Scene Scale / Fire flies problem

[Peter Grealy]

Hi,

Does anyone know of an easy way to reduce/eliminate "fire flies" from a render?
I think my scene size is to blame for these hot pixels. I've modelled a pedestrian bridge located near a city, the over all size of the model is about 10 kilometres square (modelled in maya's default cm units)

Any advice would be tops.

Thanks

[snivlem]

Can you post a screenshot? Sometimes these hotpixels can come from sharpen aa filters and glossy materials.

[Peter Grealy]

I think you may be right. The hot pixels seem to be occurring on glossy materials (main bridge which is supposed to be a glossy-like plastic). I'm not using any sharpening filters in the sampler though. See attached, Thanks

[Peter Grealy]

Well after 3hrs of trying to fix these hot pixels I've finally come to the conclusion it was the fatal combination of the Vray Sky system and a large water body. Not really sure what exactly was to blame, I've since replaced the vray sky (kept the vray sun) with hdri lighting and problem solved.

Cheers

[snivlem]

Strange, I thought that HDRI would be the problem at least thats what I was thinking originally.

After seeing your problem Im not really sure what else to do here, I haven't seen this before. (Turn off sun's visibility might help).

[Peter Grealy]

My gut feeling is that the refractive water body was causing the weird results. Anyway, I've got nice clean renders now YAY! Thanks for your help.

Cheers, Peter

[Reaversword]

A closer friend who codes, the first time he saw this kind of stuff, told to me:

Did you tried it on a Registered Memory (ECC) machine?.
He believes this kind of non registered memory is the guilty, because if exist an error, memory doesn't reports the error in this pixel.

Have you tried in a mechine with ECC registered memory?.

[aweidenhammer]

turn off the sun's visibility.

[Peter Grealy]

Reaversworld, Thanks for your insight, I do not have EEC (Registered memory) on my machine, just the regular unbuffered DRAM. I did a little recon on this and you are probably right in general when individual bits of data get corrupted. My understanding though that this is a very rare occurrence ("chances for a single-bit soft error occurring are about once per 1GB of memory per month of uninterrupted operation"). So I can rule this out due to the number and frequency of the hot pixels occurring. I will check into EEC ram next time I upgrade though.
Checking the sun's "Invisible" box was the correct fix, all hot pixels disappeared immediately. Thankyou very much aweidenhammer!!!