GPU Accelerated OpenSubdiv is coming to Blender! Sergey Sharybin shares an update of his latest work in this video. The amount of acceleration is breathtaking and will change the way you work with highly subdivided models.
Ton Roosendaal even suggests that this will be available this summer (at SIGGRAPH, I bet!):
OpenSubdiv coming to Blender this summer! Sergey and Francesco made a cool video to explain how it works. http://t.co/K1NjV1cImI #b3d
— Ton Roosendaal? (@tonroosendaal) May 12, 2014
24 Comments
I might not always see eye-to-eye with direction of Blender development, but this steady development of OpenSubdiv is certainly one I think is on the money. Talk about a game changer. This will do wonders for improving polygon performance within Blender.
Fine job, Sergey!
Indeed, I foresee this could possibly do away with having to do re-topology to do rigging and animation after sculpting, which will be such an timesaver.
Don't forget about the Quad-Dominant remesh modifier (Interactive Quadrilateral Remeshing), too! That, coupled with OpenSubdiv, might just revolutionize Blender towards becoming a major player in the industry. I think it's the start of a nice new direction for Blender.
Not totally sure what the Quad-Dominant remesh modifier is, but it sounds like it would help too. I agree with you, these are extremely important steps towards Blender becoming an "actual" competitor in the field of 3D creation software... well, more so than it is now.
You know the Dynamesh and ZRemesher features in ZBrush? Quad-Dominant remesh is something close to along those lines.
It's a new remesh development that works better than the current Remesh modifier in terms of producing quad-dominant topology, interactively. Seriously-exciting stuff.
http://code.blender.org/index.php/2014/05/google-summer-of-code-2014/
Ahh, I get it, basically it does the retopology work for you by generating an animation friendly topology as the sculpting is done. Thats really cool!!!
Yeah, and it also makes things like merging meshes together a far cleaner result. I would imagine that DynoTopo sculpting can now be cleaned up sooner, making possible a phase for cleaning topology as you go along.
Basically, we're looking at the world's first freely-available "Dynamesh"-esque solution...which is huge. This feature alone has the power to bring new interest to Blender, both increasing the professional user base and motivating better interest with third-party support.
We have V-Ray officially supporting Blender now. We have Epic Games helping Blender with its .fbx support. And Robert McNeel & Associates hired a Blender developer to integrate the standalone version of Cycles into Rhino.
I think we're finally starting to see Blender gain recognition of its potential to other major businesses, as well as our Blender developers becoming a major contributor to the industry. All thanks to Blender taking on game-changing features. All very encouraging stuff!
Blender for President 2016! ;)
This is really great news!
Anyone know if it is possible to download a test build?
don't know if it's included already in development versions, but here's a link to the latest builds
http://builder.blender.org/download/
You made cry Sergey! ..in a good way! ;)
This is great.
I was wondering if crease support will be added.
Sigh... with my old HD Radeon 3450... :/
Won't make a damn difference.
Didn't you see the 20x speedup for CPU only, at least for animating/editing subdivided models, you'll see a huge difference ;)
I hope so. My hardware is terribly old :(
OpenCL performance seemed to be pretty quick.
I wonder if the GPU has to be connected to the display for OpenSubDiv to make use of it...
That would really help if it did, especially with sculpting.
Just curious, is this again implemented in a modifier way, or will we finally have partial sds with open subdiv?
That's pretty sweet. Can't wait to try it.
As much as I love the development of Blender I believe we will very soon have to face that almost every part of Blender needs to be rewritten. The code is a mess, same goes for the documentation. This makes it very hard for newbies to contribute code. Blender is just not ready for professional production.
I sorta agree with you on that. I foresee a code and documentation rewrite in the near future. Though I still think newbies can make contributions to the code if they have a solid understanding of coding. The documentation is sorta a mess though, and needs to be cleaned up, as it is can be sometimes quite hard to find what you are looking for.
This is what the releases from 2.7 through 2.89 are all about. Putting a freeze on frantic development to slow down and refactor, document, maintain and clean up.
And this is to be expected. You keep adding new features to the software, things break in the progress and you fix them. Sooner or later though, the software becomes bloated and you end up spending more time fixing things that broke than you would going back and doing a code rewrite. Its the natural course of events. It has to be done at some point with software that is in a constant state of development. Minecraft is a good example of this. There have been at least two massive code rewrites in Minecraft's history and after it is done, it always runs significantly better than before.