[Bf-cycles] CPU Threads benchmarks

[Greg Zaal]

Interesting. I tested some scenes, both simple and heavy, on both windows
and linux and noticed only tiny changes in render time (2 seconds for a 9
min render, 0.3 seconds for a 1 minute render) - using 3.4GHz 8 core
i7-2600k.
Neither auto-threads or manually using 8 was consistently the winner
either. In some scenes auto (9 tiles) was quicker, and others 8 tiles were
quicker.
With such small differences and inconsistent results, it seems more likely
that it had little to do with the number of threads used and could probably
have been caused by other background programs doing their thing.

Greg


On 8 July 2013 00:12, Matthew Heimlich <matt.heimlich at gmail.com> wrote:

> I'll compile the times in a pdf when I get a chance, but on my i7 2600k @
> 4.6GHz setting manually to 8 threads was faster for every test in the
> cycles suite than using auto-detect.
>
> Matt Heimlich
>
>
> On Sat, Jul 6, 2013 at 3:48 PM, Thomas Dinges <blender at dingto.org> wrote:
>
>> Hi,
>> some time ago (summer 2012), we changed the amount of threads for CPU
>> rendering to "number of threads + 1", when using the threads "Auto
>> Detect" feature, which is default.
>> This topic has been raised a couple of times since then, especially
>> users were wondering why they have e.g. 5 tiles on screen, although they
>> only have 4 CPU cores.
>>
>> I did some tests today, and it seems there is no real performance gain
>> by doing "threads + 1". It would be good though to re-confirm this on 8
>> thread systems too (Core i7, Hyper Threading).
>>
>> Here my benchmark results:
>> http://archive.dingto.org/2013/blender/code/Cycles_Thread_Bench.pdf
>>
>> Best regards,
>> Thomas
>>
>> --
>> Thomas Dinges
>> Blender Developer, Artist and Musician
>>
>> www.dingto.org
>>
>> _______________________________________________
>> Bf-cycles mailing list
>> Bf-cycles at blender.org
>> http://lists.blender.org/mailman/listinfo/bf-cycles
>>
>
>
> _______________________________________________
> Bf-cycles mailing list
> Bf-cycles at blender.org
> http://lists.blender.org/mailman/listinfo/bf-cycles
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/bf-cycles/attachments/20130708/3148331a/attachment.htm