Speed of work vs ZB & MB

[Andrew Shpagin]

I have heared several times that ZB or MB are faster in work then 3DC. Is it really so? If you will work with mesh 8M does ZB and MB work faster? They need to show all 8M of polygones and it could not be done realtime. What operations are actually faster?

Also what about mesh size on disk? I think that for ZB & MB mesh size on disk should be really huge (on 8M).

[BluEgo]

I have heared several times that ZB or MB are faster in work then 3DC. Is it really so? If you will work with mesh 8M does ZB and MB work faster? They need to show all 8M of polygones and it could not be done realtime. What operations are actually faster?

Also what about mesh size on disk? I think that for ZB & MB mesh size on disk should be really huge (on 8M).

ZBrush is absolutly fantastic when handling large polycount. All brush strokes are so fluid in a way that is difficult believing that this stuff is real time.

Anyway I can sculpt perfectly and in a smooth way with 8M meshes, but I can subdivide without problem my mesh until 15 M even if here I can notice little slowdown.

A clever method ZBrush uses is to display a lower level of subdivision when panning, rotating and zooming.

The limit in which ZBrush starts to display a lower resolution while performing these operations is set to 100 K as default.

A basic cube subdivided until 6M of vertices is about 50 Mb of size on disk

My PC here at work where I'm making these tests is a 8-core with 4 GB of ram, videoboard NVidia quadro FX 5600

And if you want, you can download the ZBrush 3 trial from the pixologic website

[SonK]

I have heared several times that ZB or MB are faster in work then 3DC. Is it really so? If you will work with mesh 8M does ZB and MB work faster? They need to show all 8M of polygones and it could not be done realtime. What operations are actually faster?

Also what about mesh size on disk? I think that for ZB & MB mesh size on disk should be really huge (on 8M).

8 million poly mesh in Mudbox would probably be close to 1 GB in file size. I can't really sculpt on 8 million poly mesh on my old GeForce 7600 GS, it;s just not realtime(30 frames per second). 1-2 million is doable though.

[UnCommonGrafx]

Yes, there is a smoothness in ZB that is not in 3dCoat.

ZB can work with virtually anything I've thrown at it and it keeps it fun by being so fluid. All at the same time you are doing some real cool sculpturing and painting.

I like 3dCoat for the use it offers me with LW. A UV has to be a bit more perfect to not crash zb while 3dCoat outputs everything. It still amazes me that zb can't get trap code in there for that...

[SonK]

alittle Mudbox 2.0 "news":

"Zargarpour gave a demo of an upcoming release of Autodesk’s Mudbox, still in the Alpha stage. No release date was offered, but the software has some compelling features. A 64 bit machine can handle 100 million polys, and he said using the program was as fast as a pencil and paper and as easy as sculpting butter."

Doesn't seems like they are using voxel/volumetric sculpting, so 3DC 3.0 still has a chance.

[rimasson]

I'm in love with 3DCoat because of some of it's tools which are UNIQUE.

I'm a Zbrush user, Mud dosen't add almost anything new to. 3Dcoat does.

[wailingmonkey]

I'm in love with 3DCoat because of some of it's tools which are UNIQUE.

I'm a Zbrush user, Mud dosen't add almost anything new to. 3Dcoat does.

I agree...as far as I'm concerned, 3dC will ALWAYS have a chance in my

pipeline since it's combined toolset makes certain challenges easier/more efficient

to overcome than Mudbox or Zbrush. It's pricetag makes it nearly unquestionable

as an asset of serious consideration. Throw in the fact that Andrew is apparently

tireless in his desire to improve/enhance it (and has an update schedule unmatched

by any other vendor I've encountered), and you have quite a unique combination

of strengths going in 3dC's favor that other software devs are either unwilling or

unable to match.

I'll of course be curious what the next Mudbox (or Zbrush, for that matter) iteration

improves upon, but I can't imagine either of them will ever make me *stop* using

3dC or render it obsolete in my workflow.

