I agree almost completely, Matt. I know MR became some attention by its developers lately with the GI-Next stuff and some other things. I also knew that XSI's implementation of MR wasn't ideal and probably the reason for the instability in many cases. Whats sad, because once softimage was famous for its MR integration. But even the integration would have been flawless, smooth and perfect over all the years, the renderer itself became too old. Too much of its algorithms and core design had to be rewritten to comply with modern demands, I would think. I always liked MR for what it does, but there are definetly scenes that you cannot render in MR whatever your knowledge is. In theory MR can do it. It can handle GI, motionblur,DOF and has some neat optimisations for glossy reflections. Combine even a few of these advanced features (that's not so advanced by today) and you will reach render times far beyond reasonable. I'm talking about the latest version (1.12) in softimage.
Mental ray has serious and painstakingly correct SSS that is "physically" correct by using photon mapping and it had it long before other renderers IIRC, but was hard to use and took decent amount of time. I think no one used it in production. They came up with the fast-sss shaders and they were more usable. MR can do path tracing and other cool things. There is a lot what MR can do on paper but in production you'll need a much bigger render farm than you would need with most other renderers. For NPR stuff its quite good an I still like it for the many shaders I have available inside XSI. Just today I switched to redshift for an animation, that would have take at least 2 days on the (small, eight machine) renderfarm, because I needed 3d motion blur. Sometimes you need it. Even with the rasterizer and unified sampling (what is btw not useable in v1.12) I hit render times of 10 minutes per frame. That's not much, but if you need to finish 2500 frames until the day after tomorrow it becomes a problem. :) I re-shaded it for redshift and with a few optimisations, a frame took 20 seconds on just the workstation. Of course that's GPU versus CPU and not comparable. Nevertheless, I would not be able to finish the project by using MR in this case. As for the compartmentalisation (love you for those words btw), I wonder if this is the reason why mental ray volume shaders were never exposed inside XSIs render tree? Sven -----Original Message----- From: softimage-boun...@listproc.autodesk.com [mailto:softimage-boun...@listproc.autodesk.com] On Behalf Of Matt Lind Sent: Tuesday, November 21, 2017 5:06 PM To: softimage@listproc.autodesk.com Subject: Re: end of another era A lot of people don't know this, but mental ray received a HUGE update last year with acceleration as much as 20x for global illumination and related. The version that ships with Softimage didn't receive this update, of course. After acquiring mental ray, Nvidia let it rot for a few years before deciding they should put effort into it after all, then put significant work into it, but not until after many people went to other options. As much flak as people put towards mental ray, it's actually a very good renderer, but it requires you have knowledge of raytracing algorithms to make best use of it. The problem is most users only used the interactive version which was hamstrung by the XSI interface which created most of the problems related to crashing due to memory constraints and other issues. If you ever used mental ray from the command line on it's own, you'd know it was actually very fast and stable. If it had continued to be included as a standalone renderer after XSI v2.x, it would've been more popular as those serious about rendering would've gone to the command line and/or written their own scripted UI for it such as with Qt. Classic case of marketing ruining a product's profitability. My only complaint as a shader writer was that mental ray was overly compartmentalized which sometimes made it difficult to access parts of the scene to create comprehensive shading effects. What you could do with a single uber shader inside another renderer, required teamwork between multiple smaller shaders in mental ray. It was also not documented very well in the advanced areas. You had to really figure it out on your own. While the documentation was accurate, a lot of it didn't make sense until you already know how the renderer worked. But once you did, you had a lot of power at your finger tips. Matt Date: Tue, 21 Nov 2017 09:01:52 +0100 From: Olivier Jeannel <facialdel...@gmail.com> Subject: Re: end of another era To: Anto Matkovic <a...@matkovic.com>, "Official Softimage Users MR wasn't heaven, but it's again another widely used abandonned software. Of course the provider will never tell the user that he has stop its development, no, instead the users finds out waiting months, then years, that nothing happen or changed on the platform he thinks everything is fine. ------ 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.
