Maxwell Vs. Mental ray
 

hey everyone, I just decided to do some simple tests comparing the render times of maxwell and mental ray. Turns out maxwell is faster for alot of things, (its like turning every one of mental rays settings to full, which I didn't even want to try)
This was my first test, and I will post some others soon.
In this example, mental ray took more than twice as long. That doesn't really go along with most peoples conceptions of maxwell, does it?
http://img77.imageshack.us/img77/7673/test3ep4.jpg

another. Mental ray has GI and caustics enabled, FG, and other stuff. the only lighting is an HDR.
again, it took more than twice as long.http://img297.imageshack.us/img297/9233/testie6.jpg

damn, i gotta gets me some maxwell!

whats your pc specs?

not that good, haha.
a gig of memory, and a dual core 4200+ processor.
It cost like 650 bucks, so its not exactly top of the line ;)

The results produced from the two renderers look different, which means that it isn't a proper comparison becuase they are not doing the same thing. One of them has to be taking shortcuts somewhere, but since I rarely dabble in rendering, I don't know which one is more realistic. Which ever one is more realistic should be the winner here, not the one that is fastest.

Its like comparing a two-year-old kid drawing a tree to a trained artist drawing a tree. They are both recreating a tree, but the kid is faster than the artist because he draws a line for the trunk and a circle for the leaves up the top, unlike the artist, who is painting every little detail... ;)

They look different because each renderer has a different process that it goes about for the render. the HDR for example comes out at a different angle. I am just trying to show that maxwell isn't NEARLY as slow as everyone says.
Mental ray was at some pretty high settings, when Maxwell had its default settings (it does everything real light does)

To the best of my understanding, you can have two algorithms (I assume that is what you mean by "process"?) that produce the same result, with the only difference(s) being the amount of resources it needs (for example, RAM) and the time it takes.

Anyways, since you don't seem to be testing for which one is faster on the exact same task (identical scene/setup, identical/close-as-possible quality) , I guess my point doesn't really matter... :)

I am using the same scene set-up, the same lighting, everything is the same except for the renderer. The mental ray ones use as high a setting as I could before it became ridiculous, and maxwell used its default settings, which follow the exact properties of light. So maxwell is always at maximum settings. What I am trying to show is that, for the quality and time, maxwell is faster than mental ray.

I'm trying to find more technical info into this "Maxwell" renderer before I say anything else...

Amusingly I can't find anything detailed (maybe I'm not looking hard enough), but it seems that Maxwell uses a raytracing algorithm called "Metropolis light transport" which basically does some basic raytracing and then determines which bits are visible before doing a more detailed raytracing of the visible bits.

I couldn't find much on mental ray as well, although it is a raytracer like Maxwell, but uses a scanline method to determine visible bits (which is better I have yet to find out) but it seems that mental ray is targeted towards massive multiprocessor systems, so your dual core box might not fall into that category, which could explain the performance hit.

Quite interesting that Joopson.

What are the setting that you used with MR?

edit - Just saw the specs that you had.

I think that the large MR times might be that you have everything on, which you dont really need. I think that MR would be faster when the render's optimised, also what happens when the geo's upped to have more materials etc? That would be good to see.

As I said it would be interesting to see what results you get further down the line.

The thing is, those Maxwell renders aren't finished and still are noisy… I'd like to see how long it takes for Maxwell to get to the SL where there's no noise, then compare it to mr.
:beer:

i dont think these are real "tests"

My thoughts are summed up by what gster123, Jr.Who, vladimirjp just said. I'd like to see the difference in a more practical scene. Just arbitrarily cranking up the MR settings is just asking for it lol :p .

That is kind-of the point. I did this because maxwell ALWAYS has max settings on, so Im trying to get an equal quality render.


And yes, the second test is a little noisy, but 2 more minutes would have solved that.

one reason why the scenes don't seen the same is because they have different things as lights. In mental ray, I used area lights, but maxwell needs a polygon plane, and it cant use Mayas built in lighting system. I did try to get it as close as possible, though.
Also, did I mention that I was using photoshop while rendering with maxwell? I wasn't able to with mental ray though.
Also, what kind of scene would you suggest for the test?

Ok, I did one with more materials. same materials and lighting in each, and the only thing different is the angle of the HDR because each renderer interprets the HDR in a different way.
http://img382.imageshack.us/img382/8726/testsxd1.jpg
And before other people say it, I know mental ray doesn't need the settings that high to look the way this render does, but as I said before, maxwell automatically does, so its only fair. for all intents and purposes, this is a test to see which is faster on an average computer, For the maxwell one, if I had rendered it for just 10 more minutes, the noise you see would be gone, but I had already set it to end in 20 minutes.