[rimasson]

I found some info about the upcoming Mudbox 2

http://boards.polycount.net/showthread.php?p=818895

- Support for realtime shaders in the viewport (cubemaps, IBL, HDR support, and I suspect the ability to load a certain kind of .fx maybe)

-Ability to rotate the relflected environment and ability to lock it at the desired angle

- Realtime ambiant occlusion

- Realtime depth of field

- texture stamping through strokes

- color painting (forgot to ask if the brush engine was anything close to photoshops)

- automatic little squares unwrapping allowing to paint on an object that has not been unwrapped by hand

- ability to paint bumb, spec and diffuse maps (dont know if you can setup a material with a specific bump, spec and diffuse at the same time...)

- SteadyStroke (=lazymouse)

- More polies on screen

- x64

- Rewritten behaviour of some brushes (I think)

- New splash logo (that was horrible)

- New startup screen for quick access to basic scenes setup

[Rodney Brett]

Andrew, if you are interested in "voxel" based sculpt program for future R@D, I suggest looking into SenSABLE Claytools/Freeform modeling. I got a chance to test drive this program/device for a few days and it was really amazing. As far as I know, it was the only software of it's kind.

As far as ZBRUSH is concerned, the beauty of the app is the speed. I don't know how they do it, but I'm often moving "chunks" of polys at 8million with no lag. It's the perfect application for pre-visualization. You can pretty much "sketch" model within the program. Here's one of my samples of doing just this:

With the current 1.07 release of MUDBOX, I cannot do this. Strictly because of two factors.. Realtime drastically drops once the mesh is past 2 million on my machine, even though I've got 8GB of Ram(due to MB 32bit limit). The other factor is that jumping up and down subdivisions can take a long time, sometimes up to a minute. Zbrush can jump up/down subdivision levels at lightning speed.

I see 3d-Coat as more of a "detailing" application. It's unbelievable for painting and adding small, intricate details without headaches. It's competing with applications like BODYPAINT more so than Zbrush/Mud. The simple fact that you can preserve your layer structure while projection painting is already enough to trump Zbrush's workflow right now for texturing.

-Rod

[Nemoid_]

I've thought a bit to how ZB operates IMO(no expert tho)

ZB partially substitutes open GL/DX with its own visualization engine, made through pixols. That's why its CPU /memory dependant but not so much graphic card dependant.

i see many apps video card dependants that have probs because of drivers , and new features added to them, but ZB has not that many problems in its case. only small quirks in 3.0 version. it doesn't generates normal or disp maps on the fly, i guess.

it just visualizes polygons thorugh the pixol environment so that it can be very rapid and powerful in sculpting response.

the app is also very optimized to work also oin common machines, and becomes really rapid when you have enough of ram and a good cpu.

could be wrong here, but apps like Modo , Silo and for now 3D coat are less powerful and smooth in sculpting for a similar reason.

but, the workflow is also what counts.

ZB is wonderful for creating characters and models from scratch with z spheres, sculpt and detail em to very fine levels, but its not the same in texturing/painting. no layers, depends on poly resolution, no classic uv unwrap and also doesn't accept overlapping uvs. its not that u can't texture with it, btw u can do wonders and has z app link for what's missing, but workflow is convoluted and just weird in this area for now.

3.1 retopo tools, as for now, work, but aren't on the same level of a Nex or of what 3d coat is going to have soon.

[awnold]

Yup, I have said it before but what 3dc adds to a workflow, and it unique (from what I know) is painting on mesh. From what I know, bodypaint is a great app but I think you paint almost as a projection, and its applied in real time to the mesh, its also 900 bucks. ZB, because of what is discussed above is mad powerful with sculpting but its painting is limited by polygon count, meaning your mesh better be really dense to get details (from what I have seen). Or you can paint with its projection feature but you need to 'fix' the model in place before starting to paint. This 'fixing in place' feature is really unintuitive compared to BP or 3dc. Plus 3dc's connection to PS is a big plus (like also stated by others above)

-my 2 cents

[lc8b105]

The speed of work and The speed of import a model

It's hard to say the difference of the speed of work between 3D-Coat and ZBrush,But the speed of import an object seems slow in

3D-coat,especially when the faces of the model is a little more.

