Most of your information is incorrect. First, I never said nor implied "you're not smart enough to use software X" mentality. I said most users do not take the time to learn the software to use it properly. That has nothing to do with being smart or dumb. It's about investing time to learn so you apply the principles correctly in context. You can be the smartest person ever to walk the face of the earth, but if you don't take the time to learn how your tools work, you should expect many failures: http://www.dailyreckoning.com.au/images/dr20120709.jpg
Majority of artists I've dealt with in production have never once opened the manuals to learn how to use the renderer. They rely purely on intuition flipping switches that sound cool, then shove the scene off to be rendered. When it fails to render, they complain and blame the renderer. Let's give an actual example: Many years ago I joined a feature film when it was already well beyond 50% through production. they had a lot of problems with rendering trees. Scenes would get pushed off to the renderer, but any scene with more than 3 trees would crash taking the computer down. For several meetings there were grumblings how mental ray should be ditched, so I was assigned the task of researching the issue. Inside of Softimage the scenes looked fairly simple and was very responsive to the mouse and timeline. A street with trees lining the sides of the road. No more than 6 or 7 trees in view of the camera at any one time. The tree trunks only had resolution of roughly 20K - 180K triangles depending on the tree, and each tree trunk had one or two bitmap textures at resolution of 2048 x 2048 pixels in 16 bit color. The scene as a whole had 30 bitmap textures, if that. I inspected the shaders and didn't see anything other than standard phong, lambert, and color correction nodes in the various rendertrees. No exotic features activated either. Hmm, I thought. So I dumped to .mi2 files and rendered using the mental ray standalone in heavy verbose mode to get more information what mental ray was actually doing. Before it crashed, the log indicated there were more than 20 billion triangles in the scene exhausting available RAM. That's right, 20 billion - and the scene was still in the process of loading. Naturally I'm scratching my head as to the source. So I reopened the scene and took another look and that's when I discovered the leaves of the trees were created with particle instancing. That is, the artist created a single clump of leaves in very high detail (~15,000 triangles for a handful of leaves), and instanced them around to every branch of every tree. When mental ray loaded the scene, it had to dynamically allocate the memory to hold each instanced set of leaves. Since each tree had hundreds/thousands of mounting points for the clump of instanced leaves, the instancing inflated the triangle count to 20 billion+ exhausting available memory causing the renderer to crash. If the artist had used a different approach to setting up the scene, such as using a delay load geometry shader or modeling the leaves directly onto the trees, mental ray would in turn use a different strategy in allocating memory to render the scene and perhaps not crash. This is my point about people needing to take the time to learn the renderer and the rendering process, and stop blaming the renderer. 99% of artists don't do that. So when they complain about mental ray in these contexts, what they're really doing is shouting out their own lack of preparation. Again, it has nothing to do with being smart or dumb, but it has everything to do with being prepared and responsibility of learning tools of your craft. Not sure of their present release schedule as I haven't kept up in the past couple of years, but when I was more involved, Mental Images was in habit of releasing patches and updates quite regularly. Often every few weeks. What you have understand is the business relationship in how you access mental ray rendering. Mental Images licenses their technology to other businesses, like Autodesk. Autodesk is effectively the customer and the one receiving the regular maintenance, patches, and support from mental images. Autodesk in turn then integrates the rendering technology into their DCC products and extends their own technical support to their customers. It is Autodesk who decides to only update their integration of mental ray once or twice per year, not mental images. So direct your complaint to Autodesk on that front. This arrangement is also why you get integrated mental ray and not standalone mental ray. That was likely a cost driven decision as standalone licenses would cost more (read that as, having standalone licenses in addition to integrated rendering would cost more). Softimage|3D was the last application to offer both. In this scenario, Autodesk is like your local reseller - they're intended to solve your problems, but actually just get in the way. If you want better support per your complaints, you can purchase directly from mental images and get mental ray standalone and more frequent updates. It's the same rendering technology, but unhindered by the overhead of the DCC which translates to faster performance and more stability - especially at load time when integrations suffer the most. Your DCC also only exposes options to rendering features which the DCC supports, but Mental Ray has many additional features beyond that - some of which would solve problems many complain about. All those check boxes, sliders, and menus you access in your DCC are command line flags you can activate with the mental ray standalone, but mental ray has additional flags. I have used both integrated and standalone rendering with mental ray quite a bit over the years, and have also written a considerable number of shaders. standalone is by far more stable, faster, easier to debug/troubleshoot, and scales very well. No it's not perfect, but for pure rendering it beats integrated rendering hands down. A lot of that is due to the integration, not the renderer itself. So about your whole 'user base is struggling' comment.... that's back to my earlier point of not taking the time to learn. when I started using mental ray back in the 1990s, I was just an artist/animator. I didn't know how to code. I was working in games and had to make cinematic sequences for a game called "SnowCrash" that never made it to market. I had to do a lot of futuristic stuff on a budget and thought mental ray would be a good medium for generating special effects with the OZ shaders and rendermap. Unfortunately, I didn't know how to use mental ray and the Softimage|3D documentation was less than useful as Softimage's concept of materials, lighting, and other techniques were often backwards. That's when I cracked open the mental images written documentation for mental ray and began reading. While some of the documentation was terse or organized in a way I didn't exactly consider user friendly, the programming documentation was very logical, straightforward and to the point which actually made it all make sense. The coded examples were very well written for the purpose of being informative how the renderer actually works. that in turn gave me insight how to use mental ray inside of Softimage|3D, and inspired me to learn to code so I could write shaders to enhance my artistic experiences. that effort to learn how the renderer worked paid dividends later in my career and prompted me to pursue a computer science degree. RedShift is in a different league of renderer, being hardware based, and is justifiable in the context of this discussion as it offers a very significant benefit over mental ray. My main argument is refuting the idea of using a competing renderer n the same class at additional cost when it has negligible advantages in the holistic context such as Arnold or 3Delight. Sure each renderer has it's pros and cons, but to spend money on them and claim they're easier to learn when you haven't taken the time to learn the renderer you already have and paid for - that's not a solid argument. I'm not saying there aren't situations where a move to another renderer is warranted. Matt Date: Sun, 29 May 2016 06:35:48 -0700 From: Derek Jenson <derekjen...@hotmail.com> Subject: RE: Anybody still using mental ray? I think the biggest problem with the stability of MR was with the concept of only releasing a single update once a year which was tided to the 3D program. That was unrealistic idealism. There was also pressure to give customers the lastest and least tested version of MR with each yearly DCC update. 3D is too bleeding edge for that release model to be stable. Being XSI's only renderer option for a long time, stability certainly became an issue. If MR updates were released with the frequency (and flexibility of rollbacks) like all 3rd party engines, everyone would have fonder memories of the software. The developers of MR also worked in complete isolation with regard to communication with their customer base. The RS guys have bent over backward to educate and update their clients, and I really appreciate the support. IMO, you can only partially point the finger at users for not using a software as intended. With information/training being so easily accessible now the "you're not smart enough to use software X" mentally of the early years is void. If a whole user base is struggling with a technology... then something with that tech is flawed; not the other way around. The flexibility of MR and 3delight are unmatched (in XSI), but the speed demands forced on this biz make Redshift indispensable for keeping pace. ------ Softimage Mailing List. To unsubscribe, send a mail to softimage-requ...@listproc.autodesk.com with "unsubscribe" in the subject, and reply to confirm.