GI and softshadows in 1.46

[flipside]

Hi,

Here two small tests which make me slightly panic (hint: I need that differences list ). All settings are identical in both versions!

Softshadows: they look worse and take 50% more rendertime in 1.46 (with same subdivs on the lights - I thought they would need less)






GI: notice the detail under the handle in the 1.09 version, and the rendertime... I thought the new build needed less hsph subdivs, but they need more actually in this example. Rendertime doubled! The difference may be small but GI detail is what it's all about and if rendertime doubles in a simple skylight scene...

[flipside]

Ok start to panic now!

Here again two images, and an swf to compare them easliy. Notice the GI detail under the small teapots.





http://users.pandora.be/stor1/vray/GItest.html

I don't know how I can get a better result with 1.46 by lowering some setting.

regards,

wouter

[flipside]

samples of the above renders:

1.09



1.46


Might be interesting too.

wouter

[Gijs]

irradiance settings?

high settings in 1.09.03 are not the same as high settings in 1.46.04 for example

[flipside]

Ok I found my answer

Vray 1.46 can indeed get away with lower Hsph subdivs for the GI, but you need to lower the interp samples also! It interpolates much better than 1.09 (even with the same interpolation type). When you leave the interp samples at 20 (as I did in my tests), you simply blur out the good GI solution completely.

I used interp=8 and hsph=25 in this image! Notice how the detail is now better than in the 1.09 image, but at the walls it is worse. But that's the compromise of interpolation vs quality which you will always have with blurry methods.

[flipside]

irradiance settings?

high settings in 1.09.03 are not the same as high settings in 1.46.04 for example

Yes I noticed... I was using custom settings as I thought I knew vray

Should have checked that first, the difference in presets is exactly what I ended up testing...

Sometimes talking to yourself helps very well

[vlado]

You can see that the 1.46 version took more irradiance samples than 1.09, so it's natural that it's slower.

Best regards,
Vlado

[percydaman]

hmmm.. I guess we're all going to have to relearn a little bit. Thanks flipside.

[trixian]

You can see that the 1.46 version took more irradiance samples than 1.09, so it's natural that it's slower.

Best regards,
Vlado

Actually...I can't really see that 1.46 took more samples....looks about the same...the bigest difference is that the sample dots are much larger....thus appearing more dense. Shouldn't the vray log window show how many samples were taken?

[dejaVu]

Trixian,
You can see that the new build took more samples than 1.09 in the ceiling.
Over large flat areas, it is obvious that there are more sample taken.