

results without the used screen resolution (or the resolution of a scaled SU window) are pretty useless.

There should be a default “faster” modeling style (with shading and textures) shipped with SketchUp, as an example to new users.

There are 5 faster default styles included with SketchUp. One of them does include textures.

1 Like

Good. I should have looked, because I usually just see my styles. I don’t see them in 2019. “Shaded with Textures” might be a good one. But if it doesn’t say faster, how would new users learn the difference?

Most pro users I know never have profiles on.

Do you know why Profiles take so long to render? To a layperson like me, it doesn’t feel like Profiles (or dashes, or other effects) would add a high computational load compared to regular edges.

I’m of the view that Profiles should be used for displaying edges we want to look thickened; not for “outlines” and to highlight non-closed surfaces.

Non-closed surfaces should be highlighted instead with some type of effect that is applied to edge gaps and overlaps/small edges that is visible during the modelling process. (similar to the Edge Tools extension).

Profiles are “heavy” because SketchUp has to figure out which lines need to be heavier every time the model moves. Based on where the camera is looking at the model, different lines are at the edge and are made heavier. Each time you orbit… each time the model moves one pixel, different lines are at the edge. It is a lot to figure out!

1 Like

I just built a fresh PC with a AMD 5600x, RTX 2080 Super, 1080p 60hz output and default SketchUp 2020 (no extensions). The Test Time Display model is the very first application I installed :smiley:

My first tests (@1080p 60hz) got 30 frames per second…very disappointing…!

Then I went into Nvidia settings and found something called Power Management… I turned that to Maximum (instead of Power Optimised) and got 49 frames per second. Much better (but still slower than my old HP laptop).

Threaded Optimisation actually lowered the score to 45fps.

After a couple of sequential runs (kind of cheating?) my best score is 58 frames/second.

1 Like

in the nVidia 3D settings SU is configured to use the “high-performance GPU” (and not a maybe integrated GPU… dunno if the AMD CPU does have) w/ settings configured to “let the application decide”?
AA settings in SU OGL prefs?
Windows power management settings configured to “High/Ultra Performance”?

I left AA settings default (App managed) and SU default is 4x.

Didn’t touch windows power management - good idea. This setting might explain why one Test Time Display run can be very different from another despite all other settings appearing the same.
I usually open SU and run the test "fresh’ each time (without any extensions, SU opens in about 3 seconds from the M.2 pcie Gen4. This makes me wonder if some extensions could be loaded when activated in SKP, instead of everything loading at startup).

After an Nvidia driver update, enabling XMP to run faster ram speed, I’m getting repeatable 62 frames/second.

The tweaking will continue once I upgrade from stock cooling and do some mild overclocking from 4.6 to 4.8ghz.

I will also do a layout rendering performance comparison between systems. 4790k, 5960x, 8650h, 5600x.

Here is the most optimistic video I’ve ever posted to the forum. Meaning, I was being optimistic when I made the video.

This is a 1920x1080 screen recording of Parallels running on my new M1 iMac. What it shows is the Insider’s edition of the ARM version of Windows 10. That version of Windows has the new ability to run 64 bit x86 applications in emulation. In this case it’s running SketchUp 2017.

With Profiles selected it measured about 21 fps, with Profiles off it went above 30 fps.

3 Likes

I get this.

72 frames displayed in 3.4108 seconds
Average frame = 0.0467 seconds
21.4026 frames/second

Is that on an Intel processor MacBook Pro? It’s amazing that I got about the same frame rate while in an x86 emulation inside an ARM Windows 10, inside a virtual machine application.

When I run the test directly in SketchUp 2021, using Rosetta 2, I get this:

72 frames displayed in 2.5374 seconds
Average frame = 0.0348 seconds
28.7699 frames/second

With Profiles turned off it’s this:

72 frames displayed in 1.5866 seconds
Average frame = 0.0217 seconds
46.0091 frames/second

Here’s the same results on my MacBook Pro, which do come out better:

Profiles on:
72 frames displayed in 2.4783 seconds
Average frame = 0.0339 seconds
29.4559 frames/second

Profiles off:
72 frames displayed in 1.2288 seconds
Average frame = 0.0168 seconds
59.4085 frames/second

2 Likes

Here are my machine specs:

MacBook Pro (15-inch, 2017) running Catalina in French Canadian
Processor : 3,1 GHz Intel Core i7 quatre cœurs
Memory : 16 Go 2133 MHz LPDDR3
Graphic cards : Radeon Pro 560 4 Go
Intel HD Graphics 630 1536 Mo

i7-10750H, GTX 1650Ti, SU '21,

1 Like

image
image

Specs:
EVGA RTX 2070
AMD Ryzen 7 3800X 3.9GHz
Asrock X570 Taichi
32G ram

1 Like

@MikeWayzovski I will point out that in your Google Form there’s not an option for SU '21.

1 Like

Dell Precision 7510
i7-6820Hq
Ram: 32Gb - 2133Mhz
VGA: Quadro M1000m - 2Gb
Setting: 64x AA + Use maximum texture size

z2576527660376_73a7cbf609f513b1792c5046c3531edd
z2576528462346_46afb59374d7bfbe4f5f287618d106b7

Would be interesting to see how you get on with a lower anti-aliasing setting. One of the main uses for that setting is so that monitors with less pixels per inch can be made to look smoother. Zooming in on your screenshot shows that the lines are made up from shades of gray:

image

If you are on a HiDPI monitor you can set 0x anti-aliasing, and it will look ok. If you’re on a regular high resolution, 2x or 4x should be more than enough to make lines seem nicer.

With 16x you may be using 256 times as much video memory than you need to. I’m curious if the lower setting would increase the frames per second on this test.

3 Likes