But that test was with version 1.0, and there were some bugs, even the guys at pixar asked me to send the scene, so they could check what was wrong.I didnt tested this new version, but i think that they have resolved this problem.My opinion, is that if you work with scenes that need a fast and accurate GI solution, theres no other render as powerful and fast as Mentalray.I agree that work with displacement maps and motion blur in mentalray for maya, can be slower than in Renderman.But things are changing, and i hope that with next release of Mentalray things will be better, and make Mentalray, The Render Engine!
Well, thats just my opinion.
hey wizzackr, yeah, the rasterizer dont work so well with FG, i know… probably AO is a much better option in this cases… u can also separate your FG pass and make it with the normal scanline, some times ago Francesca Luce posted a network to do this kind of things, but now is not available, she says that we have to wait a tutorial… so, just wait
Nice way of getting the tone down to have a decent discussion :hmm:
I personally think it looks pretty good / comparable - hard to tell with so many watermarks
what are your settings? what artifacts are you getting? I think it is pretty painless to use and produces really smooth results with a visibility sampling of 4+
Plus it is not cheap 2D AFAIK:
[I]"rasterizer: Its primary difference to regular scanline rendering is its separation of sampling and sample compositing (also called sample collection). Without the rasterizer, mental ray selects spatial and temporal sample points (eye rays) on the image plane in the shutter interval. If an object moves, it will appear in multiple samples at different points in time. Since rendering time is roughly proportional to the number of samples, it rises quickly if an object moves quickly.
The rasterizer works by sampling all objects at a fixed time. If the object moves, these sample results are cached and re-used at every point the object passes over. The cache is tied to the geometry:
* a number of sample points is selected on each triangle based on the visibility, size, and orientation of the triangle. These points are then individually shaded by calling the material and volume shaders as usual. The results are stored. Care is taken not to shade points hidden behind other geometry.
* The image plane is scanned, and all cached shading results are composited to pixels. If an object moves, its shading results are simply used multiple times instead of re-shading it."
I see it more as the same sarcasm dagon1978 used in the sentence I am referring to. No harsh feeling or anything.
Btw. dagon1978 got me right:
Yes the renderings above look good, definitely. I was referring to artifacts I am getting whenever I try to use Rapid Motion Blur, which also was the reason for all the heated discussions. I got artifacts with it and therefor didn’t use it for the test - as I was interested in rendertimes for good quality images. Anyway, no need to start all over again.
You’re right. For the most part it’s the same technique as the Pixar motion blur. There is one exception that I can think of.
RfM has a checkbox that is on by default called Ray Traced Motion Blur, which blurs moving objects that are seen in reflections. The MR method does not do that. I think if you uncheck that box, they will be roughly the same.
Dragon, how on earth did you get those results with MR?
I did the exact same test with the same scene and I got 1:01 secs with RfM, and 8:59 secs for MR! Keep in mind that’s a single threaded render with RfM and a multi-threaded render for MR. (I have a dual xeon machine). Resolution was 640x480.
I did very very little to the default scene, however. I did not add any approximation nodes, all I did was set the render global defaults to RapidProductionBlur, turned off raytracing, and set rapid blur settings to 4 visibility samples and a shading quality of 1. Nothing else was done to the scene, what other adjustments did you make to the scene to get render times from almost 9 minutes to under 2 mins?
At 2 visibity samples, render times were cut to 3:55 secs, but the blur quality doesnt match renderman’s at this setting.
Sorry, I would upload images, but i have no hosted space at my disposal at the moment. Hopefully this is enough info for you to work with.
Okay, it is great to see the conversation arriving at this level. A few pages back it was one big rant. In the end, we all want to see the renderers we use give great results, fast, so it is so much more constructive to do some trouble-shooting, rather than render one or two random scenes without much tuning and declaring a winner.
Anyway, if you or others could post Maya scene files or mi files, then others could have a look and see which settings might be at fault. If you export to an mi file, you could also just post the options block here, perhaps 20-30 lines of text, and others might have a chance at figuring out why you don’t get the performance.
The RenderMan standard and the prman renderer have been around for a very long time, and it is very well tuned. It is also relatively easy to use at the hobbyist level, but if you are in production at a high level, it can also be very difficult to get the results you need, even with prman, just as with mental ray, or any other renderer. It is hard to sit down and make a test scene which demonstrates this properly.
Some people prefer prman, partly because it is a known quantity, which has gathered a large body of knowledge about how to render most efficiently.
mental ray is much newer, at least at the level it exists today, ie. in most major animation packages, and getting into more all the time. If it was as bad as some people in this thread have claimed, it would simply not be in all these packages. But it is a more complex product to tune, and there isn’t so much general knowledge out there. It is probably also more sensitive to correct settings, so new users will sometimes see great performance, sometimes less great performance. Btw, there is an increasing amount of mental ray knowledge being accumulated at the http://www.lamrug.org/ website. Lots of good reading about tuning and how things work.
