Maxwell 2.0 - 2x to 5x speed up

[Zom-B]

http://www.maxwellrender.com/version2/n ... rison.html

More of the new features will be revealed soon they say...



I hope Indigo gets some noticeable speed optimizations too...

[Stromberg]

I have seen this a few days ago
Looks great, and since i have used maxwell befoire, it will be great to play with the demo version when it gets out

And i do hope indigo gets faster aswell, we all want speed

[CTZn]

I want safety and accuracy.

[neo0.]

Speed would defenitely be nice. Im guessing certain things can be approximated with no noticeable difference due to the limitations of the human eye...

Though, what I would really like to see is better/more user friendly materials. A separate, package independent material editor would be great. Also automatic previewing...

[fused]

Speed would defenitely be nice. Im guessing certain things can be approximated with no noticeable difference due to the limitations of the human eye...

I guess thats what the Maxwell guys thought, too

But thats strictly against Indigos philosophy (if it didnt change with commercialisation).

[CTZn]

And anyways... how fast was MW until... this ? I never used it, but there is no guarantee it is reaching Indigo's abilities.

Someone here can guarantee these multipliers ? Not less than twice as fast as before on the same CPU ? Hah !

Speed is definitly on the CPUs side, given than the renderer code is optimized. I'm still using two P4 CPU's (2003 design), do you own I7's so you are entitled to require even more speed ? Get the right computer before you ask for more, efforts must be dispatched

There has to be a balance between speed and accuracy, mind you. Hence my post above.

[neo0.]

Speed would defenitely be nice. Im guessing certain things can be approximated with no noticeable difference due to the limitations of the human eye...

I guess thats what the Maxwell guys thought, too

But thats strictly against Indigos philosophy (if it didnt change with commercialisation).

True, but "physically accurate" is only as physically accurate as the eye can see.

In real life, there are trillions upon trillions of photons emitted by a single light source, and yet indigo probably doesn't use nearly that many because you your eye can't notice the difference and, with this in mind, indigo presumably saves users unimaginably huge amounts of time.

[Borgleader]

True, but "physically accurate" is only as physically accurate as the eye can see.

In real life, there are trillions upon trillions of photons emitted by a single light source, and yet indigo probably doesn't use nearly that many because you your eye can't notice the difference and, with this in mind, indigo presumably saves users unimaginably huge amounts of time.

It will, if you let it run long enough

As far as the "your eye can't notice the difference" type thing...is that, taking shortcuts will always make a difference either that, if not most likely the pain of coding said shortcut is pointless. I would honestly like Indigo to stay physically accurate and slower rather than faster and inaccurate. Then again, I dont own a copy so ... I've no right "to vote" on this matter.

[jim_archviz]

Well, yes. The goal of a photo-realistic renderer, even physically-based, is to produce an image which the human eye cannot distinguish from an actual photo shot.
The physical model used is an approximation in any case. The issue is to find "good" approximations.

As for Maxwell, I think what I'm reading here is pure speculation. Until someone can produce an image showing differences with previous versions and artifacts, no-one can tell that Maxwell 2 is "cheating".

[Stromberg]

Well i agree with Jim here, i can realy se why mawell would start cheating i mean, it's the light simulator and they have been that for years now.
And it is a unbiased render engine and thats what they want to be, if they would start cheating for speed they might aswell do that from the start.

[lycium]

1. maxwell claim to be unbiased - this means they should get the correct solution + some noise

2. their new version produces substantially different images compared to the old

clearly we have a contradiction, which necessarily implies that they are cutting corners or have bugs in their engine. at any rate, there is consistent error in the image not due to noise (i.e. converging to different results), and as long as this is so their claim to fame of being unbiased is false advertising.

[Stromberg]

I feel like there is no reason for us to talk about this, i trust next limit to know exactly what they are doing.

[lycium]

don't see how we can't talk about it, because nextlimit do know what they're doing...

logically it cannot be that there are two different, and both correct, solutions to the same scene. that doesn't require any knowledge of rendering systems, it is a simple matter of consistency and correctness; either they have used a different set up between the two renders, or the two versions are producing different results.

to give another example, if in indigo you get a different converged result using uni- and bi-directional path tracing, there is an issue in one of the methods (or you are trying to render with point lights, which are anyway unphysical).

maxwell is an awesome piece of software and nextlimit are smart guys who know what they are doing, nobody is disputing this.

[Zom-B]

You have here a point for sure lyc!

unbiased render engine should produce always same visual output if setup is the same!!

Indigo 0.3 vs indigo 1.1.14 (only problem here was the old color problem solved years ago)


So it can be that there are just bugs fixed since the last version to get real unbiased results, and it was "wrong" before.... but I doubt somehow... since the NL guys are a bunch of pros!

something smells fishy here!

[CTZn]

something smells fishy here!

Users continuously asking for more ?

Faster cars, but no deadly crashes ? Tastier food, but less colo-rectal cancers ? Wireless, but EM waves should stop their way before they touch you ?

Haha We dream of non-sense and pay for it becomes real but it's still a dream, only it ends into a nightmare. I'm refering to our everyday life, not specifically to Indigo (or Maxwell) customers

Seriously (sorry for I was sarcastic) lyc said it all. Year by year Boole's law is pretty much verified and that's the future of unbiasedness. Developers should not aim to a wall meanwhile imho.

only problem here was the old color problem solved years ago

Namely, reverse gamma correction was introduced