Method comparison

[OnoSendai]

Glass acceleration for single dir path tracing (including GPU accelerated tracing) will be added soon, maybe next release.

[Alain]

Thanks for the clear answer

Kind regards
Alain

[Headroom]

@Alain

As stated before, all tracing methods ultimately converge to the same end result. Some just take a little longer. In case of your test scenes simple path tracing is simply not a very efficient method. While the noise in most of the image clears up pretty quickly the area behind the glass appears at first to receive no direct light.

However, your test renders simply did not render long enough to determine if there actually is light getting through the glass. I recreated your scene in Blender only covering half of the opening with glass.

This is rendered with Path Tracing (non GPU as it's actually slower for this scene on my iMac). As you can see it took quite a while for anything to show up in the area in question with PT while in BIDIR it shows up immediately.

So in essence, there is nothing wrong with your scene and there's nothing wrong with Path Tracing it just takes it's time!

While a Single PT Glass Acceleration would help this scene, i'd assume a GPU supported BIDIR (perhaps also with glass acceleration) would be even more helpful.

[Zom-B]

As you can see it took quite a while for anything to show up in the area in question with PT while in BIDIR it shows up immediately.

Use ExitPortals to fix this speed issue

[Headroom]

Yeah. I would have but....

Blendigo (for Blender 2.5) currently does not export the material definition for exit portals.
I got a little lazy of patching it in by hand after the second time.

[Zom-B]

The hint for Exit Portals was also pointing to Alain,
pure PT should really benefit strong using them...
(check Exporter Manual for details Alain!)

[Bosseye]

07_Pathtracing_GPU_OpenCL.jpg

I get exactly the same screwy results when I set my GPU acceleration to OpenCL. No idea why and no one has shed any light on it yet. No issues with CUDA acceleration.

[Alain]

OK, I'll do some tests with Blender.

It shocks me a be to hear that Blender 2.5 does not export exit portals.
Are there some more "bugs" in the Blender 2.5 Exporter ?

Does Indigo RT support instances for Blender 2.5 ?

Kind regards
Alain

[Headroom]

Blendigo 3.0.10.1 for Blender 2.5 DOES support exit portals.Tested and verified

It is, however, not a material definition anymore but a mesh parameter - a check button that you'll have to enable in the same panel where you find the Indigo internal subdivision paramenters.

Also according tho the product comparison on the website Instancing for RT should be supported.

[Alain]

@Alain

As stated before, all tracing methods ultimately converge to the same end result. Some just take a little longer. In case of your test scenes simple path tracing is simply not a very efficient method. While the noise in most of the image clears up pretty quickly the area behind the glass appears at first to receive no direct light.

However, your test renders simply did not render long enough to determine if there actually is light getting through the glass. I recreated your scene in Blender only covering half of the opening with glass.

This is rendered with Path Tracing (non GPU as it's actually slower for this scene on my iMac). As you can see it took quite a while for anything to show up in the area in question with PT while in BIDIR it shows up immediately.

So in essence, there is nothing wrong with your scene and there's nothing wrong with Path Tracing it just takes it's time!

While a Single PT Glass Acceleration would help this scene, i'd assume a GPU supported BIDIR (perhaps also with glass acceleration) would be even more helpful.

Did it take 9h CPU computing until directlight came through the glass ?

[Headroom]

@Alain
I am not sure I understand your question. Whatever can be seen in the status line on the bottom of the image is how long it took to render that image. It rendered overnight.

IIRC in another thread it was mentioned that there is currently a bug in Pathtracing that causes this atypical behavoir. I am sure the folks at Glare are working diligently on finding this bug as it obviouslt limits the usefulness of pathtracing.

Again if you use BIDIR the direct light shows up immediately.

[Alain]

OK, pathtracing with GPU acceleration is still bugy (no directlight).
I did a test running it about half an hour and no directlight shown.
I can do a test with longer rendering, but I doubt that there will be shown some directlight anytime.

Your tests probably are useles because you tested it with CPU computing not with GPU.
I'll test it with CPU only....

Kind regards
Alain

[PureSpider]

As many said before, direct light through glass with pure PT (with or without GPU) is hell of slow.
So your own test is probably useless saying you waited for 30 mins only...
Switch to MLT if you can't stand that

[Alain]

OK, I see there is no way to get directlight through glass within a usable amount of time.
I rendered a testscene for about 2.5 hours and I see only some very noisy areas where direckt light should be.

There seems to be no renderer at the moment which supports GPU acceleration, directlight through glass (in a usable amount of time) and instancing support.

Luxrender does not support instancing for GPU acceleration.
Octane does not support instancing as well.
Indigo supports instancing but no directlight through glass in a usable amount of time.

I guess I have just to be patience until the first renderer supports all needes features with GPU acceleration
Or does anybody know a renderer which supports all these features yet for Blender ?

Kind regards
Alain

[PureSpider]

Vray :<
Wait till Indigo supports MLT on GPU or at least glass accelleration on the GPU