Progressive path tracing

[Metin_7]

Hi guys,

I'd love to know a bit more about the new Progressive Path Tracing. I've been away from the beta versions for a while and resorted back to 1.09, but now that the completion of 1.5 is really nearing I'm going to go back to the 1.5 beta.

I've checked out Vlado's PPT tutorial in the online reference, but I've got two questions:

1: Is PPT only possible with Light Cache in both primary and secondary GI or can you use Photons, Irradiance and/or QMC as well?

2: Is the result the best possible result you can achieve with the VRay engine if you wait long enough, comparable with the non-biased approach of Maxwell?

Thanks and cheers,

Metin

[priad]

hehe, metin what about your sheeps test?
I'm not astonished by your bridge with maxwell
You can set only the fist bounce in lightcache and second in none of course. I don't think that others work.

I think that its GI approach is similar to Maxwell for the accurate but without the Physical sky of Maxwell, perhaps a futur addon to Vray?
I have make several test and I want to know more about it.
You will try it?

PS:I love your work

[vlado]

Re: Progressive path tracing

1: Is PPT only possible with Light Cache in both primary and secondary GI or can you use Photons, Irradiance and/or QMC as well?

Currently you need to use the light cache both as a primary and secondary engine. I'm thinking of a hybrid light cache/photon mapping approach (something like bi-directional path tracing) but this is yet to come.

2: Is the result the best possible result you can achieve with the VRay engine if you wait long enough, comparable with the non-biased approach of Maxwell?

No. Often you can get better (and probably faster) results with traditional methods, like QMC GI (although these may be biased). Further on, currently VRay can generate up to 2^32 different QMC paths internally. The max value for the light cache subdivs is 60,000 which allows for 3,600,000,000 different paths. When you distribute these over the entire image, this may turn out to be insufficient. For a 2000x2000 image, you can only get 900 paths per pixel, which may not be enough to bring noise to an acceptable level. Traditional methods are not that limited and are able to put more samples into it.

Best regards,
Vlado

[Metin_7]

Thanks very much for your informative replies, Priad and Vlado! Somehow this new forum doesn't notify about replies by e-mail, did others notice that too?

Last time I used the VRay 1.5 beta my favorite GI method combination was Light Cache + QMC: it offered speed as well as the right amount of GI detail / accuracy. What are your favorite combinations if I may ask? Please explain your choice as well, I'm very curious what the pros around here recommend.

Thanks again!

Cheers,

Metin

P.S.: Priad, thanks for your compliment! Much appreciated.

[priad]

For my part I use first irmap and second lightcache for the fast result and the good smoothed GI. I found lightcache very fast and easy an other advantage the good viewing during the render.
And with the irmap in first either with a medium setting you can have a very good GI with lightcache in secound bounce.
I use it for exterior and interior render.
That's all that I can say for my little experimentation.

for PPT you can take a look at this, there is opinions about it.
http://new.chaosgroup.com/forum/phpB...4e7c2067541332

[Metin_7]

Thanks again, I'll check out that thread.

Cheers,

Metin

[Metin_7]

By the way, Priad, my cartoon sheep were rendered with Scanline, not Maxwell.

Cheers,

Metin

[priad]

I know

[Metin_7]

I thought you were serious about the Maxwell sheep, sorry.

I suspect that some people actually thought I had tried to create a cartoon look rendered with Maxwell.

Cheers,

Metin