Jump to content
3DCoat Forums

V4.1 BETA (experimental 4.1.17D)


Recommended Posts

  • Reputable Contributor

I have to test further before making a full report, but I'm having a lot of trouble with the new Splines e-panel tool and making Pose selections. Particularly with the spline wrapping around an object (character). It seems if I create it, then switch to another e-panel tool and back, the curve gets messed up, the part that was on the "back", furthest from the camera, gets moved to the front. Some very unexpected results are also coming from using it with Pose selection as in the pic below.

2012-12-23_1927.png

I agree. I was trying to get familiar with the changes in order to demonstrate the changes, but I got some weird, inconsistent behavior. I even had problems with the closed spline tool...not closing.
Link to comment
Share on other sites

  • Contributor

Andrew, thank you for fixing the bug with repopping windows. :)

By the way Andrew (I know this thread has your attention), I just bought a new custom-built machine based on a 3930K processor. I expected a major gain in 3D Coat voxel modeling and overall performance, but the performance is very similar, if not identical, to my 5-year old Q9300 unit.

I loaded up a 15k tris scene and navigating through it simply chokes 3D Coat. Rotating the camera around the object results in no more than 10-15 fps, even though the fps meter in the lower left corner of the window states that it's 30. LC brush strokes in higher detail are as fast as they were on my Quad Core. What's funny is that the CPU load never exceeds 10%. It looks like sculpting in 3D Coat doesn't take advantage of multiple processor cores. Hell, even rendering doesn't use multiple cores!

True, I still use quite an old video card - 9800GTX/512MB to be precise, but I've browsed the forums here, and users claim that they didn't notice any difference between, say, 200 series cards (which are very near performance-wise to my card) and 400/500 series.

So what kind of a machine should a user have to run 3D Coat V4 smoothly when working on hi-res objects?

Merry Christmas everyone. :jester: (sorry, no Santa emoticon...) ;)

Link to comment
Share on other sites

  • Contributor

Andrew, thank you for fixing the bug with repopping windows. :)

By the way Andrew (I know this thread has your attention), I just bought a new custom-built machine based on a 3930K processor. I expected a major gain in 3D Coat voxel modeling and overall performance, but the performance is very similar, if not identical, to my 5-year old Q9300 unit.

I loaded up a 15k tris scene and navigating through it simply chokes 3D Coat. Rotating the camera around the object results in no more than 10-15 fps, even though the fps meter in the lower left corner of the window states that it's 30. LC brush strokes in higher detail are as fast as they were on my Quad Core. What's funny is that the CPU load never exceeds 10%. It looks like sculpting in 3D Coat doesn't take advantage of multiple processor cores. Hell, even rendering doesn't use multiple cores!

True, I still use quite an old video card - 9800GTX/512MB to be precise, but I've browsed the forums here, and users claim that they didn't notice any difference between, say, 200 series cards (which are very near performance-wise to my card) and 400/500 series.

So what kind of a machine should a user have to run 3D Coat V4 smoothly when working on hi-res objects?

Merry Christmas everyone. :jester: (sorry, no Santa emoticon...) ;)

I have a 6 years old cheap 2.3 quadcore with a gtx280 and only 6 gig ram and I work on 5 millions vox layers in 30-50 million polys scenes

using LC and Surface tools at blazing speeds when detailing,I use proxy mode to make broad radius changes when needed....

On a 15k scene I would get something like the speed of light or near it....

You really have trouble navigating around a 15k scene?! :blink:

Something is definitely wrong...

Sorry,that Im not able to help you. :(

Link to comment
Share on other sites

  • Contributor

Oh, sorry Artman, I misspelled it. It's 15 million tris scene.

Nevertheless the issue remains. With 30-50 million polys scene (when all are visible), at a current state, I have a feeling that 3D Coat would BSOD my system.

