I started taking screen shots to continue the discussion about CPU and 
compilers.

Instead, I ended up making a page of screen shots of bsac doing its 
network render thing...

http://www.buckosoft.com/bsac/flyby/


(
In the Framelist, note the 17 minute M$ render times vs. 50 minute gcc 
times (vs. 176 minutes for the Celeron (sven...)). 3 to 1 increase for 
M$ [1].

I have another scene, looking at a tile floor of superellipsoids that M$ 
has a 10 to 1 speed increase! [2] My linux boxes hate that floor. :)

And then I have a similar scene (looking through glass at nothing too 
complicated) where M$ has only a 50% increase in speed, 34 minutes vs. 
43 minutes. [3]

I think what this means is, I've got to get my hands on the icc/Ubuntu 
compiler.
)

dik

[1] 3 to 1
http://www.buckosoft.com/tteoac/video/frames/ttho0705.png
[2] 10 to 1 for M$
http://www.buckosoft.com/tteoac/video/frames/ttko0524.png
[3] 50% increase
http://www.buckosoft.com/tteoac/video/frames/ttho2946.png

Post a reply to this message

Am 18.02.2016 um 05:38 schrieb dick balaska:

> (
> In the Framelist, note the 17 minute M$ render times vs. 50 minute gcc
> times (vs. 176 minutes for the Celeron (sven...)). 3 to 1 increase for
> M$ [1].
> 
> I have another scene, looking at a tile floor of superellipsoids that M$
> has a 10 to 1 speed increase! [2] My linux boxes hate that floor. :)
> 
> And then I have a similar scene (looking through glass at nothing too
> complicated) where M$ has only a 50% increase in speed, 34 minutes vs.
> 43 minutes. [3]
> 
> I think what this means is, I've got to get my hands on the icc/Ubuntu
> compiler.
> )

Are you sure that's what it means? Microsoft and Intel aren't really the
same, you know.

Post a reply to this message

On 2/18/2016 1:44 AM, clipka wrote:
> Am 18.02.2016 um 05:38 schrieb dick balaska:
>>
>> I think what this means is, I've got to get my hands on the icc/Ubuntu
>> compiler.
>> )
>
> Are you sure that's what it means? Microsoft and Intel aren't really the
> same, you know.
>
No they aren't the same.  But there is no M$ compiler for ubuntu and I 
want to try something other than gcc.  Given that we expect the i7 to be 
about 20-40-60% faster than the i5, and I'm seeing up to 10x speed 
increase, one possible explanation is that the M$ compiler generates 
superior code than gcc (for povray anyway).

Reading this red vs. blue page
https://software.intel.com/en-us/forums/intel-c-compiler/topic/327585
it looks like maybe I can play with the optimizations on gcc. -- I 
haven't checked if povray _already uses_ the suggested optimizations, or 
if they're even viable, in which case I'm back to pining for icc.

I've generated 94,772 frames [1] on my 4 Ubuntu boxes in the last 4 
months.  Anything I could do to speed that up would be great. (other 
than changing my superellipse floor to boxes ;) )

dik

[1] data derived from here, Servers tab
http://www.buckosoft.com/bsac/meta/

Post a reply to this message