[Bf-committers] need help testing sculpt tool performance

Tom M letterrip at gmail.com
Fri May 28 04:09:59 CEST 2010


Hi all,

with the recent sculpt calculation optimizations there have been some
nice performance gains.  I'm curious how much performance gain is for
really good hardware.

I'm on a

iMac 2.8 GHz Core 2 Duo, with 2 GB Ram

with the following settings


start with a moderately number of face mesh (real faces 384 after 3
levels of multires then apply or maybe one level above that 1296 for 4
levels).
turn off double sided (object data>double sided)
turn on smooth shading (space bar>shade smooth)
enter sculpt mode
enable fast navigate in sculpt mode
add a new multires modifier and add multiple levels of catmul clark

I can get 25 million quads it is still fairly buttery sculpting for
clay brush of various sizes.  Unfortunately the amount of time
swapping going from 6 to 25 million quads takes 15-20 minutes so I
can't really go any higher.

If someone has a piece of beast hardware (8GB ram or more, fast quad
core or better), I'd be curious how many polys you can go to and what
level is comfortable

ie how many quads can you get and have buttery smooth sculpting with
the clay brush (try different radiuss and strengths)
how many quads can you sculpt on and it still be 'ok' performance (a
very slight lag)
how many quads can you reach 'period' if sculpt performance wasn't of interest.

You will need to 'switch' to the sculpt SoC branch for this experiment.

https://svn.blender.org/svnroot/bf-blender/branches/soc-2010-jwilkins

Thanks,

LetterRip


More information about the Bf-committers mailing list