PS. It's actually a 18 million tris scene (I had some layers hidden). But this doesn't change a thing. Scene navigation as well as the speed of LC hi-res sculpting remains the same as it was in my Quad. What a bummer. :cray:

Link to comment
Share on other sites

  • Reputable Contributor

Oh, sorry Artman, I misspelled it. It's 15 million tris scene.

Nevertheless the issue remains. With 30-50 million polys scene (when all are visible), at a current state, I have a feeling that 3D Coat would BSOD my system.

PS. It's actually a 18 million tris scene (I had some layers hidden). But this doesn't change a thing. Scene navigation as well as the speed of LC hi-res sculpting remains the same as it was in my Quad. What a bummer. :cray:

Scene navigation is all on the graphic card. No wonder it's exactly the same, if your card is the same. Brushing is CPU multi-threaded, except in cases where CUDA is used. I recently asked Andrew and he said most Voxel brushes were CUDA accelerated. Try and get your hands on a GTX 670 2 or 4GB (VRAM), and you'll definitely see a difference in viewport navigation complexity handling. Should see some improvement in Voxel Brushes as the 9800 only has about 140 CUDA cores. The GTX 670 has over 1300! If Andrew will ever get around to recompiling 3D Coat's CUDA instructions (hopefully to CUDA 5, so it takes full benefit of some exciting new technologies that the Kepler cards introduced), there should be a big boost there alone. Would love to see CUDA expanded to brushes in the Paint Room.

If you would like to see a recompile for CUDA 5, make sure to add your +1 to the request, here.

http://3d-coat.com/mantis/view.php?id=772

  • Like 1
Link to comment
Share on other sites

  • Contributor

Thanks! This explains a lot! I planned to order a newer card right after the Christmas, but GTX670 4GB is far too expensive for my wallet right now, so I plan to get Gigabyte's GTX660Ti 3GB. With 192-bit memory bus it might not be as fast as its older brother, but I hope it will be enough to handle the scenes of relatively high complexity.

Speaking about multi-threaded brushing... I've checked this and it seems to only fully utilize one thread (even with CUDA disabled). The remaining 11 are almost idle (max. 20% usage on 2-4 out of 11 remaining threads).

By the way, any chance of OpenCL support in the future as an alternative to CUDA?

Link to comment
Share on other sites

  • Reputable Contributor

Raul said he was going to look into it at some point, but then he had to leave (he thought he had a 6 month visa, but turns out it was only 3). I mentioned Open ACC, which is a new standard that has been co-developed by a number of corporate giants, including NVIDIA and CRAY. The big benefit is that instead of a developer like Andrew having to dig deep down into the code and re-write a lot of it, Open ACC allows the developer to just add directives (hints) in their current code, and it will offload the work from the CPU to the GPU. The biggest benefit is the ease and speed of implementation. Hope Andrew can research it sometime. Nevertheless, CUDA is far more mature than OpenCL, and he already did a lot of coding work on it. Hate to see him have to scrap all that hard work.

Also, even though the card nomenclature may be the same, not all cards are. The ones with aftermarket style cooling tend to be both overclocked from the factory (extra performance for free), silent and stay much, much cooler...even under a heavy load. I never buy the reference card versions anymore. You can see the difference in the two different links. Notice that the price is often the same.

http://www.newegg.com/Product/Product.aspx?Item=N82E16814162119

http://www.newegg.com/Product/Product.aspx?Item=N82E16814130811

Link to comment
Share on other sites

  • Contributor

Yeah, I've heard about Raul. I hope he'll be joining us soon. :(

On the other hand I've never heard about Open AAC. What you said is very interesting though, so I guess I'll have to browse for some resources to read about this new standard.

But the thing that's good about OpenCL is that it's supported by cards from opposite camps - nVIDIA and ATI. And from what you said I have this bad feeling that it's just going to replace CUDA and be nVIDIA tech only.

Link to comment
Share on other sites

  • Reputable Contributor

It's GPU agnostic. It sounds like just the thing for a small developer like Andrew. It can even work with current CUDA code, if need be.

