[Bf-cycles] Cycles: Add an experimental CUDA kernel.

[Vlad MS]

Hi,

So what exactly is causing the huge memory spikes when using GPU SSS?

Thanks

Vlad Mafteiu Scai

Mobile: +45-5333-30-67


*Mobile UI Designer, Product Visualization Artist*
Email: mafteiuscaivlad at gmail.com

Portfolio: http://www.behance.net/motionseadesigns


This e-mail is confidential and is intended solely for the use of the
recipient.  If you are not the intended recipient, be advised that you have
received this e-mail in error and that any use, dissemination, forwarding,
printing, or copying of this e-mail is strictly prohibited.


2014-08-28 6:50 GMT-07:00 David Black <db4tech at yahoo.co.uk>:

>  Hi Carlo,
>
> Ah, yes, memory usage statistics provided were from Cycles reported usage.
>
> David
>
> 3d-designs-davidblack.blogspot.com
> <http://www.3d-designs-davidblack.blogspot.com>
>  On 28/08/2014 05:08, Carlo Andreacchio wrote:
>
>  Just tested the memory usage, there still is a massive increase in
> memory usage when using experimental vs supported kernels...
>
>  Generic Test scene:
> Supported - 850mb
> Experimental - 1408mb
>
>  Tried to run Mike Pans BMW benchmark but I could not get it to load
> under experimental kernel (1500mb ram limit, kernel was too big)
>
>  Are you using a external program to report the memory usage for the GPU?
> Cycles does not report GPU memory correctly to blender.
>
>  Carlo
>
>  >Hi Thomas and Sergey,
> >
> >http://lists.blender.org/pipermail/bf-blender-cvs/2014-August/067323.html
> >
> >Firstly, thank you kindly for adding experimental GPU SSS cubin files,
> >what an exciting, unexpected, surprise!
> >
> >Amazed to find Experimental cubin using exactly same amount of memory as
> >Supported cubin in a complex scene containing SSS. Also, less memory
> >than Blender 2.70a and 2.69.1 (GPU SSS test version).
> >Understandably Experimental cubin are a little slower than Supported
> >cubin due to SSS calculations, but not by a massive amount!
> >If helpful/interested added GPGPU memory consumption and speed test
> >results at the end of this mail.
>
>
>
> _______________________________________________
> Bf-cycles mailing listBf-cycles at blender.orghttp://lists.blender.org/mailman/listinfo/bf-cycles
>
>
>
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> http://lists.blender.org/mailman/listinfo/bf-cycles
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-cycles/attachments/20140828/9f511deb/attachment.htm