This is an scaned model from DAVID.

In ZB,only 2 seconds to import this model,almost just click the button then it imported.

But In 3DC, 2 hours and 50 minutes only load 99%,then the systerm seems to be freezed.

I know this model is really huge for 3DC,but it can't be imported through the "import huge mesh",because it's a scaned object,

we can't provide two meshes(one huge mesh and one base mesh).

The reason we want to import a scaned model to 3DC is the new retopology tool.The new retopology tool is wonderful.

And retopology is very important for the scaned model work flow.

But the problem is : How do we import the scaned model to 3DC first ?

Hope one day 3DC will also join in the scaned object work flow(Retopology).

DAVID_Angel.rar

[Andrew Shpagin]

Try to import model with keep-uv. I plan to add possibility to import reference mesh to be used in retopo tool.

Your mode was loaded sooo slow because of auto-uv mapping (I think).

[lc8b105]

Try to import model with keep-uv. I plan to add possibility to import reference mesh to be used in retopo tool.

Your mode was loaded sooo slow because of auto-uv mapping (I think).

Yes,I just left the default setting，it's auto-uv mapping . I just made a test to select keep-uv,this time only 57 minutes load to 99%.

"I plan to add possibility to import reference mesh to be used in retopo tool." That's great! :lol:

[philnolan3d]

Just out of curiosity I downloaded that model. It took about 30 seconds or so for it to get up to 99% using Keep UV. After which it appeared to be frozen. Same results with Keep Clusters except that after a moment I could see a black silhouette of the model And could rotate it around just fine, as can be seen in the attachment.

I decided to try what I had done with the Make Human model and loaded the angel cherub into Accutrans, then saved it out as an LWO. Upon loading that into 3DC I had the same result as the last test. So I loaded it into LW and looked at it. It doesn't have a UV map on it, so that's why it was having trouble loading with the Keep options. Lastly I went back and tried the original model with auto-mapping. 3DC just froze up immediatly.

So I'm sure the reason why ZB and LightWave have no problemn is that they don't need the UV map to display the model. I hope this helps.

[philnolan3d]

Here's a new demo of the upcoming MudBox's performance

http://area.autodesk.com/mudbox_preview#performance_preview

[Rodney Brett]

Here's a new demo of the upcoming MudBox's performance

http://area.autodesk.com/mudbox_preview#performance_preview

Yeah, that was an impressive demo. It seems that Mudbox 2009 is going all OpenGL and will probably require the same system stats as Maya. This will be good for Maya users and for my own work, since I'm running QuadroGL cards, however Dave Carwell also mentioned in the interview that it would run on an 8800 GTS. I still really appreciate the fact that Andrew added OpenGL version of 3d-coat. There are lots of 3d users that don't have gaming cards and one of the biggest issues right now is the split between high-end gaming GPUs and high-end workstation GPUs. It seems like Nvidia and Ati are both in a little conspiracy to "cripple" gaming cards performance in Maya. It's been a topic of grief here at my work.

[philnolan3d]

Some additional Mudbox 2009 info I just picked up:

Texturing is coming.

Press release:

pressreleases.autodesk.com/index.php?s=press_releases&item=447</td>

Features list:

usa.autodesk.com/adsk/servlet/index?siteID=123112&id=10708002

Honestly in my current project I used MB for sculpting, 3DC for retopo, and I plan to use 3DC for painting.

[rimasson]

I checked the Mud 2009 release :

I have a 'clay brush' like Zbrush, but No psd layered textures support.

[Improv]

Here's a new demo of the upcoming MudBox's performance

http://area.autodesk.com/mudbox_preview#performance_preview

And here's the latest Mudbox 2 video (August 11):

http://area.autodesk.com/mudbox_preview#siggraph

[Rodney Brett]

And here's the latest Mudbox 2 video (August 11):

http://area.autodesk.com/mudbox_preview#siggraph

Still no word on a retopology tool which is very disappointing. Don't get me wrong, what they've shown is pretty amazing, it's just that to get it rigged and animated in Max/Maya/etc., you still need to depend on some middleware solution for resurfacing. I was also sad to see that surface topology tools are not in Maya 2009's release. I was betting in it being there for sure.

