Krakatoa

#521

I loaded my test scene in Max 2008 32 bit using Krakatoa 1.1.2.32992 and could repro the crash you reported when it hits the Matte Objects.
I deleted all teapots but one (262,144 faces) and no crash. I cloned it once - no crash. Tried 3 teapots - still ok. 6, 9, 15 - still working… It even worked with 25 clones. It looks like slowly increasing the memory consumption of matte objects within the session would make the crash go away. I opened the original scene within the same session and it crashed again… Memory management is a funny thing!

I decided to launch the same Krakatoa build in Max 2008 64 bit to cover all cases. As I hoped, no crash. The timing was as follows:

PRG: Rendering frame 0
STS: Section “Retrieving Particles”:
STS: Total 00h 00m 22.156s Called 1 times Avg 00h 00m 22.156s
PRG: Rendering 50000000 particles.
STS: Section “Render:Matte Objects”:
STS: Total 00h 02m 06.234s Called 1 times Avg 00h 02m 06.236s
STS: Section “Render:Sorting Particles”:
STS: Total 00h 00m 7.375s Called 1 times Avg 00h 00m 7.375s
STS: Section “Render:Drawing Particles”:
STS: Total 00h 03m 09.047s Called 1 times Avg 00h 03m 09.047s
PRG: Finished rendering frame: 0

This equals 5m46s.

Then I tried an internal Beta build of a future version of Krakatoa in Max 2008 32 bit and it did not crash! The times were longer than the 2m45s for the 64 bit version of the same build I posted earlier, pretty much 4 minutes (which should make you excited for what is coming later this year, and it is even faster when using materials with heavy mapping and culling!)

PRG: Rendering frame 0
STS: Section “Retrieving Particles”:
STS: Total 00h 00m 20.250s Called 1 times Avg 00h 00m 20.250s
PRG: Rendering 50000000 particles.
STS: Section “Render:Matte Objects”:
STS: Total 00h 02m 00.453s Called 1 times Avg 00h 02m 00.453s
STS: Section “Render:Sorting Particles”:
STS: Total 00h 00m 35.703s Called 1 times Avg 00h 00m 35.703s
STS: Section “Render:Drawing Particles”:
STS: Total 00h 01m 04.562s Called 1 times Avg 00h 01m 04.562s
PRG: Finished rendering frame: 0

The conclusion is that there is a memory issue in Krakatoa 1.1.2 32 bit when preparing heavy Matte Geometry but it has been resolved in the current internal build. 64 bit appears to be working in both the shipping and the future versions as advertised.

#522

Chad Capeland of Anatomical Travelogue has once again pushed Krakatoa into unexpected territory. Enjoy:

http://www.anatomicaltravel.com/research/2009/02/fun-wit-voxels-chameleo/

#523

Holey crap. i was thinking “whats this a splash screen?? I dont get it… oh… wait holey crap”

Nice job. Bobo it must feel good to have krakatoa pushed to lenghts most of us cant even imagine.

#524

Wow SSS with Krakatoa! Dudes doing some amazing stuff :slight_smile: , man need more time in the day…really

#525

heh i know johnny, perhaps if i stop sleeping i can catch up…

#526

That’s insane! How many particles?!

#527

24 Megaparticles. They’re uniform and isotropic, which is not ideal, of course. I could easily optimize it for better distribution later, but keeping the particles uniform makes testing much easier.

And there’s no SSS, just S, being the “Scattering” part. There’s no “Surface” for which to be “Sub” of.

  • Chad
#528

But there is the fantastic “illusion” of it, which was possible earlier, but looks so much different with the voxel engine :slight_smile:

#529

This isn’t voxel. It’s still just point based. The point based rendering has some new features, but more importantly, it’s whipping fast.

  • Chad
#530

No kidding! Did I read write, < 3 minutes for that frame? Impressive, definitely…

#531

Shhhh…

And time is a relative thing, Krakatoa isn’t comparable to other renderers because of many factors. And my computer and scene isn’t comparable to anything else either, so it’s not much for metrics.

#532

Render time is always relative and dependent on many factors but that fact is that time is still impressive for 24 mil particles with such a nice result. :slight_smile:

#533

We are working on making that time shorter, even if it costs us sales :wink:

#534

Ha! We all know that Krakatoa is a loss leader designed to drive Deadline sales, right?

Krakatoa is really a pleasure to use when you have a smoothly running Deadline farm. And we’d gladly take quality (and flexibility) over speed any day.

#535

I was obviously talking about the “multi-shadow-map” request you made recently as it would theoretically produce the same result with 7 times less sorting, so you would get both quality and speed :wink:

#536

I’m still shocked how fast point rendering is, cool to watch it render too :smiley:

Well certainly no where near as cool as Chad’s chameleon but a quick test with voxels and fume driven smoke.
VoxelSmoke QT 4mb

#537

nice snow cannon :smiley:
now go and make one with the cuda-pflow connection…i wanna see some avalanche action!

#538

Hey Johnny, I really like that, it has a very foamy look/feel.
How did u achieve that? FumeFx sim loaded into PFlow?

Cheers

#539

Good idea Anse :slight_smile: something to mess around with.

Carlos, strictly fume, a couple of simple sources, the catch is the new voxel engine in Krakatoa.

#540

Damn I feel so much like skiing now :frowning:

Hey Anselm, did you try that CUDA thing for PF? How’s it doing?