I guess I knew about this, but finally, like a lot of things in blender, got a chance to play with it. Base 500 face Suzanne model.
is it or only microdisplacement ?
which has been there for ever!
happy cycles
Lol hahahahahahaa. Oh dear, you obviously don’t understand open subdiv at all. Read, read, read, and then Read some more.
This has been in Cycles since its first release, it’s just a simple subdivision (like the Subsurf modifier, just on the Cycles side).
And OpenSubdiv is not implemented in Cycles yet, at least there is no public code.
DingToooooooooo!, Dude been meaning to catch you at some point. What do you think the odds are of implementing an GLSL Uber shader in Blender/Cycles and a Bake down utility (there’s open code online for Maya in this department, GLSL can do this but needs some TLC). Been working on a Real time PBR renderer with a modified Blinn phong Linear shader system, but as is Blender is useless to me too create any maps. Save my life and have a look lad, your one of the shader boys on here that gives hope. If not ill have to give a try at porting my GLSL shaders through Blender internal ( waste of time in my eye’s). Also there’s no reason why such a shader can’t be run through cycles with the right GLSL tags within the editor (e.g each light you use also gets it’s own Open gl GLSL light settings).
It seems to get to this quality I need to subdivide to 7 which seems to make my machine less happy than using the micropoly displacement. Maybe just an illusion.
@3dluver I didn’t say anything about open subdiv. I said pixar like/looking. I haven’t seen much user art in cycles that is using the micropoly options so I thought I would post my findings. I’ll be very interested in the development of opensubdiv. Thanks.
Here is a test. Left is micropoly (dice .005) - right is subdivide modifier 7x (8 million faces) Right uses over 3 gigs mem and takes longer and my machine grinds down. Left looks better. How is it the same? Thanks Jay
Yeah, we all knew about it.
Try to displace using a UV displ map and see if you’ll still happy.
This is an experimental project and we have to wait.
I’m actually quite happy with Blender, just trying to understand better how it all works together. 
By “Pixar-looking” do you mean perfectly smooth surfaces? Because as Dingto explained it is the same old subdivision technique applied at rendertime.
Pixar’s subdivision method (Opensubdiv) implements GPU-powered subdivision calculation, better creasing and the ability to have the same subdiv results in the viewport as in the final render. Also it seems to open up for a good Ptex implementation.
I think what jaypjohnson is trying to point out before all the snarky and rude comments came flooding in is that it is not simply the same method of subdividing the mesh as the simple subsurface modifier. Yes, it’s the same algorithm but it’s not subdivided the same way as the modifier. The modifier statically subdivides the mesh in the scene. Those polygons stay in memory and are duplicated in the memory of Cycles. it always stays the same resolution no matter how close or far it is from the camera. What this is however, is render-time subdivision. Yes, it’s meant for displacement and it doesn’t actually work to well for that. But, as a way to dynamically subdivide your mesh using the built-in subdivision surface algorithm, it can actually be more memory efficient and give superior results to the modifier.
There is one fatal flaw as I have just found out: It somehow disables all sub materials other then the first. So, if you have a single character mesh with multiple materials, it renders the entire mesh with the first material. You could probebly still get around that by using maps to mask off the individual surface areas but it’s still a pretty serious limitation.
Now, does it compare with Pixar Subdivision surfaces algorithm? No, of course not. But did he say it did? No and I don’t think that was jaypjohnson point. Until we actually get OpenSubdive in Blender (which may be a ways off still) this can help out a lot in certain situations. I find it really upsetting to see this kind of reaction from the community. What is going on here is that someone is thinking outside the box and has found a good use for a feature we all thought was broken. Yet we are so bent on shaming others here that you all jump the gun and chastise him for even suggesting it. I would expect people to at least be respectful.
Send a PM. There’s no reason derail someone else’s thread.
Thanks Indy_logic. 
It just goes to show the difference between face to face conversations with all the social cues and etiquette vs. blasting off an email before reading what was written. I guess it’s all in how you title the post. If I would have asked, “Does anyone know the difference between subdivision modifier vs. dicing?” the thread may be better.
Microdisplacement in Cycles is stupid: because it smoothes the mesh the same way that the subdivision modifier does with catmull-clark algorythm instead of doing simple subdivision.
So if you want to had very fine details to your mesh with microdisplacement, you will obtain a detailed surface but with a too smoothed global shape, very far from your model in solid view.
The result of your displacement map is totally unpredictable.
LOL! Indeed. It blows my mind how people talk to each other on the internet. Whatever though, it goes with the territory I guess.
So zeauro, I guess you missed my earlier post. He’s not talking about using the dicing for displacement, he’s talking about the “Dynamic Subdivision” that results from just turning it on without even using displacement at all. This is similar to the dynamic subdivision feature of OpenSubdiv and it’s currently not a feature that people even know exists in Blender.
If you enable “True” Displacement and lower you dicing value, you can dynamically smooth your models at render time without even using a subsurf modifier. It saves a lot of memory and some render time in the process. But, like I posted above, it has some limitations.
All of this will be nullified once OpenSubDiv in Blender but it’s a nice little workaround int the mean time.
In practice, I see no advantage to use a subsurf modifier at a level of 7, except for displacement.
Smoothing is so important at this dicing rate that your mesh is ridiculously scaled down.
OK, it can looks like a funny thing. But there is no pratical case for that.