Blender benchmarks

Jarod, of the german Blender community has startet to collect some
rentering times for a simple test file. Here are the result. Perhaps some of you are interested:

My machine rendered the test file in 3:40.

Pentium 4 3.2GHz
1 Gig RAM
Windows XP SP2
Blender 2.37a (Intel optimized)

Just a little FYI. :smiley:

My machine rendered the test file in 4:00.

Pentium M 725 (1.6 GHz)
512 MB RAM
Windows XP SP2
Blender 2.37a (Intel optimized)

7m44s on my box.

Athlon tbird 1.4ghz
768MB ram
gentoo linux
Blender 2.37a, compiled from source

These are my machine’s results.
Mind the difference between the one with treads swithed on, and the one without.

test.blend rendered in 6min12s,
Intel P4 hypertheading, threads button swiched off.
WinXP SP2, GAIM, FireFox and WMP running. 1GB of RAM, Blender 2.37a

Same machine, now with threads button on.

using treads-rendering speeds my machine up with about 20% . Isn’t much, but is sure is something.

4min08s with 1.25 GHz G4, 1GB Ram. 2.38 build.
Mac OS 10.4.2.

It took 7min10s with 2.37a though.

Altivec optimisation would make quite a significant difference on top of that. I wish gcc had better auto-vectorization.

My machine rendered the test file in 5:50.

Pentium 4 1.80GHz
128 MB RAM
Windows XP SP2
Blender 2.37a (Intel optimized)

wth? :o

Pentium IV 2.0Ghz
Windows XP Home SP2
Blender 2.37a

rendering time: 30 mins 27 seconds

a pentium 3 450Mhz 64MB RAM did better

maybe u had SpeedStep activated???


thx for your render values!

now it’s possible to add render times to the table by yourself!



Why did it add two entries for each of my values?

It automatically put Apple G4 1.25GHz and IBM G4 right below it. If someone could delete the Apple ones and leave the IBM entries, that would be great. Or delete the IBM ones, but the chips are IBM and some Macs may use Motorola chips so I would delete the Apple ones.

HI Jarod,

I mistakenly wrote Athlon XP instead of Sempron. Will it matter much, if so I can resubmit it, if you cannot delete it or modify it.

I am currently at 19number in the list.

You could try turning on the treads option in the render controlls :wink:

getting approx. 10min39sec on gcc-o3 compiled blender 237a, ubuntu linux, single p3-733.

You could try turning on the treads option in the render controlls ;)[/quote]

I think that would work for dual-processor machines. I meant special vector instructions that make code execute faster on a single processor. They make loops (for loops, while loops etc) go faster by sometimes as much as 4x. It’s like SSE2 optimization on Intel machines. Manually it needs someone to know the specialized code but auto-vectorization is supposed to do most of it on its own. Unfortunately, the feature is in its early stages and doesn’t really support most of the loops.

Not too bad for a… 3? year old computer that wasnt even that great when I built it?

Athlon XP 2200+ @ 2ghz
768m DDR (maybe i should try OC’ing this too?)
Ubuntu Linux 5.10 (Breezy) Preview release (well, it was hoary originally, but I enabled breezy reps and update constantly)
Blender 2.37

I’ll try this in windows too, as soon as I install one of those lite versions.