[ShockFire]

So I decided to do a 'simple test' with the beta
As you might have guessed from the title, it is now at 1000 hours render time
I know this is way too long for this scene, but I can't really figure out why
I was wondering if this belongs in the Breaking Barriers forum since both render time and times continued must be records
Anyway, here are the stats:

Resolution: 1600x1200
Render time reached: 1009 hours
Sampling level reached: 24.62 (so close to 25!)
Meshes: 174
Triangles: 582740

Processor: Pentium 4 2.0GHz
RAM: 1.0 GB

MXS size: 27.3 MB
MXI size: 210 MB
Continued: 212 times
MaxwellBench 1.0: 11.454237


And here it is:

[Thykka]

geesh.. And there's still noise

[aitraaz]

yes! It's a world record

[Maximus3D]

My god.. a 1000 freaking hours! boy you're nuts but that's one massivly huge sampling level you hit with it. The crazy amount of hours is the most impressive, although i could think of better stuff to do in a 1000 hours.

/ Max

[ki_cz]

I'd love to see the increments of sampling reached at the end, was it even increasing anymore?

[ShockFire]

i could think of better stuff to do in a 1000 hours.

Well I made a little program that will automatically keep rendering Maxwell scenes
It makes sure to take up all CPU idle time with maxwell renders until the end of time (in a sort of intelligent way)
This is the render that comes in after all the other renders are done and there's nothing to do

I'd love to see the increments of sampling reached at the end, was it even increasing anymore?

I think it currently takes about 4 hours to advance 0.01 sampling levels
Real change hasn't been visible for weeks now unfortunately
However when you do a difference in PS on two renders that are 10 hours apart and increase the contrast to 99 you can see it's still doing something
(even though it's invisible to the human eye):


I still have all the original PNG files for each step (continue)
These start at september and are already over half a GB
I haven't saved the corresponding sampling levels or render times
It is however possible to graph the file size, since this directly relates to the sampling level
It won't be perfect, but I could give it a try and see how it works out

[Thomas An.]

Yeah, this scene has 1028 hours on it (waiting for caustics to show up in reflections) http://www.maxwellrender.com/forum/view ... dispersion

It is however possible to graph the file size, since this directly relates to the sampling level
It won't be perfect, but I could give it a try and see how it works out

This would be very very interesting. Yes the entropy of the image should corolate to file size. The lower the entropy (less noise) the less the file size.
One thing to watch out is that many cuastics do not show up until level 20 or more. Which means on some occasions the file size will decrease until sl20 and then go up again and start decreasing from there.

In any case, here is an old thread about this.
http://www.maxwellrender.com/forum/view ... ht=entropy

P.S. Here is another thread that attempts to predict noise via statistical methods: http://www.maxwellrender.com/forum/view ... 9849#79849

[def4d]

[quote="Thomas An."]Yeah, this scene has 1028 hours on it (waiting for caustics to show up in reflections)

I understand now why i couldn't achieve my version of your dispersion of dispersion

[rivoli]

Yeah, this scene has 1028 hours on it

wow. no matter what you do, thomas has already done it first. even a 100 million hours render.

[Thomas An.]

I understand now why i couldn't achieve my version of your dispersion of dispersion

Well, you can see the dispersion early on (within a couple of hours at least in the preview window) the main problem was to see reflected caustics and for the grain of the caustics to clear a little.

I have to say, caustics in this beta is a severe pain. Everything else could be ready within 6 hours and then have to wait a week just for some grainy caustics.

[aitraaz]

even a 100 million hours render.

sample level: 24.9x10^infinity

[rivoli]

sample level: 24.9x10^infinity

well yes, he started rendering a loong time ago.

[giacob]

[quote="Thomas An."]Yeah, this scene has 1028 hours on it (waiting for caustics to show up in reflections) http://www.maxwellrender.com/forum/view ... dispersion


that is what i 'd call a dispersion of time

[Thomas An.]

sample level: 24.9x10^infinity

well yes, he started rendering a loong time ago.

No, I just use the T900a dongle device.

It basically sends the PC back in time ... so it has been rendering in the past centuries ... went through renaissance, french revolution, first WW, second WW, it goes past year 2003 while NextLimit invents Maxwell render and then comes back to me. All in all it feels like 2sec have past.

These dongles are pretty inexpensive these days the are called TeDiQuBs (temporal distortion quantum bridge warp devices).

[rivoli]

No, I just use the T900a dongle device.

does it look like this?

http://www.xcelco.on.ca/~stevbike/delorean.jpg