Sorry correct me if I’m wrong, because I might have missed it. No performance patch is planned for 2.93 and I’m cool with it. But is there ANY toughts at least on 3.0?
Don’t want to be repetitive with spam and whine, but today it’s been an especially harsh day. Trying to hit tight deadlines with these performance is… stressfull. I’d like to know if someone has some info as I might have to revert back to 2.79 or even find other alternatives for the sake of my mental sanity.
I didn’t read the whole thread, sorry if I’m repeating or missing something here.
Just wanted to add that performance has become quite a bottleneck in my workflow as well. I agree with what has been said: the work that has been done so far by the devs is huge and remarkable with all the new features added.
Unfortunately, it gets frustrating when it comes to dealing with the edit mode performance even more so considering how much Blender now has to offer.
More often than not, I end up resorting to C4D for heavy tasks, and C4D is not exactly the fastest 3d software.
Performance is undoubtedly the keyword that comes up every time Blender is mentioned in other 3d apps forums: it seems almost everyone now agree on how much Blender has advanced and gained over the past few years though they always mention the one thing that’s keeping them from adapting Blender to their pipeline, that is performance.
I hope there’s still room to allocate some resources in this area.
It’s very often the unflashy features that make the flashy ones really shine. Performance and quality of life. When sorting by “Top” on RightClickSelect the majority of feature wishes aren’t super obscure things. Most of them are things that improve basic and very frequently used tools or workflows.
This was also why the papercuts initiative for 2.80 gained so much traction, I think. Frustration arises from things that go wrong or are unpractical in workflows which need to be done over and over again. These things become apparent throughout the whole work experience the whole day long.
There are two reasons to come back to a software each and every day:
Ideally a mixture or ballance of both 
Is it me or has edit mode performance gotten far, far worse? I’ve been using Zbrush exclusively for the last 6 months and have just returned to Blender for my current project and the perf is absolutely diabolical. Much worse than before. I’m using 2.93, and what’s more, the machine I’m using is much more powerful than the one I was using last time I worked in Blender.
What is going on? Even on a simple mesh with a Bevel mod and 2 levels of subdiv and only 64K verts(when subdivided) the whole thing is bogging down adjusting the bevel width.
2.8 has been a wonderful experience for me, having switched from almost 20 years of 3dsMax, I couldn’t imagine going back to model in Max because I LOVE Blender’s workflow, but wow, the terrible performance in Edit Mode(which has obviously been there since I started using the program 18 months ago) is so frustrating, considering it’s such an important aspect of any 3D program. I realise it takes time/resources/etc and understand, but surely it should be a top priority at this stage? Even chatting among pro artists from Maya/Max they ALWAYS, ALWAYS raise the issue of the poor perf in Blender. The fact that it’s actually gotten worse(from my example/experience above) is quite worrying because before I could cope and keep working, but now… it’s bringing me back to 3dsMax level perf from 15 years ago.
Reading the commit logs and the developer site, I couldn’t help but notice how in the case of geometry nodes, performance is being seen as a priority.
All we are asking is that this also becomes a focus when it comes to editmode and OpenSubDiv. We know you guys are more than capable of bringing performance up to and beyond the industry standard. Having high performance in the node system is great, but remember to also give attention to what is usually the bread and butter of a DCC app.
I open 2.79 every once in a while, and I am actually amazed at the snappiness of everything, from moving around in the viewport to editing a mesh. And 2.79 wasn’t the snappiest of the bunch either, so would like to chime in an extra voice for the huge importance of performance improvements in the viewport.
Since devs decided to shift priorities and did not give any time estimation regarding addressing the issue, I’m thinking if it might be possible to kickstart a crow founded project (between us artists) to hire an external developer to address this issue.
Is this a valuable option?
You are free to do it if you think it’s worth it but honestly, given the complexity of the task, good luck finding a developer that has this kind of skill and that hasn’t already been hired by another company.
With that said, i agree that performance is a very critical issue right now and even as a very casual user i have had to put up wit it so I can imagine it is much worse for professionals.
Maybe the switch to Vulkan will help a bit?
Since every change in the code needs an approval from the developers it is simply a waste of time to make any patch even moderately complicated.
It has always been a long debate how this kind of “patreons” could be run and in the end the only clear thing is that only the foundation can organize that or a single investor with money and good friends with the developers.
… or fork it and most likely be responsible for maintenance on every new update you want to use (since there won’t be many people willing or able to help with deeper optimizations, I guess.
Or have the funded developer in question get in touch wit the Core Devs if you plan on developing this patch for the main branch, I guess.
Or better yet - get in touch with the module team/owner directly before trying to start a funding campaign and ask what the recommended approach would be in that case.
I assume @dfelinto might be able to answer this question most directly.
Can’t Patreon Members like Facebook and Epic require that their money goes into certain aspects of development?
If that is the case, couldn´t a bunch of smaller users finance a developer to work for the Foundation but for a certain area of development?
I remember reading somewhere that a couple years ago a company/or someone specifically gave Blender money to develop a module in blender or to further develop it . . . i think it was the Blender game engine.
Which blender dropped as we all know
The way I understand it they can make requests but it has to align with what the developers are doing or be compatible with that. Donating a large sum gives the advantage to actually make specific requests but it does never grant the right for these requests to be actually implemented.
So yes - you can nudge and are at a better position to have that nudge be heard but it’s never a transaction.
Which is a good thing in the regard that nobody can simply buy into changing Blender their way but is a bad thing for when you want to give specific money for a thing that the core developers simply don’t want or currently can’t for one reason or another.
Bottom line - everything you donate via the fund will always remain that: A donation. Not a transaction or rights
I think if there really is the intent to have money involved or hire somebody to contribute to the core development then actually talking to the module owners and coordinators first would be the best thing to do. They know what’s going on and how it could be included or if it would collide with their own plans etc.
Isn’t Kevin Dietrich sponsored by Facebook in order make something specific (Alembic?) work better?
If so, couldn’t a pool of Blender users do something similar? Would be difficult because you’d need at least 6K Euros or so per month for a full time developer but the initial question would have to be if the blender foundation would even allow such a constellation.
Maaaybe? I don’t know how much or if in that case it would be better to have an external developer contribute from the outside like any other user contributing to an Open Source project. I don’t have enough knowledge in this field. I would love to know, though.
The way it is now I don’t even know how the internal developer status is regarding speed ups. Maybe they are working on it but not full force or are stuck in a way. Maybe an external dev would collide wit te curent progress …
That’s why I was trying to summon @dfelinto or maybe @brecht to this thread to clear things up.
I tried this at one time, and it was a waste of time. As soon as the module owner knew that I was not a programmer, he understood that this approach would have no future and preferred not to continue talking.
Yes, I got your point the first time you wrote it.
I do think that it’s still worth trying again since circumstances change over time and may be different here.
Asking with a specific plan doesn’t hurt but may lead to results.
Giving up is the only surefire way for no change. 
It was about a related issue that we are talking here… so, I don’t expect a lot of changes… 
Poor performance may be caused by the python data structure.
Blender “official” developers hired by the BF are not standing against external patches, however yes, the problem is they want the patch to be coded exactly the way they want.
Regarding OSD, a guy named Jack Quiver had started a patch with interesting improvement regarding performances:
https://developer.blender.org/T68996
But he stopped because according to BF dev, he was not going in the correct direction (and that is certainly true) for optimal performances and code management.
Then yes, finding someone skilled enough for this is hard, and someone who knows Blender code enough is even harder, and someone who accept to follow line-to-line BF devs instructions and corrections (often for free!), it’s almost impossible.