Advanced Member JamesE Posted January 10, 2010 Advanced Member Report Share Posted January 10, 2010 Does surface mode use the GPU for mesh rendering or is it more like Zbrush? Just curious, as I was thinking that it would probably be much better for detail sculpting if it was GPU independent the way zbrush is. Not sure what kind of issues that would create in switching back and forth from Voxels to Surface mode though. Just kinda wondering how it all works I guess - or will work. I'm also wondering about the proposed multi-resolution capability for Surface mode. How will this work when switching back and forth between voxels and surface? Or is it a one way thing? Do you lose your subdivision history if you switch from surface to voxels and back? Quote Link to comment Share on other sites More sharing options...
Reputable Contributor AbnRanger Posted January 10, 2010 Reputable Contributor Report Share Posted January 10, 2010 Yeah, I wonder exactly what operations involve CUDA...brush speed? Calculations where you see the progress meter? Converting to surface mode? The reason I ask is that it helps the user to know what sort of hardware upgrades to make. Right now the biggest slowdowns are the calculations; so if 3DC uses strictly the CPU than that would naturally be where a user may want to spend a bit extra, to get a higher performing CPU. Quote Link to comment Share on other sites More sharing options...
Advanced Member splodge Posted January 10, 2010 Advanced Member Report Share Posted January 10, 2010 Does surface mode use the GPU for mesh rendering or is it more like Zbrush? Just curious, as I was thinking that it would probably be much better for detail sculpting if it was GPU independent the way zbrush is. Not sure what kind of issues that would create in switching back and forth from Voxels to Surface mode though. Just kinda wondering how it all works I guess - or will work. The GPU is used in both modes. From a rendering point of view there is no difference between surface mode and voxel mode. The only difference is that in voxel mode 3D Coat stores a voxel representation of the model alongside the mesh you see on screen. When you make a change to the underlying voxels then the corresponding part of the mesh is updated. When working with multi resolution you'll probably be able to switch back to voxels but it will depend on how much memory you have as to how much detail you'll lose. Voxels take up much more memory than just the raw mesh you have in surface mode, so obviously if in surface mode you have a billion poly mesh and you decide to go back to voxel mode then it's unlikely that there's gonna be enough memory to store voxels of such a high density. So either you accept a huge loss in detail or you go buy a big memory upgrade. I'm also wondering about the proposed multi-resolution capability for Surface mode. How will this work when switching back and forth between voxels and surface? Or is it a one way thing? Do you lose your subdivision history if you switch from surface to voxels and back? When you go back to voxel mode then you'll lose your subdivision history. Although I suppose it's technically possible for 3D Coat to temporarily dump the surface mode mesh to hard drive when going to voxel mode. Then when you eventually switch back to surface mode 3D Coat could load the dumped mesh back in and project it back onto the new surface. But this could only work for very small changes, so I'm not sure it'd be worth the effort because if you're only gonna make small changes then you may as well stay in surface mode and make those small changes. And in future the whole point of surface mode won't be just the brush speed increase and more room in memory for a detailed mesh. It'll also mean 3D Coat will be able to drop down to a lower subdivision when panning the camera. Whereas on today's machines even if you had 100 gig of memory that could hold a billion poly voxel model then the GPU just wouldn't be able to render the whole mesh fast enough, so panning/rotating the camera would be painful.. Quote Link to comment Share on other sites More sharing options...
Advanced Member splodge Posted January 10, 2010 Advanced Member Report Share Posted January 10, 2010 Yeah, I wonder exactly what operations involve CUDA...brush speed? Calculations where you see the progress meter? Converting to surface mode? The reason I ask is that it helps the user to know what sort of hardware upgrades to make. Right now the biggest slowdowns are the calculations; so if 3DC uses strictly the CPU than that would naturally be where a user may want to spend a bit extra, to get a higher performing CPU. 3D Coat only uses cuda to increase brush speed on some of the more demanding voxel brushes. There's a really informative section in the manual about it. I reckon a memory upgrade may speed up those parts where you see the progress meter. On my machine I have a feeling 3D Coat is often relying on virtual memory when exporting/quadrangulating/baking high detail models. I often get lots of agonizing white screens and "3D Coat is not responding." messages. And I got me 4 gig of ram!. Sometimes it gets to 90 percent and then just hangs for ages. That's where it's probably doing the virtual memory shuffle. Quote Link to comment Share on other sites More sharing options...
Advanced Member JamesE Posted January 11, 2010 Author Advanced Member Report Share Posted January 11, 2010 When you go back to voxel mode then you'll lose your subdivision history. Although I suppose it's technically possible for 3D Coat to temporarily dump the surface mode mesh to hard drive when going to voxel mode. Then when you eventually switch back to surface mode 3D Coat could load the dumped mesh back in and project it back onto the new surface. But this could only work for very small changes, so I'm not sure it'd be worth the effort because if you're only gonna make small changes then you may as well stay in surface mode and make those small changes. And in future the whole point of surface mode won't be just the brush speed increase and more room in memory for a detailed mesh. It'll also mean 3D Coat will be able to drop down to a lower subdivision when panning the camera. Whereas on today's machines even if you had 100 gig of memory that could hold a billion poly voxel model then the GPU just wouldn't be able to render the whole mesh fast enough, so panning/rotating the camera would be painful.. Yeah I suspected as much when I read about that... it's not really a big deal to me as I've grown quite accustomed to working without subdivision levels as it is in 3dc. By the time I get to a detailing phase I'm pretty much beyond the need to go back and forth, so I'm fine with finishing off in surface mode. What I do like about it even now is the ability to use both modes as I see fit. Concepting and roughing out forms in voxels, refining them a bit in surface mode, and then jumping back to voxels if I want to make a drastic change - it's all fairly seamless. The topology independence is still the most important part for me when sculpting in this app. Enhancing viewport interaction will be a nice bonus too. Anything that brings performance closer to zbrush when detailing is good. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.