[Andrew Shpagin]

At least I have understood technology used in MB 2009

It seems to be based on multiple render targets technique.

1) render depth in one screen-size render target

2) render color in second RT

3) maps (1) and (2) will be updated only on changed areas, no full rendering performed

Realtime operations:

5) render shadow map using mesh of lower resolution (to speed up)

4) generate normals on the fly in pixel shader using depth of nearby pixels - from rendered in (1) image, apply shadow map, blur using depth map (DOF), apply lighting.

I don't know if they do in exactly this way but this way can give good performance even without cuda etc.

[Rodney Brett]

At least I have understood technology used in MB 2009

It seems to be based on multiple render targets technique.

1) render depth in one screen-size render target

2) render color in second RT

3) maps (1) and (2) will be updated only on changed areas, no full rendering performed

Realtime operations:

5) render shadow map using mesh of lower resolution (to speed up)

4) generate normals on the fly in pixel shader using depth of nearby pixels - from rendered in (1) image, apply shadow map, blur using depth map (DOF), apply lighting.

I don't know if they do in exactly this way but this way can give good performance even without cuda etc.

Yes, I believe this is the reason why Dave Cardwell stated that Mudbox 2009 will use the GPU for certain tasks. Performance for Mudbox 2009 will be scalable, so future hardware/CPU/Ram upgrades will have more of an impact on the program's performance. The realtime render looks almost like a high quality game engine render. It's very fast.

Currently, Zbrush tech is heavily dependant on it's "Pixol" technology. The only GPU usage in Zbrush is with Zmapper for generation of Normal Maps, but most of the program is software based. This can be an advantage or a disadvantage depending on how you look at it and what kind of hardware you are running it on. I don't know if Pixologic plans on writing Zbrush 4 from scratch or if they will continue to go in the Pixol 2.5d direction.

[philnolan3d]

I don't know if Pixologic plans on writing Zbrush 4 from scratch or if they will continue to go in the Pixol 2.5d direction.

I hope they go from scratch and redo the interface while they're at it.

[Rodney Brett]

I hope they go from scratch and redo the interface while they're at it.

Ha! They are pretty stubborn with that interface. I just got back from teaching Zbrush to new students and explaining "Zoom" to them always makes them scratch their heads.. I've been using Zbrush since 1.5b, so I got used to it, but I often forget how weird it can be.

[Andrew Shpagin]

Also, I see they are using screen space ambient occlusion technology. You can read what is it there

http://rgba.scenesp.org/iq/computer/articles/ssao/ssao.htm

All that features like SSAO, DOF, HDR are not complex things to do... Just need a decission where to put time (1-2 weeks) - in visual appearance in viewport or to edit capabilities.

[SonK]

Also, I see they are using screen space ambient occlusion technology. You can read what is it there

http://rgba.scenesp.org/iq/computer/articles/ssao/ssao.htm

All that features like SSAO, DOF, HDR are not complex things to do... Just need a decission where to put time (1-2 weeks) - in visual appearance in viewport or to edit capabilities.

I vote visual appearance, make it look as good as Mudbox 2009 viewport :lol:

[rimasson]

I vote too. ,maybe after voxel integration.

I may help to better visualize a model, but also to attract new users.

Th realtime demo iwith Ao is simply amazing.

[Andrew Shpagin]

I vote visual appearance, make it look as good as Mudbox 2009 viewport :lol:

Maybe it is like a question - what is better for car - how it looks or how it drives or how much fuel it takes

[SonK]

Maybe it is like a question - what is better for car - how it looks or how it drives or how much fuel it takes

lol. okie, maybe if you explain what you meant by "edit capabilities", i'll make a better vote. I guess either way i vote, this will be added to 2.10? If that's the case, i vote both lol. I don't mind waiting abit longer if those features are added. Having that realistic viewport is important for sculpting/painting, its not just eye candy.

Also Mudbox 2009 allows the user to use their own custom cgFX shader.