i saw a video on this and seems you can get very high poly count and still have a fast render time but also a fast response form blender which would not be the case if using subsurf
but what would be the maximum poly count and still have a good response time from blender using theses new tools ?
Example i tried to do the top of a corinthian coloumn i think
with acanthus leaves and i ended up with around 50 000 vertices and blender was slow
if i re do it with multires and sculp would this be a lot faster
and can you have more then one model in scene
not certain here if i can keep the symetry on the model around the column!
i mean if i want to do several complicate columns design for a building would it gives a fast render times even if i get around 1000000 vertices for instance ?
I was getting pretty smooth(as in useable) sculpting with 6.3 million faces, on an Ati Hd3650 mobility. It crashed though when I tried a long stroke, guess it ran out of memory or something.
Thanks to the ‘fast navigate’ option navigation/rotating was quick too.
I just tested with a heavily subdivided cube (edit mode) then 4 levels of multires if i can remember correctly.
The sculpting was actually useable but the gui (value sliders and so forth) was not reacting quick enough:no:.
I can’t get more than 1.5 million faces before it goes so slow that it’s hard to work. Another subdivision crashes blender. But then again, I got vista >_>
are you using a heavy mesh or a very light mesh…like the default cube. Someone stated that it is better to use a heavier mesh, and after I tested I found it to be true…which is better for real world situations anyway.
When i open up this thread i should have mentionned that i was refering back to the 2.5 video done on sculp and multires
which seem to demonstrate that under 2.5 this way of using multires with sculp was supposed to be a lot faster then doing it with subsurf and sculp!
and that it would allow to work with much higher vertices count for modelling and still have a good ersponse time from blender
now this is obviously for high rest modeling and not very low poly models!
I’ve just been sculpting about 5 million polygons in blender 2.5, running on windows XP 32-bit.
I know for a fact on a 64-bit OS such as Linux or Windows 7 I could go much higher than that.
Some tips to speed it up while sculpting are, lower the sculpt resolution for preview viewing, I usually set this quite low so when you view the object outside of sculpting mode it shows no sub-division. Enable optimal display, this increased my frame rates dramatically, even in solid drawing mode. Enable fast view rotation, or what ever it is called and also if your graphic card supports it and works well with it enable VBO’s.
Personally I leave VBO turned off as it has some serious glitches with my GPU (Nvidia 260GTX)
The glitches when VBO are turned on are mostly in edit mode; Selecting edges doesn’t work, often selecting edges or faces fails to register (Doesn’t do anything) or randomly selects a face or edge on the mesh. Vertex selection works like normal though.
There are other glitches but I haven’t used VBO’s in a while so I’ve forgotten exactly what they are.
The options for the preview resolution and optimal draw are selectable in the multi-resolution modifier panel.
The quick/ smooth (?) rotation option is available in the Toolbox panel (T-key).
It seems to be working with the latest graphicall build, I guess it was the older revisions that had the issues, I wasn’t aware of any bug fixing regarding the issue.
I have a dual core 2.4 GHz, nvidia 9600 GSO 384 MB and 6 gigs of ram. Running vista 64.
Multiresing the default cube isn’t optimal, that I know. Then having a modeled head should be better, right? Because it’s still slow. Actually I haven’t had a model that that I can go high polycount like in those videos and comments. Like 6 million polys is 10 times more then what I can get before it lags like there’s no tomorrow.
Here’s a rough guide to the effect your base mesh has on the speed of sculpting:
When you sculpt, it optimizes by ignoring parts of the mesh that aren’t near the brush. There are two resource-intensive things that sculpting does: the actual brush calculations needed to modify the surface, and redrawing the areas that have been changed. For this optimization to work, it needs to break the mesh up into pieces. That way, just one small piece can be sculpted or redrawn, rather than millions of polygons in a detailed mesh. So the question is how is mesh split up?
When you multires a mesh, it gets split up into grids. You know how when you apply one level of subsurf to a model, each quad turns into four smaller quads, and each triangle turns into three smaller quads? Each of those quads counts as one grid in multires. So if your input mesh is 6 quads, like the default cube, you’ll have 24 grids, whereas a monkey (500 faces), produces 1968 grids.
One grid is the smallest chunk of the mesh that can be sculpted on and redrawn independently. So no matter how far you subdivide, you have the same number of grids, and you have the same limit on the number of chunks the mesh can be split into for optimization.
Note that only the topology of the base mesh matters, not the shape.
As I said, this is a rough guide – it doesn’t contain any numbers! So you still have to experiment a bit to see what level of detail in the base mesh works best on your system.
I have one more question.
What decides the speed (framerate) when rotating around an object? Does that have anything to do with the sculpt mode or is it all about good hardware?
Well, you can turn on fast navigate, then it’ll drop down to a lower multires level when rotating. Other than that, it’s mostly about hardware (and of course you should have VBO enabled in the user prefs.)