David ?  I know they call me Julie on Thursday's, but David ?  8-)

Glad to be of help anyways.

Err Jason ....I think.  



Hi David :

> Basically you need to turn off the safety tab in rendering/post
> settings.

  Thanks for your reply , it tweaked my memory and made me realise
why I gave up on post-processing per box . I included the images .
The speedup was awesome , from 230 seconds to 160 seconds , but
obviously much more is now required in ... some other Control
Panels .

  Arjo suggested using per-box post so maybe he has some settings
or a maybe the benchmark project set up to show a user how to
effectively have high quality along with better performance ?

> Set all cpu threads to be used in the rendering settings and
> set post processing per image for low resolution (e.g. PAL 
> or NTSC)

  OK , this is what I've been doing all along after trying per box
post processing , years ago , and it's OK but as you say , this
means only 1 single CPU does the Gi .

  This is why I started this thread , to see if we could simply
tell V5 to render every 2nd or 4th (etc) frame they way it's
done with WCS .

I'm certain there must be a way to command RSV5 to skip frames,
but then again , maybe this could be something for V6 .

  Anyway , thanks again for your reply and also your tips on
rendering large images with Gi per-box post .

garry



> Hi Garry
> 
> Sorry don't know what happened in last post.  I typed a reply, but it was
> gone when it arrived in the user list ? :(
> 
> Basically you need to turn off the safety tab in rendering/post settings.
> 
> Set all cpu threads to be used in the rendering settings and set post
> processing per image for low resolution (e.g. PAL or NTSC)
> 
> The safety tab is for when you use post per box as it renders four times
the
> area per box to allow seamless blending when the boxes get stitched at the
> end of the render.  That's why renders can take so long with this on.
> 
> Setting post to per image negates this, but the downside is only one cpu
can
> complete the post processing stage.  We have asked for multiple cpu's for
> post processing in v6.
> 
> For high res GI images of 4000 pix or more I suggest you use per box post
> processing (still with safety off) with GI blur slightly higher than
normal.
> The max ram limit of 3-4 gigs can cause problems when the post processing
> stage begins and you can run out of ram at this kind of resolution.  You
> will get the 'unable to allocate ram' error as there is no space for all
the
> post image channels to be loaded into physical ram.  Painful after a 12
hour
> render that seemed to be going well.  Don't set 'backup render' for this
> option either as even with per box post realsoft still saves the entire
> image r3d file which will fail in post the same way.
> 
> With the resolution this high the box seams are less obvious between boxes
> and can be easily fixed in photoshop.
> 
> Hope this is of use
> 
> Regards
> 
> Jason
> 
> 
> 
> 
>> Hi Garry,
>> I don't understand why you're only using 1 CPU with gi rendering. Why 
>> don't you use per box post processing?
>> Arjo.
> 
> Hi Arjo :
> 
> I have had no luck with that .
> 
> For instance , I just downloaded the alien/ufo benchmark again and
> tried box rendering the Gi postprocessing , it takes over an hour
> to do . Select "Effect/Image" and it does it rather quickly , but
> just 1 CPU is working for the Gi .
> http://www.realsoft.fi/updates/benchmark
> 
> I guess I must be doing something obviously very wrong , considering
> the benchmark scores for the "UFO" . No idea what's wrong , and I
> just gave up on it a few years ago and kept using "Effect/Image"
> which works perfectly every time .
> 
> Windows XP (sp1/sp2)
> RS V5sp3
> Dual Xeon 3.2 Ghz
> 
>  Please forgive my stupidity . It's been a while since I've hit
> the Gi render button and perhaps the cobwebs are thicker than
> thermal paste .
> 
> garry
> 
> 
> 
> 
> 
> 
> 
>> studio wrote:
>>> Hi :
>>>
>>> When doing RS Gi , 1 CPU will be working and 7 CPU's will
>>> be sitting idle !
>>>
>>>
>>> garry
>>>
>>>
>>
> 
>



Reply via email to