LuxRender vs. Cycles speed

[Paapaa]

What is the main factor making Cycles render a very simple scene faster than LuxRender? Scene contains just a plane, emitting sphere and a white cube. Using CPU with 2 threads.

Cycles renders the same simple scene at about 1 sec/pass and LuxRender uses about 6 sec/pass. Using default settings (bidirectional + Metropolis).

Obviously LuxRender is much more complex and unbiased than cycles supporting many things Cycles can only dream of. But what are the biggest reasons for the slower speed? Is there a way to make LuxRender faster?

[Carbonflux]

Luxrender is not slow, this is a common misconception because the perception of realism does not require accurate simulation, if you had bothered to read this forum and other related forums before posting you might have noticed this has been discussed many many times, unless your goal is just to be critical. If you are getting the results you want just use Cycles.

[Paapaa]

I did not mean to offend anyone. Sorry if that sounded like that. I'm simply comparing the time it takes to get smooth results. Could you point me to threads that cover this issue? And I thought that Cycles is simply giving "good enough looking" results as opposed to truly physically based in this very simple scene. But sorry if this has been discussed many times.

[SATtva]

Besides everything else Lux is spectral. All things being equal, it will always be slower than a simple RGB renderer.

[Paapaa]

Yes, spectral rendering might be one reason. I also guess the film interface (filters, such as Mitchell) takes its toll.

[Carbonflux]

The real question is what are you going to do with all this speed? Is this just speed for speeds sake?

Someone can write a book on the complex issues related to gpu's, Cycles uses the gpu to accelerate rendering so on a single PC it seems faster, also Cycles uses CUDA which is not open source. But honestly I am tired of all the good vs bad issues, the question is are you getting the results you want? That is all that matters, if you can't tell the difference between Cycles and Lux whats the point of comparing them?

If you really need lux to be "faster" you can use network rendering, lux will scale over 100s of processors, Cycles has no network support yet iirc. for a pure gpu option you can try SLG2 ( small lux gpu. )

The real issue here is the same issue every time this comes up, its about human perception, you can make a oil painting look photoreal if you want too, your mind by default is not well trained enough to not be fooled by visual cues. To assume you are born with the ability to judge what is a fake and what is real is a mistake that is very common among artists.

If Cycles fools your eye why ask for anything more? Unless you want something more?

Lux has the integrity of physically based simulation without compromise.

You yourself stated the difference.

[Pilchard123]

Luxrender is not slow, this is a common misconception because the perception of realism does not require accurate simulation, if you had bothered to read this forum and other related forums before posting you might have noticed this has been discussed many many times, unless your goal is just to be critical. If you are getting the results you want just use Cycles.

Ouch. A little harsh, Carbonflux? All the poor guy was asking is "What exactly does Lux do that makes it slower than Cycles?". Surely that's a good thing to be asking - it means that they can make a more informed choice about which renderer to use. Yes, it may have been covered by other people, but at least it wasn't a "WAA WAA CYCLES IS FASTER AND LUX IS A LOAD OF ROTTING HORSE URINE" post.

Just out of curiosity, is there an (an? a?) FAQ post anywhere around here? A cursory glance around the site, wiki and forum doesn't seem to find one.

EDIT: Semi-ninja'd. Still feel your first reply was a little acerbic though.

[Carbonflux]

I don't mean to be harsh but at this point its hard for me to imagine anyone posting a question like this unless they are trolling, at the very least I think its fair to challenge them to test if they are trolling.

Maybe in this case its not trolling.

But the point is, Lux is not slower than Cycles.

Lux is not slow.

This is a totally invalid comparison and at first it was ok to tolerate it but I refuse to regret being suspicious of a post like this.

[Paapaa]

I'm not trolling. I have never used Blender before, I just learned about Cycles and LuxRender. Then I compared how long it takes to get similar looking images from Cycles and LuxRender in this very simple scene.

I trust you when you say LuxRender is not slower. But can you point me to a discussion which provides me the information to understand that I'm comparing apples to oranges and my whole question was invalid?

[Carbonflux]

There is no one definitive thread, its more like the death of a thousand cuts, try using the search for cycles and luxrender.

Its impressive that you got lux and cycles working so easily being new to blender

The answer is simple, luxrender is as fast as it can be for doing what it does, its open source so if anyone has a faster way of doing things they can present it.

Lux physically simulates light flow on x86 general cpus and scales in a linear fashion to any number of cpus.

One example of why your metric is invalid is that you compare lux with cycles on a single PC, as I implied above.

You used your PC with its gpu, lux is not limited to one PC so you can't really say how fast it is. You could use the Amazon cloud with lux and it would render millions of times faster than Cycles.

Also your plane and cube test scene has no meaning, try something more complex with physical materials like glass.

Anyway, I could talk about this for hours but you are missing the point so I will restate it...

Its not about speed its about appearance.

If you are happy with the results just use Cycles.