I fail to understand the purpose of this test... why does it matter which one is faster?

It also appears that mental ray has better quality, at least in the latest test. The results have more depth, the shadows look more realistic and have are more dynamic and looking at my own desk and the shadows it casts (its sunset atm), mental ray's results do look somewhat similair.

Ok so I guess just out of curiosity you wanted to see which was faster on pretty much max settings rather than practical settings, problem with that is that you set a 20 min cap for Maxwell while MR is left to actually complete the image. And in this particular case your Maxwell time is probably going to end up closer to the MR time if you decide to correct the weirdness with the geometry for the drawers and get the noise completely out.

I think to accuratley test them you would have to set the mental ray scene up better. Just setting things up to the right dosn't mean it is the right setting. If the renders treat certain lighting setups differently you should compensate.

Im also not sure what it is you are trying to test? Becase if you are testing speeds i would think that the end result would have to be identical. At the moment the pictures are too different to be accurate. Showing two completley different pictures and saying one is superior because it is faster is not accurate. If you could show me two Identical Pictures and say that one is done with Mental Ray and the other is done with Maxwell then maybe the test will have a point.
At the moment it just looks like you are trying to make Maxwell look much better than Mental Ray.

Just my little opinion

The whole point of this post (which I now give up on, haha) is too disprove the myth that maxwell is amazingly slow compared to other renderers. If I had left either render for 2 more minutes, the noise would be gone too.

The problem is, maxwell bases everything on how real light interacts, but mental ray goes off of what YOU want the lighting to look like, in a way. this means it would be VERY hard to get the images to be exactly the same. The main difference between the 2 in the 3rd image is the angle and interpretation of the HDR by the renderer. Mental ray's image turned more pink because of it, but maxwells didn't. its just the way maxwell uses HDR. They were as close as possible though, with an identical lighting set-up, and the same HDR, so I don't see what the fuss is about. As I said though, everyone thinks maxwell is a waste of time because of how long it takes. I was just trying to show it isn't nearly as slow as most people think, and it could be used for animation.

-Andy

Maxwell does not behave like real light. I don't think any commercial renderer behaves like real light. Renderers however do try to act similair to real light. My understanding is that Maxwell fires off a few rays to find out which bits are solid and then guesses the color values of the everything around it before doing a bit more raytracing to find out how good its guess was. That is hardly how real light behaves. If one wants to test the "realism" of Maxwell, one should write a renderer that traces the path of every photon and its interaction with air molecules, dust particles and the surfaces of materials at a molecular level and compare the results (after a few months of rendering at a large supercomputer) with Maxwell. I'm very confident that one will find the results to be different.

OK, what I just said sounds like technobabble from Star Trek, but I hope it does give an idea of how far Maxwell is from simulating real light.


Also, I suspect the reason why mental ray is so slow is becuase it wasn't tweaked. Tweaking is very important.

I think that MR works in a similar way.

As people have said tweekning the render in MR is part and parcel of using it, If you tweeked the settings I bet you could get it faster than maxwell with very similar results on the MR render.

I would be more convinced with the tests if you did the mental ray one, and then took that same amount of time to do the maxwell render, that way they're rendering the same amount of time, then we get to decide which one looks better

Maxwell and MR are both raytracing renderers, but I think they use entirely different methods to make raytracing more effiecient and to add all those fancy effects.

Yes I know.

I was pretty sure you did, but you did say, "I think that MR works in a similar way." in a previous post...

Yes there both Raytracers. Obviously there different when it comes to calculating specific parts, but the overall theorys are similar.

joopsoon, stick with Mental ray. The longer wait is worth the effort and the renders will look better.

but, Maxwell gets better results for stuff, if I tweak it a bit. It also allows me to change the lighting when its done. Much more effective. plus you know if its gonna suck the moment it starts rendering the scene.

Haha I like the choice of words, "stuff". Well it all comes down to personal preference, if you feel its the renderer for you then by all means have at it. I just don't see myself shelling out $$ for Maxwell anytime soon when MR gets things done.

well, neither can I, my brother got it awhile back, but since then he changed to doing web-design, so he gave me his license. but yeah, my choice of words was great. What I meant was, The results I get from maxwell is more satisfactory than what I get from mental ray.

So would mental ray. The key is to Tweak it!

but with maxwell, I can tweak it afters its been rendered ^_^
I loves me my multilight:beer: