Jump to content
3DCoat Forums

Performance Tips


CyBorge
 Share

Recommended Posts

  • Member

So, as many others have stated in the forum - I love 3D-Coat but, I'm often perplexed/confounded by performance issues.

 

I've spent time reading about VRAM, about Page Files/Virtual Mem, System Mem, CPU specs, CUDA Performance, Lack of OpenCL support, extra channels per layer etc. 

 

Thus, as I've been reading these various threads it occurs to me:

 

I can't seem to find a consolidated and authoritative resource on the subject. (Yes, I will read the manual but, it's not as COOL as a video.)

 

It ruins my experience with the program when I'm forced to troubleshoot performance.

It breaks my creative excitement and I imagine it does the same to others.

 

In fact, Ignorance ( my own and others ) of 3D-Coat's performance nuances may be a main factor in keeping it from being more widely adopted. 

 

Moreover, while watching the fantastic official tutorial/feature videos on YouTube ( Note: A big "THANK YOU!" to the 3D-Coat team for all the videos they are a great feature. ), I've noticed that he's often operating at a much higher voxel resolution than I can achieve without experiencing issues etc.

 

I do have a decent system so I'd expect to be able to at least follow along with tutorials etc. and I imagine others may be affected by this as well. 

 

 

Summing up - Given all the questions about what Video Card to use etc. maybe some experienced Guru or Gurus could compile a useful resource.

 

I'd be willing to pay for such a resource. 

 

Something like "3D-Coat Performance Tips - Hardware and Concepts".

 

Now, if this exists - please by all means post your "Look here dummy..." links below because, these questions keep coming up on the forum.

 

Or if there's a sticky "Performance Tips&Tricks Thread" please list that with an appropriately pithy insult for my laziness at finding it. :D

 

If not, maybe someone is due to make some sweet cash by sharing their expertise.

 

 

 

 

 

 

 

 

 

Link to comment
Share on other sites

  • Advanced Member

What's exactly is your current system?

What sort of issues exactly do you experience?

 

Something I only discovered after a while was the CUDA smooth boost which helps performance with the brushes a great deal I've found, if it's an option for you.

post-38412-0-02653800-1428527448_thumb.p

Edited by arumiat
  • Like 1
Link to comment
Share on other sites

  • Member

Hey arumiat - thanks for that tip! 

 

My main system specs are:

 

Intel Core i7-3770K  32GB RAM EVGA GeForce GTX480 with 1536MB VRAM and I run off a Crucial M550 SSD.

 

I know that my video card is long in the tooth and light on VRAM so I plan to upgrade that soon.

 

I typically work on a 24" monitor at 1920x1200 resolution.

 

Typically, I'll experience slow-downs or freezes when upping my voxel res to get smooth results from various operations. 

I'll have to get you some exact levels/numbers where I start to feel the issues.

Edited by CyBorge
Link to comment
Share on other sites

  • Contributor

First thing that came on my mind reading your post was drivers , are they up to date?

 

Second is Ram both system and GPU but i see you have plenty of it.

 

On my system with quad xeons 32 gb Ram and GTX 470 (use it as a second , as 780 is kept for rendering) i got no issues at all.

Edited by Michaelgdrs
Link to comment
Share on other sites

  • Member

Thank Michaelgdrs,

 

I'm glad to hear your system with a 470 is still a viable solution.

 

I do keep my drivers updated etc. - Currently running version 347.52 of the Nvidia drivers.

 

I'm sure most of my issues are due to my ignorance of exactly how 3D-Coat works underneath or my failure to follow best practices when adding resolution and layers etc.

 

However, I do think a lot of newcomers to 3D-Coat basically have the same questions about performance, hardware, and the aforementioned best practices and hence the call for a resource on the matter.

 

Out of curiosity, how much VRAM on your GTX470 and which version of 3D-Coat do you run on that system?  (Cuda+DX64 or OpenGL+Cuda - I've heard arguments for both)

 

I seem to have better performance with the OpenGL version, like the documentation suggests, and I'm running the latest stable 4.1.17D(Cuda)(GL64)  - but I was trying the DX64 version when some issues occurred.

Edited by CyBorge
  • Like 1
Link to comment
Share on other sites

  • Advanced Member

I'm pretty sure that your GPU's vram makes a difference with using voxels.   I'm using a 460 and it works pretty well, I can go up to couple of million polys and not have much lag.   My 460 has 2gb of vram though, maybe that is the difference?  I use the OpenGL version, but on Linux.  I don't know if that would have any effect though?  

  • Like 2
Link to comment
Share on other sites

  • Member

He's a GRIMM!!! :D

 

Yes, VRAM is one of the most discussed aspects of performance and "On screen polycount" here in the forums.

 

I think Michaelgdrs is right that my GTX480 is still valid for at least most projects but, that's not going to stop me from upgrading! ;)

 

I've been looking for a good excuse to get a 4K capable card and Monitor. - ( I want to spend late nights basking in the glow of a 40" 4K screen! )

I'm just waiting for the rumors of 8GB GTX980's and 970's to become a reality!

Edited by CyBorge
  • Like 1
Link to comment
Share on other sites

  • Member

If i am not mistaken , i remember reading  , that fermi gpus works better.

 

Can someone verify please?

 

That does seem to be true at the moment. The current versions of 3D-Coat aren't compiled for the newer versions of CUDA that the Maxwell chips would benefit from.

Plus NVIDIA basically crippled Cuda performance in the newer mainstream cards - choosing to focus on gaming performance instead.

 

So an older FERMI GTX580 can outperform most newer cards when it comes to Cuda. 

Titan cards are an exception - but at a hefty price.

The other key exceptions here are the GTX 780s (Which you seem to have) - These cards have remained expensive to buy even though they're replaced by the newer models.

The main reasons for their continued high price is that they have higher overall Cuda Performance and a full 384 bit Memory bus for their VRAM - (compared to 9XXs  - their on par with a 580 more or less)

 

Look here: You'll see the 580 hanging in there: http://www.tomshardware.com/reviews/geforce-gtx-780-performance-review,3516-26.html

 

So if anyone can find the 6GB version of the GTX780 it should be a very good performer with 3D-Coat and offer a nice amount of VRAM, but, it will cost.

Another caveat: 7XX series are pretty power hungry compared to newer cards so you'll want a Power Supply that's ready.

 

As for OpenCL and AMD Cards - It doesn't seem to be an option just yet. - 

However, given that some programs are starting to support OpenCL for rendering, I believe it's only a matter of time.

Edited by CyBorge
  • Like 1
Link to comment
Share on other sites

  • Advanced Member

It is going to depend on the application.  I have a 980 in addition to my 460 that I just use for rendering in Octane.  The Octane devs have optimized for Maxwell to the point where my 980 beats a 780 and the Titan speed wise.  Although the 780ti, Titan Black, and Titan X are faster.  The Maxwell architecture is different enough to warrant CUDA re-coding.  I don't run the CUDA version of 3D-Coat because it requires version 3, Octane requires version 6.5+, so I am unable to run it as the CUDA version I have installed is too new.  So my speed with 3D-Coat is completely reliant on the speed of OpenGL that the 460 can handle.  

Link to comment
Share on other sites

  • Contributor

It is going to depend on the application.  I have a 980 in addition to my 460 that I just use for rendering in Octane.  The Octane devs have optimized for Maxwell to the point where my 980 beats a 780 and the Titan speed wise.  Although the 780ti, Titan Black, and Titan X are faster.  The Maxwell architecture is different enough to warrant CUDA re-coding.  I don't run the CUDA version of 3D-Coat because it requires version 3, Octane requires version 6.5+, so I am unable to run it as the CUDA version I have installed is too new.  So my speed with 3D-Coat is completely reliant on the speed of OpenGL that the 460 can handle.  

Fortunately, I'm too ignernt and run both 3DC and Octane with the same 560TI in CUDA.

Link to comment
Share on other sites

  • Advanced Member

Fortunately, I'm too ignernt and run both 3DC and Octane with the same 560TI in CUDA.

 

It all comes down to your card's compute capability version.  Your 560 is the same as my 460, compute version 2.1, which is right at the edge of Octane's compatibility.  Octane no longer supports cards with versions lower than 2.0 compute capability.  So I'm guessing that is why your card works using CUDA 3 just fine with Octane and 3D-Coat.  My 980 has a compute version of 5.2 and will not run with the older versions of CUDA.   :(

Link to comment
Share on other sites

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.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

 Share

×
×
  • Create New...