I have to admit, since MR is a old renderer it has its flaws. I work with it since 1999 and I would say I know it quite well, even I never programmed shaders for it. One issue is DTR/ distributed tile rendering, it simply doesn't work reliable. Mental images/nvidia solved the memory overflow problem some versions ago (slave machines ran out of memory and crashed), but there still seems to be a problem when using satellite rendering together with BSP2. In some cases, the rendering hangs at 100%, not written to disk and you have to kill the process. It seems to not happen with BSP1 but since it's much slower especially when using displacement, it's not an option to use BSP1 just because of having satellite rendering to work. This happens with the render region as well as with batch rendering. I cannot say if this is entirely a MR bug or a softimage-MR thing or maybe a windows network issue.
However, it works with batch rendering when not using more than one satellite machine. The render farm here is set up this way (I'm probably one of ten ppl in the world who uses mental ray satellite rendering on a farm). It would be nice though to have rock solid distributed rendering when dealing with large print resolution for example. I don’t think there is an internal limit in MR regarding the amount of machines. Of course its not unlimited, at some point Amdahl's law will kick in. The limitation of using five machines was simply a business/licensing decision, I guess. Imagine, beeing at a company with a larger render farm that artists can use interactivly. Would cause a bit of network traffic, yes but the workflow would be improved. It's even practical if the farm is rendering regular render jobs simultaniously. I once did some tests and the raysat.exe will use a higher priority than the batch.exe by default. There would be only peaks of CPU utilization on the farm when an artist draws a render region, so the delay for normal renderjobs would be negligible. A potential problem would be, that artists, having vast amount of render power, they'll put shitload of stuff in the scenes thats simply not renderable in the final animation. But I disgress. Another issue probably is GI for animations, as it was discussed many times over the years. Even you can have flicker free GI in animations, it depends heavily on the scenario. The classroom scene, for example. Seems an easy scene,no? Well it's not. Add a skylight system and animate the sun from sunrise to sundown. I did many, many tests with FG and even with irradiance particles. At sundown you will have only a few, very bright spots that have to lit the entire room. Therefore you need an insane high amount of FG rays to capture the light an even then it will produce splotches in the last few frames. IP adresses this problem by analyzing the screen space, firing more IP rays towards bright spots. Similar to portal lights. It is great in terms of general light distribution and does not produce light leaks. But since it's also screenspace dependent, it will create problems where a lit surface is not directly seen by the camera. In this case, at certain areas at the window frames at sunrise. Btw. while testing Redshift with the classroom scene I noticed a bug when using caustics. Nothings perfect. (However, the GI solution was blazingly fast and super clean). All in all, I use MR in production successfully and I like it, but I would not say I love it :) sven From: softimage-boun...@listproc.autodesk.com [mailto:softimage-boun...@listproc.autodesk.com] On Behalf Of Mirko Jankovic Sent: Saturday, May 28, 2016 9:25 AM To: softimage@listproc.autodesk.com Subject: Re: Anybody still using mental ray? 1. I don't want to throw away a expensive CPU based renderfarm. It will be obsolete soon anyway so just make good planing for next render tool, CPU or GPU road. 2. mental ray still has more shaders. It's slow sometimes but it is rather good when you want to do non physically based things. A bit like Softimage itself, the all-purpose, swiss knife. First time I heard something like this for MRay. Mostly it is in line I wanna puke or quit 3d completely due to rendering part :) When I discovered Arnold and then even more Redshift that is when a big issue I had with SI, ie rendering was solved and 3d was fun again. MRay was PAIN non stop! Swiss knife.. yea could say that, got ton of things but nothing good for most of things :) Redshift saved my 3d ;) On Sat, May 28, 2016 at 1:16 AM, Sven Constable <sixsi_l...@imagefront.de> wrote: I use mental ray almost exclusively for any project so far, even I'm evaluated arnold, redshift and maxwell. Maxwell is so far the most accurate renderer I've seen so far in terms of light distribution in a scene and nothing comes close to it in my opinion. It's amazing if you do product rendering only. But it lacks in shader variety, softimage integration and general tweaking, as mental ray has and allows. Arnold may be a killer for big projects, heavy scenes but its expensive. Redshift is affordable,has good GI for animations and the best integration in Softimage besides mental ray (I'm still wondering how they managed to get the round corners shader into RS! I was thinking it's a mental images/NVIDIA patent) There are only two things that stopped me switching from mr to RS: 1. I don't want to throw away a expensive CPU based renderfarm. 2. mental ray still has more shaders. It's slow sometimes but it is rather good when you want to do non physically based things. A bit like Softimage itself, the all-purpose, swiss knife. So yes, I use mental ray. sven -----Original Message----- From: softimage-boun...@listproc.autodesk.com [mailto:softimage-boun...@listproc.autodesk.com] On Behalf Of Matt Lind Sent: Saturday, May 28, 2016 12:28 AM To: softimage@listproc.autodesk.com Subject: Anybody still using mental ray? While on the subject of nostalgia and recent release of various tools for XSI. I have a selection of mental ray shaders I wrote long ago still perfectly valid as they're utility nodes for accessing renderer preferences, lights, performing math, or other basic features missing from the native shader library. Some have unique features, but also limitations due to mental ray's architecture. If released, would anybody actually use them other than for tinkering? As in, does anybody still use mental ray in a serious production context where you'd benefit from such shaders? Don't say yes because you want free digital swag. Matt ------ Softimage Mailing List. To unsubscribe, send a mail to softimage-requ...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm. ------ Softimage Mailing List. To unsubscribe, send a mail to softimage-requ...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm. -- Mirko Jankovic http://www.cgfolio.com/mirko-jankovic Need to find freelancers fast? www.cgfolio.com Need some help with rendering an Redshift project? <http://www.gpuoven.com/> http://www.gpuoven.com/
------ Softimage Mailing List. To unsubscribe, send a mail to softimage-requ...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.