Anyway, if some people who are seeing poor performance with mental ray are willing to post scene files, others, including myself, can look at them, and make suggestions. There is clearly a lot to learn for many, since ray is so complex, but it could easily be knowledge worth accumulating. For one thing, you can save $1000 right off the bat by learning the renderer you already have, rather than buying a new one.
About the artifacts in some rasterizer images: there were certain bugs in earlier versions, especially the version in Maya 6, but also to some extent in Maya 6.5, where large chunks of it were rewritten. Maya 7 has a much better version, and Maya 8 will have a better version still, including speed improvements. Many bugs have been fixed. Keep in mind that the rasterizer is much newer than most of mental ray, and has had to mature. It is unfortunate that bugs slip through, but any long-time user of prman (eye-splits anyone) will know that problems appear in all products, not just mental ray, and don’t necessarily get fixed fast.
Anyway, post those scenes, so this thread can go from useless bashing to learning.
Dagon1978 Can I use two processors/cores on a dual processor/core machine?
Yes and no. RenderMan for Maya will support a muti-process mode of operation. Presumably all Maya plug-in renderers support this mode, but Maya doesn’t make it trivial to utilize this mode…
As I said in my post I wasn’t trying to correct you I just found your first statement to not be clear and possibly misleading. This information was much more clear and concise then the original message on multithreading with RfM.
The scene in question is the motion_blur.ma tutorial that you can download with the Renderman for Maya evaluation from Pixar’s web site. Somehow Dragon was able to get obscenely good render times from Mental Ray compared to Renderman and I’m simply dying to know how he did it, because I can’t duplicate his results.
NOTE: My tests only modifed the render globals. No changes were made to the scene or geometry, which is possibly how he was able to reduce render times. But right now Im simply guessing, as Im not sure how adding approximations to a relatively low poly model would help render times significantly in this case.
I tried to tune to the extend of my knowledge of BOTH products. And I did never declare one a winner. It’s a pity people still haven’t read any of the original statements.
I will try to do that, but cannot make a promise. Tomorrow is my last day before I go on a three weeks vacation, so there is a lot to wrap up.
I don’t recall anyone in this thread calling MentalRay a bad renderer. It is a great product, but not the cure-all rendering solution. Nothing more, nothing less. Same goes for any renderer out there (yes, including Renderman).
Yep, that is a known problem. But we still have 6.5 at work, so that is what I tested against. And I mentioned it in my second sentence, so I would not consider this an unfair challenge as the testing conditions were known. Not ideal, but known.
Yep, it started out so nice and shiny and I was already hoping that it might stay that way. Well, maybe it will stay more civilized now.
I’ve been using mental ray in production for a couple of years now, and I’m putting renderman for Maya through its paces on a home project. So I’ve been able to put them both through real world non-test situations (read non-bullsh*t scenes). My biggest concern with renderers is not how well they produce a good still but in how well they produce animated frames. I could careless about still frames. I am not a guru of either renderer but a generalist who needs to get the best performance I can.
The number one determining factor on getting great images with either renderer is talent. Enough said on that.
What I have found is pretty much what others are saying here. Mental ray is a great renderer but not artist friendly. A ton of time will be spent optomising renders and fiddling with settings or surface/displacement approximations all of which may or may not drastically affect performance/ results. It really is a renderer for TD’s who can optomise it out of the ying yang and use the full range of custom MR shaders to create awesome but confusing as hell shaders.
Renderman for Maya on the otherhand is dirt simple and gives great results at the cost of being able to really get under the hood. Saying that I have seen where there are lots of opportunities to get under the hood that are hidden away from the average user, this could be my niavety of using the renderer saying this. It seems to have been designed for animation as its first priority (go figure from a product by Pixar). I particularly like how well it handles all types of geometry (nurbs, SuB-D, poly and has its own built in poly sub-D algorythem). But it is a version one render with plenty of little gnatts and bugs to get out of the system. The basics are all good but when things get a bit more complicated, things start to not work so well together. This is typical of version 1 software and I am actually very suprised that RFM is as far along as it is. They also have a great system of tech support going for it (also more comon with a v. 1). I predict within a version or two it will be a very robust production ready renderer. So far I am rather fond of this piece of software even if I have been tempted to kick it to the curve every now and then.
My verdict is that renderman for maya will be a valuble renderer for the small shops, boutiques and freelancers. It’ll end up paying for itself for the average user in terms of added productivity and allowing the artist to devote more time to the work and less to fiddling. For the TD/MR guru’s, it would be a waste of money as they could acheive the same results and prossibly more (depending on the task) with Mental Ray at no additional cost.