-AMD FX 8370 8 core CPU @4.2Ghz
-GTX 550ti (upgrading soon to a GXT triple windforce 960 :D)
-12 GB of CL9 1600mhz system memory
A large mesh, well actually, even meshes of only moderate poly counts will lag a notable amount in edit mode. I’m very disappointed. I took this one mesh that was giving me this lag trouble in blender even just on some of the sub-parts of it, combined it all into one mesh, sent it to maya, and entered face mode and it ran just as smooth as it normally does. Same situation in 3DS, Modo, and C4D. All of the other programs ran perfectly smooth with this huge mesh in their respective wire-frame overlay face selection modes or what-have you.
Blender is a fault here. It does not handle the polys well at all in edit mode. I heard of some sort of viewport improvements were coming, perhaps they intend to sole this issue with that? I sure hope so, I can hardly bare editing larger poly meshes >:C
The Blender viewport can be quite slow in modes like editing because the drawing code, for the most part, is basic GL_begin > GL_end immediate mode stuff.
There are moves though to eventually drop the fixed function pipeline in its entirety and expect the user’s machine to support Open GL 2.1 and above (Blender only requires version 1.1 as of now). The Viewport FX project is currently seeing the first bits get committed to Master, but the main purpose of it is to create a good foundation for massive quality/speed improvements in the future and as expected we’re not seeing miracles right away.
Actually I’ve had a similar problem, an imported mesh (*.obj) with 128.000 faces get’s almost unusable in edit mode, the viewport just can’t handle it. It becomes really hard to navigate the viewport.
The weird thing is, that if I go to edit mode on a dyntopo sculpture with 1 million faces it works fine, a little slow but not unusable… Is it because one mesh was created in Blender and the other one was imported…?
My apologies. It read to me a lot more like you were looking for a solution to a problem.
In any case, it may be worth it for you to respond to ideasman42’s question. What, specifically, is lagging? And can you be more specific about what you’re classifying as “large poly” and “moderate poly”? Actual numbers would be useful.
It’s so slow because the array modifier is enabled in edit mode. Don’t do that with a mesh like this. That’s not to say that editmode performance is stellar with dense meshes, but who edits those dense meshes by hand anyway? Just because “the other apps can do it” doesn’t mean it is important. In sculpt mode, it is important, but that mode also uses a different datastructure with different tradeoffs.
Also, a lot of what people file under “viewport performance” has nothing to do with display. If you kick off a big computation, Blender won’t be responsive, because the UI can’t update in the meantime. That’s unlikely to change any time soon, so arrange yourself with that fact.
It’s so slow because the array modifier is enabled in edit mode. Don’t do that with a mesh like this. That’s not to say that editmode performance is stellar with dense meshes, but who edits those dense meshes by hand anyway? Just because “the other apps can do it” doesn’t mean it is important.
Agreed, it’s the array modifier. (BTW if delete array mod and enable instancing… wow, I mean in object mode viewport, and while building cycles BVH of course)
However, apply the modifier and see what’s happening.
“who edits those dense meshes in edit mode?”
Well, any suggestions to close possible holes? Or to join, bridge two meshes?
Do I hear “boolean modifier”? Really?
Anyway, a question arises (about instancing performance in 3d object viewport)
It works! Why not using instancing code in array modifier?
I’m not saying there are absolutely no use cases. If you want to do these things, you can, it’s just going to be a bit of a pain. Is it important to optimize for that? I don’t think so. BMesh is designed for flexibility, not performance. If you want to fix up e.g. scan data, you’re probably better off with a dedicated tool, maybe Meshlab.
Why not using instancing code in array modifier?
Because the output of a modifier (and the input to a modifier) is supposed to be a mesh, not a bunch of instances. That means you can freely stack modifiers and everything works as expected, both from a user and developer perspective. If you want the efficiency of instances in conjunction with the array modifier, you can use duplifaces.
Ok, this seems to be view-port performance then.
We’re well aware this is a weakness in Blender at the moment.
There are some plans to really ramp up effort here and invest time to see improvements get into master,
they will be announced in more detail soon.
I was curious if this was viewport performance (GPU) or edit-mode performance (CPU), they are fairly separate, both can use some work but mesh optimizations can be done any time and don’t depend on larger view-port projects.
Semi old thread, should this reply go elsewhere ??
Most likely same reason as our studio, that in zbrush its almost impossible to work on the inside of a mesh ( think tunnels) , whereas blender does it seamlessly, but as noted zbrush handles a HUGE # of meshes with ease, although zbrush ref. is a bad example I imagine since zbrush doesn’t have a editing mode ? ( not sure , I’ve not used zbrush much at all).
I’ve seen great increase in perf. for sculpting , so ty so much devs for that! ( what a relief)
Editing is still impossible for me here as well, phenom II 3.2ghz ( 4 cores) gtx 580 8gb ram , on a say, 878k mesh.
Don’t ask , its for a game , where detail IS extremely important in some areas, and no normal maps only work, sometimes and are often a joke when you need very precise details.
Anyone know of a eta on edit mode improvements , would love to know because I may have to put some cash down on ‘name 3d app’ if its going to be a while. No 3d software is perfect, different needs different software needed, at least atm right ?
One last thing, sculpting here in blender is much better, but it takes forever ( unless I started a sculpt here ?) to get Into dynamic mode on a big mesh ( current one: 878 k tris) which I can deal with whatever, but is there no way to use a sculpting ‘brush’ similar to what sculptris does, which follows topology exactly ?
I love that about sculptris, and I use it daily here but I"d love to be able to do it all in blender, but I get that adding these surely is NO small task.
This issue is beginning to affect my workflow quite a bit - editing reasonably complex objects (~400.000 - 600.000 verts) in edit mode is far too slow at the moment.
In fact, it is embarrassingly slow when compared to mainstream commercial software (max, maya, modo, c4d), up to the point of being completely unworkable. I downloaded the trial of Modo (just to compare), and I could not believe the performance of poly editing medium heavy objects (~600.000 verts) - an utterly smooth experience - no difference in performance between object and poly editing.
I know Ton and the developer gang are acutely aware of Blender’s Edit mode performance issues, and I would like to ask here whether anyone knows what the progress is to resolve this? It really is becoming a major issue (for me, and others I know as well), and this has been the case for the past years.