Terrible Performance on a very new Imac?

Hi folks,
i’ve come to see if anyone has had similar problems to mine - i recently shelled out quite a bit of cash for a new Imac (late 2020, i7 8core, 64gb ram and 16gb radeon 5700xt) after i realized that my old workhorse had started to get rather sluggish (2015 imac, quadcore i5 with 32 gigs of ram and a 4gb graphicscard).And whilst all my work in the Adobe Suite has absolutely profited from this level-up, sketchup seems even slower than on the old machine …i’m a tad lost here…???

my benchmarks are quite decent (at least the ones i tried so far)… systemwise its 22 Sketchup pro running on monterey with the obvious 5k monitor and a 27inch cintiq on the side.

if anyone has any hints at why the performance would be so catastrophic (which it is) i’d be very glad to know them (or be pointed in the right direction)

sincerly

markus

Can you check if any of your slower models have Profiles turned on? Turning that off can make huge difference.

yes colin, turning off profiles does of course help (a tiny whee bit) - but the whole machine shouldnt be a laggy suckmonster if handling just 4mio edges…amiright?

Since i have no reference points to how many edges/faces should be a good count to handle i browsed the forum for a bit and found a quote in another “laggy mac” thread:

(…I’ve had for just over four years an iMac late 2015 Retina 27" i5, Radeon R9 M395 2GB, main memory 24GB with a Fusion drive.
That’s arguably a slightly better spec than your machine - faster single thread CPU GHz, but i5 vs i7. It bogs down on a HUGE model I’ve helped to develop - 800MB and still heading bigger, but works reasonably well on a part of that model.
The sub model is ~300MB. 27 million edges, and over half a million component instances, loads in under 2 min, and with about half of its MANY layers/tags (too many to count accurately, but at least several hundred) turned off, is quite speedy in response to orbiting, zooming, and editing.")

so…hes basically on an even more outdated system than my old one and is able to work with 27mio edges quite reasonable…while my machine starts to cough when handling 4-5mio edges …somethings pretty much off here and i’d totally like to know how to f*cking fix this…any takers? are the new versions diddleduddle? do i need to go back to mountain lion and sketchup 17? is sketchup written for pc and just poorly ported to osx? whats the deal…

I would call a 4 million edges model enormous instead of “just…” Are you sure that the figure comprises all components and groups too?

1 Like

Hej Anssi, i’m not necessarily into debating wether thats “enormeous” or not (many of the files i have to work with are roughly in the area from 3-10mio) , but more into - why could my 5 year old computer (and obviously other peoples older workhorses) handle that kind of stuff without bogging down too much and why isnt there a significant increase in usability with a modern multicore processor, double the ram and quadruple the gpu ram…(especially when all other applications are in joy about that (ps, ai, c4d, fusion360, etc)

i screenshotted the file info for you. (i’m a graphics guy, not a computer whizz…albeit i know that you have to be a bit of both these days but…well my biz is aesthetics, not hardware.)

(and yeah, i know its more than 4 mio know cause the client just updatet the file to include even more stuff…)

Any chance of sharing the file?
Is it different when the cintiq is disconnected?

1 Like

Not sure if you knew that Sketchup can not take advantage of multi-core processing. Its limited to a single core. Not knowledgeable enough to explain why, but it would certainly explain why your Adobe suite loves more cores and Sketchup just shrugs and says, “whatever”.

2 Likes

3D modelling applications (SketchUp too) are single core, so it is single thread performance that matters.

RAM is important only when you are running out of it.

My own models are usually much smaller but yes, the model in your statistics should run OK, if a little clumsily, on most better modern hardware.
I don’t know the exact difference between your old and new computer, but it sounds to me that you have been expecting a larger performance boost than what actually can be delivered. i7 processors are not automatically faster than i5 or even i3 in these single-core tasks. Especially processors with a design priority of many cores seem to have a slower clock speed or other things that hamper single core performance.

Hello Keith, thanks for that info…no i didnt know, that might explain a bit (and lets me wonder since we’re in 2022 and multicore cpus have been around for a while…why thrimble aint supportin that stuff…^^) but anyhow - the cpu should still be faster than a 5 year old one, no?

Hello Mike, the perfomance is not significantly better when i turn off the cintiq (and i certainly wouldnt wanna miss my second monitor) or switch monitors and put the viewport on the cintiq. For this peculiar file i’m afraid i cant, nda and such but it doesnt really matter - i tried old work files and also created new files with random assets from the warehouse - once it reaches roughly 5mio edges the machine goes to its knees…

There is a benchmarkmodel that tou can use:

I don’t use a Mac but I seem to remember one thing: When you have several monitors connected, you should run SketchUp on one that is designated as the “main” monitor. I wonder how the system decides which graphics system (AMD or Intel) is used on what monitor and application. You can check what SketchUp is using on the Preferences>OpenGL>Graphics card details window.

