Progressive path tracing renders

[tom schuelke]

Hi natty

yes i did try this allready, but what doesnt work is to do it in only one step. this way, the i-map gets calculated, but the final image isnt rendered

[tom schuelke]

Ah

i´ve got another idea. Wouldnt it be simple possible to build in a button for immediate Scene feedback ? This way we could get something simillar to fprime while i am sure fprime is much more optimised for this.

only a suggestion.

[plastic_]

i think this PPT thing should be seperated from the lightmap, a new GI method in the drop down rather.
things are getting confusing again, this must be a nightmare for newbies.

[Trapezium]

My PPT Test



This was after about 7.5 hrs, rendered on a dual 1GHz Xeon with 1GB of RAM.

--Jon

[natty]

i think this PPT thing should be seperated from the lightmap, a new GI method in the drop down rather.
things are getting confusing again, this must be a nightmare for newbies.

totaly agree .. then it also could be used as a second bounce maybe , with out saving it first.

[Nicolaas]

Hiya,

Didn't had time to check it out BUT..
What's the use of this method anyway..
It's a very grainy kind of thing..
I really don't like it.. it's like applying a PS filter in post.. maybe it's personal..
Had to rebuild my system today... maybe I'm very nervous.. ehhe..

Nicolaas

[tom schuelke]

Hi Nicolaas,

there are several advantages

1. you can start it and stop it whenever you want. restarting it and going on isnt possible untill now

2. setup is very very easy, turning on gi caustix will make caustix appear without any settings, while there arelimitations, due to mathematical reasons, so a light must have an area, the bigger, the faster causix will appear. pointlights would never be seen.

3. i think this is a first step. several possibilities for the futuer may be possible.

a: distributed ppt
b: ppt in secoundary bounce without having to load it manualy
c: maybe an interactive scene feedback when anything changes. so you use it like fprime maybe.


hmm. vlado told this isnt bidiractional path tracing as with maxwell. so maybe this is the reason, selfilu objects and lights doesnt get rendered in caustix as fast as in maxwell. am i wrong ?

[gadzooks]

Trapezium what were your settings for your ppt. After saving the light map file did you use high settings onthe IR MAp. Very nice images BTW. Gadzooks.

[Trapezium]

Thanks

Light Cache:
Subdivs = 20000
Sample Size = 0.005 (because of the indoor/ outdoor borders)

When I rendered the examples in my other post I used Light Cache for both but just loaded the PPT Light Cache File and set my AA @ Fixed = 4

--Jon

[TRGraphics]

I tend to agree with Nicholaas. I don't see the attraction myself. I would have prefered that they make Lightcache work with DR then to come up with this "new feature" this late in the build.

DR with Lightcache is a very critical issue with me and others here, judging by the posts people have made.

We were told it was high on the list, but apparently thats not the case. I am not happy about this, at all!

Sure a new feature is fun to play with. But, I'm getting tired of playing. I want the existing features to work properly, simple as that. I make my living with this software and playing is not high on my or my families list of things to do. Lightcache with DR would put additional money in my pocket by reducing the time it takes me to finish a job and get on to the next one without sacrificing quality.

Sorry if I put a downer on the new toy. I guess I'm getting a little frustrated with were this software is headed and when or if it will ever get where the existing tools actually work the way they should.

Tony

[Gijs]

Tony,

I think making it distributed is not very easy. Furthermore I don't think this new 'toy' has cost vlado more than a couple of hours.

[natty]

i agree with Gijs .. PPT was already there in the lightcache.. because it seems to do the same thing.

[TRGraphics]

I'm sure that none of the programming for vray is easy. I firmly believe they work very hard at what they do. There is no question of that in my mind.

I would just like them to concentrate their efforts on existing features for the 1.5 release. They have said that DR can be done with lightcache. I just want to see it done. In my humble opinion, that is one of the very few things lacking in the program.

If I offended you with the the use of the word "toy" I guess I'll just have to live with that. If I can't use a feature to make my job faster or achieve better quality than before it was added, I consider that a toy and useless for production until it's perfected.

I'm sure that since this feature is now a part of the build they will want it working properly before they release 1.5. Do you really believe that can be achieved in a couple of hours? I hope your right. But I have serious doubts that it can.

Tony

[vlado]

They have said that DR can be done with lightcache.

Yes, in principle. Some steps towards distribution were made with the introduction of the light cache passes (they distribute the light cache among the processors on one machine) but there is a lot more research that needs to be done.

As for the path tracing, one of the early 1.45 betas last summer used to do it always when calculating the light cache, so what we did was simply expand that idea.

Best regards,
Vlado

[Gijs]

No Tony, I am not offended at all. I placed it between '' because it is partly toy (because of slow speed) and partly usable (because of accuracy)