On a positive note, Raul did a fantastic job optimizing the LiveClay code before he left. It gave LiveClay performance a big kick in the pants. Really happy about that.

Link to comment
Share on other sites

  • Contributor

Thanks for linking the video. I'll check it out as soon as (now, this is embarassing) plug my speaker system to the new unit. I'll do this tomorrow, because it's 5 AM here right now, so I think I have to... let the excitement go for now and put my brain into sleep mode. :crazy:

Link to comment
Share on other sites

  • Reputable Contributor

Thanks for linking the video. I'll check it out as soon as (now, this is embarassing) plug my speaker system to the new unit. I'll do this tomorrow, because it's 5 AM here right now, so I think I have to... let the excitement go for now and put my brain into sleep mode. :crazy:

Sure. Here is some links to the OpenACC website and some videos on it.

http://www.caps-entreprise.com/products/caps-compilers/

http://www.openacc-standard.org/

http://www.openacc-standard.org/video

http://nvidia.fullvi...wed-derose.html

http://developer.dow...ch_overview.mp4

http://nvidia.fullvi.../203-wolfe.html

Link to comment
Share on other sites

  • Applink Developer

Should see some improvement in Voxel Brushes as the 9800 only has about 140 CUDA cores. The GTX 670 has over 1300! If Andrew will ever get around to recompiling 3D Coat's CUDA instructions (hopefully to CUDA 5, so it takes full benefit of some exciting new technologies that the Kepler cards introduced), there should be a big boost there alone. Would love to see CUDA expanded to brushes in the Paint Room.

AbnRanger I have seen you saing this before but you can't compare of how manu cuda cores there is in 9800 and 670. Nvidia changed cuda architecture into 660 so 1 cuda core in 9800 is not the same with 1 core in 670. One cuda core in 660 is not as powerful as it was older cards. What I have red about the issue is that 660 and 670 and mainly targeted into gaming. And they are very fast in those tasks. But for example 580 is better for 3d render engines that use cudas.

for me better cuda support is not the magic button to make 3d-coat better. There is many other directions that Andrew could take and it would benefit 3d-coat.

For example opencl because then all the ati users would benefit it too. Everytime when you saying that Andrew should make better cuda support starts to sound

a broken record. And Andrew has already answered about this so could we move on please.

Link to comment
Share on other sites

I've noticed something strange in the current version. I'd swear on drop down boxes previously you could click anywhere to open it. In the current one I must click the triangle on the right side of the box. Is this the same for everyone and has it always been this way?

Link to comment
Share on other sites

  • Reputable Contributor

AbnRanger I have seen you saing this before but you can't compare of how manu cuda cores there is in 9800 and 670. Nvidia changed cuda architecture into 660 so 1 cuda core in 9800 is not the same with 1 core in 670. One cuda core in 660 is not as powerful as it was older cards. What I have red about the issue is that 660 and 670 and mainly targeted into gaming. And they are very fast in those tasks. But for example 580 is better for 3d render engines that use cudas.

for me better cuda support is not the magic button to make 3d-coat better. There is many other directions that Andrew could take and it would benefit 3d-coat.

For example opencl because then all the ati users would benefit it too. Everytime when you saying that Andrew should make better cuda support starts to sound

a broken record. And Andrew has already answered about this so could we move on please.

CUDA is already a technology integrated in 3D Coat. It just hasn't been updated (recompiled) for newer versions of CUDA...ever. It's important to me, personally...so I don't have to blow a lot of money on hardware upgrades....especially if it's not going to help much. It's not an easy task, and that's why it continues to get pushed down to the bottom of Andrew's To-Do list. By the way, you make assumptions that a recompile wouldn't benefit, and you have absolutely no way to know that. Need I remind you again, that me and few others had to be a pain in the bum, for Andrew, in order to get Multi-Threading in 3D Coat (a few years ago). Folks scoffed and ridiculed me over it, then, the same way you are now. Andrew didn't see the major benefit in that either...until he actually did it. And for the same reason, he was just as reluctant to do it. You'd still be waiting several minutes to hours just to switch between Surface mode and Voxel, among other things....if we hadn't been so repetitive, even if it's annoyingly so.

