LAmrUG

Los Angeles mental ray^® User Group

Agenda

7:00-7:30PM Registration/Refreshments/Raffle Entry
7:30-7:45PM Introduction
7:45-8:45PM XSI Presentation
8:45-9:30PM Panel Discussion
9:30-10:00PM Open Discussion

Notes

While everyone was showing up for this meeting, we enjoyed the snacks and refreshments, which included a healthy veggie tray, in addition to beer and pizza. This was also the time attendees put their business cards in for the mental ray^® Handbook series raffle. A good group of 80+ people showed and filled the room into which we moved for the presentation.

The introduction started with a brief explanation from Bart Gawboy about LAmrUG, as a newly formed community group, providing local support to those interested in mental ray rendering technology regardless of which 3D package is used. Yet, at the same time, we're dedicated to exploring deep into each package with a render-centric point of view. So, with that in mind, we thought Softimage appropriate for our first vendor-focused meeting, given their long standing relationship with mental images^®.

We also introduced Rolf Herken, President and Director of R&D from mental images, fresh off receiving the Academy Award for technical achievement on Saturday, and Marc Stevens, Director of Development from Softimage. In addition, Bob Shafron from the local Softimage office acknowledged Ray Feeney and RFX for providing help setting up and a great place to have the meeting. [Note of thanks also to Bob, who did a lot of the work getting this meeting together, and getting the word out.]

Marc introduced our main presenter, Olivier Ozoux, Director of Special Projects, who is based in the local Santa Monica office. Olivier had prepared an excellent power point presentation, which depicted the high level of mental ray integration into Softimage. He noted that by using raylib and incorporating mental ray native data structures, XSI reduced the duplication of 3D data, which naturally occurs when communicating to a renderer. It also provides XSI with a highly interactive rendering environment.

Using a running XSI, he showed how XSI provides a transparent GUI to the mental ray rendering interface, exposing much of the functionality to artists. He also showed how one could create Phenomena by combining shaders together in the GUI and writing them out. We also talked a bit about taking advantage of using multiple processors for rendering, and the many ways in which a system can be setup for interactive as well as batch rendering. For more details on this, here is an interesting white paper titled SOFTIMAGE|XSI Processing & Rendering Solutions.

As Oliver wound down his presentation, more questions were being asked and we eventually took a slight break to move into the panel session. Some attendees went to scarf up the remaining refreshments and to socialize, while others remained to ask more in-depth questions. Please send Bart salient factoids to add to the list, but these are some of the tidbits I remember ...

• The version of mental ray supported by XSI is 3.1. When 3.2 comes out of beta, it will be supported in a timely manner, and there should be no transition issues w.r.t. 3.1 shader compatibility.
• There are more than 250 shaders. When connecting shaders together, the input and output types are color coded to aid in proper connection. XSI tries to insert the correct adapter if it can, eg., scalar to color conversion, but otherwise will not allow mismatches.
• The XSI GUI connects to mental ray's diagnostic modes, so one can visualize everything from pixel coverage based on samples settings to BSP structure. I assume this would also be quite useful in tuning global illumination and final gathering by observing photon distribution.
• In the current release, to speed up per frame rendering across a network, 10 machines per frame seem optimum. Those systems can be multi-processor systems themselves, so beef'em up to 8 procs each if you really want to crank.
• Rather than trying to temporally filter Finalgather results, try using higher numbers of photons. Jim R. states he uses 4 million photons and a 2000 finalgather accuracy.
• As graphics hardware acceleration becomes more accessible (eg, via OpenGL, cG), mental ray 3.3+ will take advantage of whatever you get in a seamless manner. You should be able to write custom shaders which can take advantage of the hardware.
• Softimage will provide us links to subsurface scattering (fake and not) shaders soon.

Finally, we held the raffle drawing, and then broke up into smaller groups trading info, while finishing off the refreshments. Time flew by fast, so we must have had a good time!