Hi.
When I use lineart with 3d meshes with subdivision surface modifier, Blender lags and my cpu is working really hard. (This is without eevee or cycles, no procedural textures).
I’d like to know what kind of performance do other people get with lineart and subdiv, and what cpu they have?
How useable is Blender?
Is it very smooth performance?
Is it a bit slow but fairly useable?
Or is it very very unproductive and it’s time for an upgrade?
My Cpu: i7-3770
Several meshes have subdivision surface on.
Level 1 subdiv
35k verts, 34.7k faces
slow. Kind of useable…?
.
.
Level 2 subdiv,
140k verts, 138k faces
too slow. Not useable
.
.
Level 3 subdiv,
452k verts, 449k faces,
even slower than Level 2!
I have to wait a couple seconds after orbiting camera around in camera view.
I haven’t found any lag with the line art modifier yet, but I don’t think it gives very good results with subdiv, so I haven’t tested that use case much. My CPU is a AMD Ryzen 9 5900X 12-Core Processor 3.70 GHz.
Right now I’m in a project with 18,000 faces, 30ish objects, every object has the line art modifier. No lag in material preview or render preview mode- navigating, at least. I’m only getting about 13 FPS playback, though.
I had thickness, smooth, and noise modifiers on the grease pencil objects- turning those off got me 16 FPS playback, which still isn’t 24 FPS, but it’s not bad at all for an intensive modifier like this one.
I tried it on a Mac Mini - I’m getting 5 FPS playback and a ton of viewport lag.
AMD CPUs work really well for Blender, and they’re pretty affordable too. You can get a 5900x for 340 right now, and with the new generation coming out, they’re going to keep going down in price.
Also, do you have GPU Subdivision turned on? That should help a LOT
Thanks. Yeah I had Gpu subdivision on (didn’t know about this). It doesn’t do anything for me, probably because my gpu is very very old (radeon hd 5770).
p.s I’m not even trying to play an animation right now. I’m just trying to model a 3d mesh with lineart lol
Ok, I think I’ve found the breaking point for my CPU- with 200,000 faces, 650,000 tris, and complex GP thickness modifiers, I’m officially seeing viewport lag when the grease pencil collection is enabled.
Sorry, one more question:
When your cpu is pushed to its limits with lineart, are all the threads working really hard?
Or only some of the threads?
Or is it just single thread working really hard?
I wanna know if good single thread performance is enough (for lineart), or if I also need as many cores and threads as your cpu (which would cost more money).
Good question. So it turns out my lag was actually from geometry nodes- with geometry nodes applied, I’m not seeing any lag at all with 500,000 tris. Here’s what happens when I disable and enable the line art collection:
The first spike is from opening the file, the big spike is from enabling the collection. It looks like it was on all cores, but it’s not hitting 100% on any of them, maybe 80%.
Here’s my test scene, you’re welcome to check it out and see how your CPU / RAM does with it. Note that I’m using 12 GB of RAM- 8 GB of that is going to Blender at the moment, so if your computer doesn’t have at least 8 GB of RAM, you might not be able to open this file.
You can toggle the GP collection on and off to stress test your CPU- this isn’t an official stress test of any sort, but it’s a pretty good example of pushing GP line art to the limits.
on my system (ryzen 7 5800x + rtx 30700) It looks like the lineart modifier uses about ~%80-90 of my 8 core cpu while rotating a ~200,000 tri mesh set as the lineart target. it does appear to use multiple cores to calculate the line art. it gets about 12.5 fps while doing this
duplicating the mesh 4 more times to bring it over a million tris drops the fps to about 2.5 fps
My test mesh is a fairly high res scan that has been triangulated. large faces can cause issues with lineart, especially when mixed with tiny faces.
My advice would be to try optimizing your scene as much as possible before dropping money on hardware. Optimization is the key to success.
If you do want to buy your way out of this problem, the biggest multicore cpu looks like it’s gonna be the key for line art performance. @joseph and I have similar GPUs, but he’s got 50% more cores than me, and even though I have a slightly higher base clock speed, he is getting noticeably better performance.
I downloaded his test scene, and I am getting .5 fps
If it makes you feel better, I get .18 FPS with that scene if I try to do playback Are you seeing any viewport lag without animation playback in that scene? I can navigate it smoothly with no issues, I just can’t do anything animation related
Navigation is smooth and silky, as long as nothing is changing. I was imagining if @A_Mix_Of_Geek_Stuff was experiencing lag, it was due to updating geometry in some way, so I was attempting to simulate that.
So I guess it depends on what @A_Mix_Of_Geek_Stuff is doing in his scene. Are you animating stuff?
Are you seeing any viewport lag without animation playback in that scene? I can navigate it smoothly with no issues, I just can’t do anything animation related