If I think it's necessary to be "the squeaky wheel," I'll do so without your permission, thank you.

Link to comment
Share on other sites

  • Reputable Contributor

sigh.jpg

I think he is saying that upon re-opening the app, 3D Coat is resorting to the default layout...not recalling the last one used. I'm also noticing that when this occurs and I try to pop up windows and move them, if they are located over another panel, I can't move or select them. 3D Coat is selecting what ever is beneath them and under the cursor.
Link to comment
Share on other sites

  • Advanced Member

By the way Andrew (I know this thread has your attention), I just bought a new custom-built machine based on a 3930K processor. I expected a major gain in 3D Coat voxel modeling and overall performance, but the performance is very similar, if not identical, to my 5-year old Q9300 unit.

I experienced the same.

Link to comment
Share on other sites

  • Advanced Member

You know this....how?

I don't KNOW it, but it is my opinion based on my experiences where I have seen perhaps a 5% increase in performance for every $150 spent on upgrades, but I have to admit I upgraded 3 years ago. But since then I've had no need to upgrade because 3dCoat runs just fine basically all the time. It doesn't slow down under reasonable strain. I might have upgraded if I felt there had been a need. But all of my programs run great on 3 year old hardware. No need to upgrade. That's just my opinion.

Link to comment
Share on other sites

  • Applink Developer

AbnRanger, I remember with multi-threading that Andrew said that he will do that when the time is right. And he did that.

It feels that you are saying that becouse you remind Andrew so often that's why he coded it, but I don't see it that way.

I don't want to see that everyone starts to remind Andrew everyday about their feature request.That would turn this

forums into chaos. I think that when there is only one main coder we have to be patient with some features. Like I said there

is other ways to improve speed in 3d-coat, not just cuda or multi-threading.

Link to comment
Share on other sites

  • Reputable Contributor

AbnRanger, I remember with multi-threading that Andrew said that he will do that when the time is right. And he did that.

I'm sure that same thing with CUDA. he knows it's important feature. No need to tell that to him everyday. Like I said there

is other ways to improve speed in 3d-coat, not just cuda or multi-threading.

We all know that low-hanging fruit gets priority. Why? Cause it's easier/faster to implement. That means the most meaningful changes/features oftentimes get completely neglected...for this very reason. Hence the constant reminders. You keep saying "it's on the to-do list". I'm saying it's been at the bottom of that list for 2-3yrs now...and it has STAYED there. Things like Voxel Painting have jumped way ahead of it in line....even though the former was requested much earlier than the latter. So, please....You're are not talking to a newbie. I know the process here.
Link to comment
Share on other sites

  • Reputable Contributor

I noticed that Voxel Sculpting does not work with flat shading, anyone know if that will be addressed? Would be so awesome! I tried to emulate with materials and light settings, but that didnt work out.

Thanks,

Jan

I don't see where that would be handy, personally. The aim in sculpting is to see the 3D sculpting changes. With no shading, it seems rather counterproductive. However, flat shading makes sense when texturing, so you can see just the textures only. If you can convince Andrew that it's worthwhile, then by all means, give it a try (support@3d-coat.com)
Link to comment
Share on other sites

  • Reputable Contributor

An old bug has returned: You cannot move floating windows when they are over a docked window.

http://www.screencast.com/t/sDRJgjsx

Password: 3dcbugs

Just added it to Mantis. Anyone else noticing this, please confirm (so Andrew knows it's not isolated to one user), here.

http://3d-coat.com/m...view.php?id=824

Also, I'm noticing that I have to reload my saved workspace every time I open 3D Coat. That's because it continues to resort back to the default. If anyone else notices this, please confirm here on Mantis:

http://3d-coat.com/mantis/view.php?id=825

Link to comment
Share on other sites

Guest
This topic is now closed to further replies.
 Share

×
×
  • Create New...