Just wanted to drop a note and say I'm very impressed with the multicore 
support in 3.7RC3.

I ran the benchmark in 3.6 on my new system (Athlon X4 3.1 GHz quad-core) 
and it took about 13 minutes to render at 384x384.

With 3.7, it took about 2 minutes.  Oddly, though, I couldn't get it to 
display the output while it rendered.  Should +d work when rendering with 
multiple cores?

Jim

Post a reply to this message

On 11/29/2011 02:02 PM, Jim Henderson wrote:
Should +d work when rendering with multiple cores?

should ... and does ;-)

Post a reply to this message

Am 29.11.2011 20:02, schrieb Jim Henderson:

> With 3.7, it took about 2 minutes.  Oddly, though, I couldn't get it to
> display the output while it rendered.  Should +d work when rendering with
> multiple cores?

AFAIR you don't get output when running the built-in benchmark.

Post a reply to this message

On 11/29/2011 02:24 PM, clipka wrote:
> Am 29.11.2011 20:02, schrieb Jim Henderson:
>
>> With 3.7, it took about 2 minutes. Oddly, though, I couldn't get it to
>> display the output while it rendered. Should +d work when rendering with
>> multiple cores?
>
> AFAIR you don't get output when running the built-in benchmark.

guess I didn't read it completely (old eye's I'd imagine) ... they say 
that's the 2nd thing to go ;-)

Post a reply to this message

On Tue, 29 Nov 2011 14:06:21 -0500, Jim Holsenback wrote:

> On 11/29/2011 02:02 PM, Jim Henderson wrote: Should +d work when
> rendering with multiple cores?
> 
> should ... and does ;-)

Hmmm, now to figure out why it doesn't here...

Jim

Post a reply to this message

On Tue, 29 Nov 2011 20:24:31 +0100, clipka wrote:

> Am 29.11.2011 20:02, schrieb Jim Henderson:
> 
>> With 3.7, it took about 2 minutes.  Oddly, though, I couldn't get it to
>> display the output while it rendered.  Should +d work when rendering
>> with multiple cores?
> 
> AFAIR you don't get output when running the built-in benchmark.

Ah, that might be part of the INI file - thanks.  That's a change from 
3.6, which did give me output.

Jim

Post a reply to this message

On Tue, 29 Nov 2011 14:53:57 -0500, Jim Holsenback wrote:

> On 11/29/2011 02:24 PM, clipka wrote:
>> Am 29.11.2011 20:02, schrieb Jim Henderson:
>>
>>> With 3.7, it took about 2 minutes. Oddly, though, I couldn't get it to
>>> display the output while it rendered. Should +d work when rendering
>>> with multiple cores?
>>
>> AFAIR you don't get output when running the built-in benchmark.
> 
> guess I didn't read it completely (old eye's I'd imagine) ... they say
> that's the 2nd thing to go ;-)

No, no, memory's the second thing to go. ;)

Jim

Post a reply to this message

Is it actually meaningful to compare the benchmarks of
3.6 and 3.7, or did the benchmark itself also change?

Post a reply to this message

On Wed, 30 Nov 2011 02:05:37 +0100, Christian Froeschlin wrote:

> Is it actually meaningful to compare the benchmarks of 3.6 and 3.7, or
> did the benchmark itself also change?

I did notice that the size of the benchmark image increased - it was 
384x384 with 3.6, and 512x512 with 3.7.

But in the end, what I was most interested in was seeing the difference 
between the two, and a reduction from 13 minutes to 2 minutes is 
significant. :)

Jim

Post a reply to this message

On 11/29/2011 08:05 PM, Christian Froeschlin wrote:
> Is it actually meaningful to compare the benchmarks of
> 3.6 and 3.7, or did the benchmark itself also change?

There were some changes ... as already pointed out the image size was 
increased, and a text object "3.7" was added. IIRC there was also some 
photon tweaks done as well. I think a some point the built-in benchmark 
and the scene file had gotten out of sync with each other ... now they 
be the same.

Post a reply to this message