From my own results on multiple screens I think resolution has a lot to do with the numbers we are seeing. Enough that perhaps screen pixels should be a listed parameter. My same laptop gets 21 FPS full screen on the retina display and 39 FPS on an external monitor. Pretty big spread. Could much of the variance we are seeing could be the uncontrolled for difference in peoples monitors?
Could well be I think. I’ll go back to my earlier post and add details of my monitor there. FWIW, it’s an iMac 27" 5K Retina screen.
All of my tests, as well as those of my alter ego @sjdorst-doppleganger, were done on a 1920 x 1080 display with SketchUp maximized and no display scaling.
Both computers, home (me) and work(@sjdorst-doppleganger) have a 2nd display. All tests done with nothing displaying on the 2nd display.
On this random PC I get 31 frames per second.
(2.080 / 0.0285)
SU Pro 2019 and Windows 10 home.
CPU: i7 4790k
GPU: 'Nvidia GTX970, 6gb
How scientific is this test?
It’s great to see people engaging with the testing process and posting some interesting results.
But I’m not quite sure what’s being tested; Im looking at my performance monitor but neither the Intel i7 or nivida GPU are being stretched. RAM isnt near full.
Maybe a longer test (more reliable) or two tests to focus on CPU and GPU performance would be very useful?
Or add a million or two Susans to the model?
I actually like the idea of testing an increasingly larger/more complex model until the FPS gets down to a particular number. That shows us how much “complexity” a PC can handle and tests its limits.
Is there an easy way to display FPS for a scene that we create ourselves?
On a scale from one to ten, none
To test scientific, you would need the same everything conditions but one, eg. Same installation of software, extensions, OS, clean install, same connexion, same screen setup, same monitor, same cable etc etc. And then replace one item (eg. Video card) to see what the effect is.
I believe no system is identical, they are all ‘unique’
The test is merely to look for ‘common ground’ factors ( this processor with that videocard give on an average this and that framerate.
You can also use this test:
Draw a line
Divide by 10000000
See how long it takes
The model is not ‘complex’ , but I bet it will bog the i4790 down for a while.
Complexity has different faces.
Benchmark results before/after upgrade:
Windows 10, SU 2019 Pro , 4k 60hz display, add default settings.
RAM 32GB DDR3
GPU GTX970 6gb
Frames per second: 30
RAM 32GB DDR3
GPU RTX2080 Super 8gb
Frames per second: 32
Just a stab in the dark: In another thread, someone with a similar setup got much better results by going to the Nvidia control panel 3D settings and turning off “Threaded optimization” for SketchUp if it is turned on. I get better results from my 2070 card, and for me that has been turned off by default.
I checked three of mine, and all of them have Threaded Optimization set to ‘Auto’
In twelve years of using both Quadro and GeForce graphics, I’ve never touched the default settings.
There’s not much difference in single thread performance between your i7-4790K and my i7-6700K
I’d be interested to learn what happens if you reduce screen resolution.
Intel Core i7-6700K @ 4.00GHz vs Intel Core i7-4790K @ 4.00GHz
Interesting feedback - thanks guys. I will post a few further test results.
I’m currently using a laptop with i8750h, 32gb DDR4 and 1070max-q
Monitors are either 1080p 144hz and 1440p 75hz
1080p I get 32 frames/second.
1440p I get 30 frames/second.
Turning threaded optimisation OFF (for Sketchup only)
1080p I get 35 frames/second.
1440p I get 32 frames/second.
So yes there is a noticeable improvement in the benchmark with the Threaded optimisation OFF.
I did have a weird set of results the other day when I plugged this laptop into the 4k 60hz monitor, I was getting consistent 52 frames/second. I’ll try to recreate those conditions. I wonder if windows desktop scaling may have something to do with it.
sidenote:| anyone noticed how Profile Edges OFF = 100% speed increase? Also, Depth Cue edges are 25% faster than Profiles.
On this 4790k and 2080 Super setup, I am not seeing any difference tweaking the Threaded Optimisation setting… nor any other settings for that matter. I tried resolution, vsync, windows scaling, and various gpu quality adjustments (and the most recent driver) and found no discernable difference.
Only global FSAA gave a -2 frame/second penalty when turned on.
During these tests the GPU is hardly working. To make it work harder I copied the scene 15 times and made the faces 50% transparent. This managed to get the GPU to work a little bit (up to 35% utilsation and at 15frames/second).
I’ll overclock the CPU next and see what that does.
I’m able to reliably recreate the 51-53 frames/second scores on my laptop:
i8750H with the 1070maxq, running at 4k 60hz (while also showing Windows 10 desktop on a second 1080p display)
Not sure why this laptop is scoring so well…any ideas?
All graphics settings are default except I did change the Threaded Optimisation = Off
So wierdly my big PC Workstation with the 64gb RAM and RTX2080 Super is a LOT slower than my 15" laptop. Yes the CPU generation is newer, but the performance should be similarish.
I have a feeling that there’s something fishy going on with differnent monitors/scaling/hertz/window modes/etc…but I dont know too much about that stuff.
it may be set to overclock when cold and throttle back when hot…