Blender edit mode Performance

Batch Evaluation of subsurfed meshes.

By the looks of it, it should offer some speedup with OpenSubDiv by way of reducing overhead (which according to developer discussion seems to be a major pain point in performance).


Hope Sergey solved this issue

I just did some tests with the latest today patches … and it seems that nothing has changed yet

I’m too(((

Bought new PC, R7 3700x, 64GB RAM, 2070 Super, new M2 drive. Performance is still bad especially using subdivision surface or undo. It’s hard to work on high detailed models.

In 2.79 modeling (with subdv) and undo was faster, but object mode with hundreds of meshes was very problematic,
in 2.81 modeling (with subdv) and undo is really slow, but there is no issue with having hundreds of complex scene and operate in object mode.

Since Blender want to be compared to softwares like Maya or 3Ds Max, I think this issue should be considered as number 1 to fix. I heard already some complaints from my friends that tried 2.81, also asking me what’s wrong :frowning:

I don’t want to be impolite. Just want to say that from professional perspective this issue keep Blender slightly in the background. If more people will notice that issue, they will bounce of from that software as fast as they wanted to test it.

I’m really happy what Blender 2.81 is right now, and looking forward very optimistic to future :slight_smile: Overall, great work devs!


these issues are known, and should be resolved in blender 2.82

They will not be resolved in blender 2.81 because they need a technical rewrite of some parts of blender, and then they will have to be tested for the right time in alpha and beta mode.

A branch already exists for the UNDO system, it should probably be made more solid and then implemented.

As you will have read on the previous posts, for the openSubdiv accelerate on GPU, first changes have already started


Happy to hear that! :slight_smile:

I’ve heard nothing about this being ‘resolved’ in 2.82, where did you read this?
They’re working on it, but it’s not an easy fix, so I’d be very surprised to hear that.

2.82 release is like halfway into next year I think so I’d be super bummed if they weren’t fixed by then. I don’t want to see them kick that can down the road another release.

Well, hopefully for the next 3-4 months they will make real progress on this … obviously nothing is absolute, I thought it was obvious by itself, but I believe that if these 3 problems are known, undo, slow heayy poly mesh editing, and slow opensubdiv perpetuate even after blender 2.82 will not be fruitful in terms of reputation, competition will fossilize these issues in the collective consciousness …
You can already imagine, “blender is a fun toy, but for serious productions it cannot be used”

So blender foundation has every interest in breaking down small dragons before they grow too much and take flight.
So even before adding new features, they have all interest in resolving once and for all these known problems in this release.

1 Like

Please don’t spread misinformation. There is absolutely no such commitment from the devs.

If I’m wrong, by all means, link it, I’d love to hear that too.


hey I’m not spreading misinformation, these issues are tagged for blender 2.82

As long term projects.

And even if that weren’t the case, a tag is hardly binding.

1 Like

Wow. There’s no need to be rude, I had no intention of angering you. I just stated fact.

Just because some stuff is tagged 2.82 that doesn’t mean it’s getting into 2.82, only that there’s not enough time to deal with it in 2.81. There is no 2.83 tag yet. This is not some remote possibility like sudden death, stuff gets moved along to the next release all the time. First they meant to fix all that before 2.8, then there was talk of fixing it in 2.81, and now the plan is 2.82. In two months the plan might be 2.83.

All the devs ever said about performance boils down to ‘we’re aware of it’ and ‘it’s hard’.


And it is for this reason that I responded to Micheal_Knudden that nothing is absolute and obvious, but that it would be interest of the blender foundation that these inconveniences are closed once and for all.

But instead of seeing this, to which you yourself came later, you thought I didn’t give you the right information …

This is ridiculous. I guess some people just want to be angry.


Man i believe that we are eclipsing on irrelevant details, maybe induced by misunderstanding.
So peace and go over.

3-4 months time is a long time. They need to show significant progress by then, or be ready to. Just today, a game company posted a bug report about how they’ve made the transition to Blender for production and are finding themselves dealing with performance issues.
There is a huge interest from the industry in Blender due to things like UX/UI changes and Eevee, but if big companies get burned by performance and it ends up costing them money, they’re not easily coming back to Blender.

That said, with performance being so fundamental to a DCC, and with companies like Ubisoft being now heavily invested in Blender, I think it’s only likely we’ll see progress soon. If not by the main development team, I should think/hope that now invested companies will find and contribute their solutions to Blender.

1 Like

They at least have things like edit-mode performance being todo items, so they recognize that it is a problem for many users. Some devs. also have done profiling on things like OpenSubDiv to identify performance issues.

The situation already is a little better in the current builds compared to 2.8, but many companies have administrators who install the software and will often stick to release builds.

Blender 2.81 in many parts is on a new planet in terms of performance compared to blender 2.80 …
The fact remains that the known problems remain, but it is normal, I know that they have made a preparation of a series of optimizations before facing them directly, also because they were more complex technical problems that require more time, and this is why we will see concrete results (we hope) in Blender 2.82