[Bf-committers] Cycles performance
mike.c.pan at gmail.com
Tue Mar 6 18:17:54 CET 2012
You are right Brecht, I just tested with a GCC-built Windows version of
Blender an the result is comparable with that of Linux/Mac.
On Tue, Mar 6, 2012 at 5:17 AM, Brecht Van Lommel <
brechtvanlommel at pandora.be> wrote:
> It's probably more compiler dependent than OS-dependent. Visual studio
> does not compile the render kernel as well as gcc. I've been
> developing with gcc, so that has some influence, and it's probably
> possible to tweak the code such that it compiles faster on visual
> studio. Thread management could be related but last I checked
> rendering threads were kept busy.
> On Mon, Mar 5, 2012 at 11:26 PM, Mike Pan <mike.c.pan at gmail.com> wrote:
> > Not to start another OS debate. But...
> > I noticed that the performance of Cycles is very OS-dependent. With the
> > rendering engine, the OS can influence the rendering time by ~10%. But
> > Cycles, i am seeing a huge difference in performance between Windows and
> > Linux/Mac, where Linux/Mac is often twice as fast.
> > This cleaned up chart by
> > Olivier<
> > shows the discrepancy:
> > http://oenvoyage.files.wordpress.com/2012/02/benchmark_win_linux.png
> > I've done my own dual-boot test and the results as as follows:
> > Cycles Render: (huge difference in render time)
> > OS X: 3:27
> > Win 7: 6:14
> > Classic Render: (similar time, as expected)
> > OS X: 1:24
> > Win 7: 1:29
> > Both OS are running natively (not under virtual machine), using the 64bit
> > version of Blender 2.62.
> > My question is, Is there any technical reason behind this? thread
> > management? malloc overhead?
> > Mike
> > _______________________________________________
> > Bf-committers mailing list
> > Bf-committers at blender.org
> > http://lists.blender.org/mailman/listinfo/bf-committers
> Bf-committers mailing list
> Bf-committers at blender.org
More information about the Bf-committers