One would want to make that conclusion, yes.

I have almost the same machine as you. Have the i9 with 32 gb ram. I don’t ever follow how many edges my models get to so can’t comment on the performance that way.

I will say that all things in your workflow being the same from one platform to the next seems to put the issue on the hardware. I have always decided that I prefer to spend more money on the top tier hardware and less on the frustrations of poor performing hardware that I chose to purchase because the cost seemed better.

That might be a bit blunt, but I purchased a 2019 MacBook Pro with the i9 and 32 gb of ram with a lousy graphics card in 2019. Tried to make an external graphics card work and nearly drove myself mad. Upgraded to the i9 iMac last year and what a shift in performance with the on board 5700xt. Having spent some time playing with the new M1 Pro MacBook Pro, I’m already planning to make the jump in the fall. I’m just done with trading budget friendly purchasing for the years of frustrations that follow. The backend costs, to me, far exceed the initial price of the top tier platforms.

This is not a mandate for everyone to follow…it’s my own experience. I think your use of the Cintiq makes your workflow a lot different than most users on this forum. Ansii might be on to something there. Might be that Apple changed how the new platform handles the external monitors like the Cintiq and it’s now causing a resource leak that wasn’t there in the older platform.

That’s a terrifyingly large file from just a file size standpoint - is suspect you are pushing SketchUp right to the brink there. Strange that you say it works worse on your new machine.

Probably a good place to start is right run ThomThom’s cleanup3

Purge Unused
Merge Coplanar
Merge Duplicate Materials
Repair split edges

Is probably a good place so start- I’d be interested to see how much alone that clears things up.

Another thing to investigate is what anti aliasing setting you have on. MSAA can be very expensive and you are running at 5K resolution . Maybe turn it to 0x whilst you are troubleshoot .

What Adam said about 0x antialiasing may help too. On Retina displays the antialiasing isn’t really needed.

thanks adam and colin, i already reduced the anti-alias to 2x, (0x looks insulting to the eyes even on a 5k monitor) and it…helped a little…

will check out the tt cleanup thing, see if that helps, but i’m not giving it much hope, tbh…

maybe will have to reconsider workflow completely and move to a more suitable archviz option (altough i’m basically stuck with it cause a handful of clients/companies i work with are throwing only skp @ me)

This is a mystery…

just grasping at straws here…haven’t seen anyone mention drivers. Do drivers cause this kind of issue if they are not updated for a new platform?

so…at least rendering is a blast - thats like rocket speed in comparison to the old machine, i guess thats got to do with doubling the ram and quadrupling the gpu ram…still, working in viewport mode is…slow and painful. as for that - i’ll widen my question to all of you sketchup wizzes (i’m just a moderate dabbler, gotten into it out of necessity due to clients demands and such) out there: what are good pipelines/workflows with sketchup in the archviz department? i’ve read a lot about sketchup ->lumion, sketchup → 3ds, sketchup - any other modeller - cause once your (skp) model reaches certain amounts of assets it becomes quite unworkable, right? (my guess/hint would be due to single cpu prgramming, this very odd thing with the circles(or not circles) that sketchup does which upps the polycount to insanity)

so if you have to deal with very geometric shapes and not much foliage/plants/sheets/funny furniture etc you can go sketchup-vray but if theres lotsa outdoorsy stuff or much polygons in the mix one would export to another, more…suitable 3d programm to finish of there…m i right about that or is there another way that have not found out about? curious for your answers

chinchin,

markus

Non-realtime rendering applications like V-ray have used multiple threads/processor cores for years so they can use whatever your computer can offer. It’s even faster on a PC with a Nvidia graphics card as it can also use that.
Modelling applications on the other hand all are single-threaded.

be that as it may - i guess mathematics are to blame (something like linear movement in space etc cant be broken into different threads) but i’m still…somewhat taken aback at the comparatively low amounts of polys at which sketchup caves in(a fact that quite a few other threads here on this board have discussed at length already), even on a fairly modern machine (yes, i acknowledge - its not a mac pro or imac pro, but still…) anyhow, not gonna change to a pc based system for now, so i guess i’ll have to work crossplattform/programs…

Maybe I missed it, but how can you tell which machine spec resulted in the fastest time?

I couldn’t find the Ruby Console in 2021 so I loaded that test model in 2018 and got 29.4169 on my i7 PC with 16gb ram. At least a 5 year old machine.