[Bf-cycles] Collecting official benchmark results

[Carlo Andreacchio]

Just tested it under linux, same sort of GPU Utilization happens....
rendering via python / command line is approximately 10% quicker then
pressing f12.

I used the command watch -n0.1 "nvidia-settings -q GPUUtilization" to
measure the current GPU usage. Same spikes occur.


On Sat, Oct 31, 2015 at 11:19 AM, Carlo Andreacchio <carlorules at gmail.com>
wrote:

> Non dedicated card... I just tested on one of our GTX590s (dual gpu on one card) just rendering off one of the cards and it is the same (tried both cards individually). GPU usage bounces all over the place but when running it through bpy.ops.render.render() or via command line it is 100% consistently.
>
> These machines do not have a monitor attached as they are headless nodes... accessing via VNC. Same results as on our artists workstations.
>
> I can provide further testing later today if needed (two GPUs in one machine, one fully dedicated CUDA card)
>
> >Did you use dedicated card for rendering?
>
> >On Sat, Oct 31, 2015 at 4:41 AM, Carlo Andreacchio <carlorules at gmail.com <http://lists.blender.org/mailman/listinfo/bf-cycles>>
> >wrote:
>
> >* >Please always do final (F12) renders with the same resolution and number
> *
> >* of
> *>* samples as stored in the file.
> *>>* We have noticed that F12 renders are not the most consistent measurement
> *>* of benchmarks (under Win7 & CUDA at least), and that rendering via command
> *>* line / python produces a more reliable benchmark result as the GPU
> *>* utilization is always at 100%
> *>>* The variation of F12 render times can be as great as 10%, at least in real
> *>* world scenes.
> *>>* Looking forward to your OpenCL developments!
> *>>* Carlo
> *>
> >* _______________________________________________
> *>* Bf-cycles mailing list
> *>* Bf-cycles at blender.org <http://lists.blender.org/mailman/listinfo/bf-cycles>
> *>* http://lists.blender.org/mailman/listinfo/bf-cycles <http://lists.blender.org/mailman/listinfo/bf-cycles>
> *>>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-cycles/attachments/20151031/3705e303/attachment.htm