The SDS cache fix would technically be a bug fix, no?
So you tell me that the 10x increase in animation playback is without gpu open subdivision!!!
1000% speedup, that sounds like a freaking dam was blocking it hahah but i wonder if it’s a general speedup or case by case!
I can’t say for sure, but I assume Sergey is working on that “Unbreak Now” report (which highlights the performance issue as being an issue with the current code, not the lack of GPU-based calculations).
4 Likes
i tested new git master
same issue, nothing new for me(
Well, the patch isn’t in yet, right? Pay attention!
The commit logs show nothing related to subsurf optimization yet. Sergey is likely working on it in his own branch not connected to the developer site.
oh, thank you
so apparently today Sergey was busy with this
https://twitter.com/tonroosendaal/status/1263120804537040897
and it’s his birthday
18 Likes
That’s really cool and generous. I hope those have no drives or SSD drives. Yahoo lost something like 20% of their server drives, back in the day, moving across a parking lot with their servers on carts. That palette jack on that brick looks like hard drive destruction.
5 Likes
Look at the smile on the face of those kids… 
That’s just Yahoo being Yahoo.
I’m sculpting a model in Blender 2.82 and it has 150,000 triangles, but the performance is quite slow despite being on an NVIDIA GeForce GTX 960M. Am I able to allocate more memory to Blender to use?
EDIT: I copied the model into Blender 2.79 and it’s buttery smooth with no stutter.
EDIT 2: I copied the model into 2.79, closed 2.8, reopened it, then pasted my model back into 2.8. Now the framerate is just fine.
Just a heads up, it was decided on Monday that numbers will be used instead of letters. IIRC one question was what would happen if there needed to be more than 26 patches to 2.83. Is aa the 27th release? Another reason was that a & b are sometimes confused with alpha and beta. So it will be 2.83.0, 2.83.1, 2.83.2, etc.
1 Like
Following Sergey’s advice mentioned here, i’ve patched/hacked the code so that the vertex sharpness comparison function always returns True.
The good news is it absolutely leads to a great speed up in the bug report file case (a wave modifier + subsurf), however it doesn’t lead to any noticeable speed up with rigged character.
The custom build, may be unstable and buggy, use it at your own risks:
(Win 64)
Edit: After further investigations, it seems it only improve performances if meshes contains a lot of non-manifold vertices (open borders). Closed meshes don’t really benefit from the improvement.
5 Likes
Messing around with it a bit, I had forgotten the feeling of having decent performances… thank you for this
this is from sergey:
Next week
- Finish OpenSubdiv performance fix
https://wiki.blender.org/wiki/User:Sergey/Foundation/2020#Week_438:18th-_24th_May
next week will be historical 
12 Likes
Will it be for 2.90 only or will they make the fix for 2.83?
1 Like
According to the rules of Blender Foundation 2.90, change are to big for 2.83 at this time. Only bug fixing, called Bcon 4.
But if they call a not good feature a bug, they will make an exception.
2.83 is long time supported (LTS) so I guess it will make it in.
Cheers, mib
2 Likes
Brecht Van Lommel (brecht): Since this is set to unbreak now priority, assuming it is for 2.83.
5 Likes