Sorry for this fairly unhelpful reply but I shall go ahead anyway - Has it
always been so memory heavy? The other night I was rendering a car scene
(cad data that I exported from rhino myself) with a similar poly count that
you mention, one minute it was rendering fine, very fast 'scene export' esq
times etc.

Then I submitted it to the farm, hoping to leave and go drink fine belgian
beers and all the machines failed, I couldn't tell why, so rendering
locally I saw that it was trying to use all the 32gb of ram... I tried
teaking the dynamic memory limits etc. but to no avail - so I tried
deleting some geo that wasn't visible and it worked... (but it's not as
though i deleted 5 million polys)

Then in the morning despite my headache I tried to debug but it rendered
fine with all the geo. So I just had mysterious 6pm problems..... (and that
is why my answer is not that helpful)





Simon Reeves
London, UK
*si...@simonreeves.com <si...@simonreeves.com>*
*www.simonreeves.com <http://www.simonreeves.com>*
*www.analogstudio.co.uk <http://www.analogstudio.co.uk>*


On 17 February 2014 10:24, Morten Bartholdy <x...@colorshopvfx.dk> wrote:

>   Apologies for the OT, but I gather there are many Vray users here, so I
> thought i would give it a try, as we are in a pinch here.
>
>
>    We have a scene with a reasonable amount of geometry ( 8 million
> triangles ) and lots and lots of glossy reflections. Vray crashes on the
> farm due to memory usage (machines have 24 GB RAM). We are investigating
> the culprit, since neither geometry nor textures should use a lot of
> memory, and the glossy reflections should not use a lot of memory, onlu cpu
> cycles. Ligthing is a mix of HDRI and a few lightsources, so currently we
> suspect either light cache and/or BSP setup. We inherited the scenes from
> someone else, so we are careful to not break the look, but are looking into
> what can be tweaked.
>
>
>    Are there some seasoned Vray users here who might be able to point us
> to what to optimize for this specific problem?
>
>
>    Thanks!
>
>
>    Morten
>
>
>
>

Reply via email to