Messiah render speed comparison.

#21

Here is a render with the default settings supplied with the scene:
Total Render Time:00:08:21
2.13GHz
2xAthlon MP2800+
1GB RAM

#22

I figured out the anomaly and why it happens.

Those weird shading errors that show up are caused by a combination of GI radius, GI gather count and GI photons.

Basically, the more photons you use, the more finely the scene is “peppered” by them (think of dick cheney).

Your gather radius is the maximum distance messiah will look for a photon sample from a specific sample point, so if you didnt use enough photons overall, and set a high GI gather count, then messiah will start to get progressively farther away until it hits your maximum gather radius. the problem here is that if you have too few photon samples, it will start to sample things you dont want sampled. such as extremely bright objects or reflections, averaging that into your surface shows up as a weird white shape.

so you either have to

A) increase the overall photon count to make sure the sample distance doesnt get out of hand.
B) decrease the GI gather radius, which will allow more noise.
C) reduce the GI gather count - which will reduce accuracy

#23

So your saying the trick is. . . to decrease your gather count and the artifacts will go away?

#24

to preserve quality, i would start by increasing photon count. for more speed, i would decrease gather radius, but you might start running into noise.

But there is more to it!

so lets say you decreased the GI photons to get alot of speed and found the sweet spot for your gather radius/gather count… now is the time to start playing with GI samples! lol. fun stuff.

oh and by the way, you know the halo around the logos in your render?

the halo is there because its averaging the Red wall on the left, the ceiling on top and the blue wall all the way in the middle of the green wall. making what appears as a halo. thats a clear indicator that your GI gather radius is too high.

#25

Mine? Dude I set mine to like. . . 3.

We should be doing this in IRC. . .

#26


Going from 3 to 300 removed the artifacts in the cube and introduced them into the sphere. I really think this is some kind of bug. 3000 looks exactly the same as 300.

#27

try reducing the GI gather count. did you get the latest project file from the zipped file i provided? because the back wall is non planar from the first link i made that also affects things.

also you are using a tremendous amount of photons. i dont think you need that many. probably around 300k

overall its a ballance of the 3. it has to be a happy medium of photons, radius and count.

#28

The more photons you use. . . the darker the shadows and the more accurate it looks. It may not make much of a difference in this scene but when I use it elsewhere it does. Adjusting the Gather Count does nothing. I tried it at 1 and 10,000.

#29

heh heh. it does affect time though. i think you will be surprised with my next render. :slight_smile:

#30


5:22

Look ma, no anomalies!


one question though… the render looked great, the shadows where smooth as hell and i was like daaamn. thats looking good. but then at the very end. as a last step, it just grained it all up??? WTF??? whatever that feature is, id like to turn it off.

#31

Those shadows are pretty close to being called “bad” though stooch. And that “enhanced” AA is always horrible. Its just a big blur pass at the end. Do that in post if you need it. Not my cuppa tea. It is cool how fast that is rendering. . . but if you look in the ball’s reflection. . . you can see the wall the camera is looking through is grey when it should be white like the floor from all the GI in the room. You haven’t really removed the artifact. . . . Just hidden it a little.

#32


4:21 note the difference… why the noise???

#33

no its removed… why should the wall be bright if it isnt directly facing the light source?? the floor is facing the light thats why its brightest…

the example from kray and LW are using an area light btw. this is with a spot light.

#34

Spotlights in Messiah work exactly the same as area lights in Lightwave. I have tested this multiple times.

Look at the renderings from all the other programs. The back wall is white.

edit* The noise you are seeing in the shadow is because you decreased your threshold on the AA. When it is too high it starts garbling up shadows. You went from .05 to .02.

#35

I just slept about an hour and only woke up from some f%$&#ng donation call ("…our poor policemen, the government ain’t interested in them, so can you foreigner please give us some money to get them their donats and let them give you some more parking tickets?")…ehm…what…what am I doing, what’s going on here? Oh, render nonsense?! I have no idea what this is you’ve discovered down there (HAHAHA…HA…ha…I mean in the thread), but it looks entertaining. I’ll see if you guys have posted that scene somewhere so I don’t have to recreate a similar scenario. As for enhanced, I never use it, because effectively not only does it look like some strange blur pass, but it also screws up the supersampling on noisey areas! So your AA level may have had the perfect settings to eliminate most or all of the grain, but as the enhanced kicks in, it takes the first coarse part somehow into account and comes up with this…bpflrrrrbt…blurp…crx…screwed up blurry granula memory of the past…nah, I don’t need it either. Instead, this is a great reminder to continue with this age old argument that we’ve had from like…hmm…let’s say day one of this pretty concept. There are some thing about the enhanced mode, that do make sense. But I’m incapable of explaining them, really, judging by what else it does. Not to mention that in the “hidden” (hit F3) area, you can see that you can adjust the range in which enhanced will sample around each pixel to come to it’s weighted results! The larger this range, the more it becomes soft. It’s actually not a blur, but the larger you make that area, the more it really acts like one.

That gi error?! I don’t even know what the heck that is? Something funky with photons? That’s a possibility.
Also, if you do not use the photons (I hardly ever use them, actually) 16 GI samples brings you to a super high sampling count 256 different color intensity levels the radiosity can detect. Going higher than that is almost certainly exaggerated. If that doesn’t result in nice renders, then something else is funky in the scene, most likely. Making 4 bounces, I think is rather insane…although it does make sense in a simple scene like that…it might even need them. I’ll have to check it out anyway.

So this is really all very interesting! Still wonder what that funny “bug” actually is. Some wicked gathering or distribution thingy with the photons I think…no idea yet.

SO THANX! :wip:

Taron

#36

hmmmm, whatever the problem may be… having photons working is pertty important IMO. the time savings are huge, even with GI depth of 4. the reason why im using 4 is because Kray is set to 4 bounces.

#37

half resolution is bad :confused:
but I use the messiah demo version :slight_smile:

That’s why I can’t play with you :confused:

#38

21:36

Gi samples 5, depth 2. no photons.

#39

AA at which level?

I need to check out that scene…and make coffee…no, coffee first! :surprised

#40

Adaptive SS
AA level 4 Threshold .02
AA pattern regular
size 1.5