SIGGRAPH: My Final Thoughts

Posted By David Blumenfeld on July 30, 2010 12:00 am | Permalink
All great things must end, and so follows that SIGGRAPH 2010 has finally come to a close. Overall, it was a nice conference this year, with some interesting new technologies to be seen and a few new techniques and ideas to walk away with. Unfortunately, it was definitely, in my opinion, not one of the best expo floors I've been to, likely due to a few things. For one, despite a few new advances this year, there weren't many revolutionary breakthroughs in the industry. In this realm, it seemed as though the biggest push of new technology which is finally starting to just take hold are the GPU embedded graphics cards and some new real-time or near-real time software that is finally beginning to take advantage of this hardware. More on this in a bit. Second, there seemed to be quite a number less vendors on the floor showing their products. While some companies were simply absent from the show altogether, others were no longer present because they have either merged or been acquired by another vendor at the show. This industry consolidation (which seems to be a cyclical thing over the years) is to be expected, but regardless has an impact of the implied presence at the show. Finally, I'm sure the recession and economic state has a great deal to do with it as well. Between budgetary cuts and marketing restructuring, the “freebie” giveaway factor was definitely much lower this year. Aside from a few random t-shirts, pens, and the ridiculously long line for the Pixar teapot toy, gone were the squishy toys, magnets, keychains, mint boxes, and random collectible things that tend to fill up the free bags of many an expo-goer.
On the topic of the teapots, what a fantastic marketing job Pixar has done with that, and kudos to them for once again, as always, being geniuses. Being a Renderman user, my “special” limited edition teapot which is somewhere in the 400's out of 1000 was obtained at the Renderman User Group meeting. I am not, however, a collector of such fine memorabilia, so I gave it to my three-year-old son who is playing with it around the house now. Watching the smile on his face as he winds it up and runs around the house shouting “Renderman Teapot” is worth far more to me than a tin box on a shelf collecting dust. While there is nothing wrong with wanting a toy or collectible like this for free, it saddened me a bit to see how long the line of young aspiring artists was for this, as perhaps more of their interest should have been in about learning more of the new tech about this field, which is supposed to be the whole purpose of this show. Maybe some of the video card manufacturers can start giving out limited edition toys of “Jimmy The GPU”, and the rendering companies can offer “Irving The Image Based Light” to drum up a bit more interest from the student crowd.

While I had intended to possibly squeeze in a few more talks, I ended up instead spending more time on the floor. I wanted to make sure I had an opportunity to take in the new offerings and get some one on one time with the developers and their products. A friend and colleague of mine, Erik Gamache, was also presenting a half hour talk on Digital Domain's making of the Hydra character in the recent Percy Jackson film, so I decided to sit in at the Autodesk booth to check it out and support him. The presentation was very well delivered, and the work looked great. A number of old friends of mine worked on that, and to them, congratulations for another job well done. To try and get the most out of the floor time, I decided to do a quick pass through all the aisles, grabbing a brochure from any booth that was a place I wanted to visit more in-depth. Afterwards, I sat with one of my coworkers and formulated a plan of attack. In the software arena, it was interesting that many of the packages were Windows only, with a Linux port either in (slow) development or not available at all in the foreseeable future. The only reason I could think of for this is that most of these tools are targeting markets with are either more involved in games production, or doing commercial/film work using Window's based tools such as 3DS Max. Being a Linux based studio, and with many of the larger studios using this platform, I would have thought more of these new tools would be available for it. This basically leaves three options if I would like to use this software: set up a few Windows based machines (we currently do this as dual boot systems or on Linux via VMWare), try to get on a beta program for the unreleased Linux versions, or simply lose the ability to take advantage of these new tools. There were definitely some interested items out there, and while I won't discuss any particular toolset at this time, suffice it to say that many of the products are definitely maturing to a point where they are becoming particularly good at what they do.

The other thing of interest, as I mentioned above, was the GPU technology and some of the software taking advantage of it. I am still a bit under-educated in how this processor works, and my attempts to get some explanation at the chipmaker booths yielded me with little more info. I simply got the runaround and was handed from rep to rep, only to finally give up when the fifth person on the hand-off chain was “back in an hour”. However, what I do know about these vector processors is that they are extremely efficient at processing many of the same type of calculation simultaneously, and this so happens to fit the requirements of raytracing. Therefore, this was the big sell, and there are a number of new software tools available taking advantage of this. Trying to figure out how this could be of immediate significant benefit to our studio was the challenge. First off, being in the commercial production and design business, we have no real requirements for real-time performance of visual display. Of course, it goes without saying that if I could obtain the same desired result in an image that I get with a software render out of a real-time hardware render, I would be all for it. However, there are certain types of rendering results that, for the time being, can still only be calculated using a software render, and until that changes, that is what will have to suffice for final images. While many of the GPU images (whether real-time or near real-time) were very impressive and beautiful, they are still not photoreal, and if that's the requirement of the spot you are making, this will not work. Of course, if the result approximates the software render in type (in other words, it is a valid representation of your shading setup), then it could serve as a huge benefit during the process leading up to the final software render. Of course, there are some renderers now which are taking better advantage of the GPU, such as VRayRT, basically being used to aid in certain parts of the render while other parts are rendered on the CPU. I am looking into those renderers as well to possibly augment our current setup. One though I had and would like to look into is potentially using the GPU to accelerate the creation of point cloud and voxel cloud data for things like occlusion, reflection, etc., and then use that quickly generated cache as an input on the software render portion, greatly speeding up the overall render time. In order for this to work, Renderman will need to be able to process those portions using the GPU (not sure if the new version will allow us to do that or not), and of course I will need to make sure we are using machines which have GPUs on the graphics cards. This also means that, unless we place graphics cards into our racks on the farm, these caches will have to be generated on our local workstations, which isn't the most efficient way to work either. Of course, I will need to do a fair amount of research on this to find out how to best leverage this technology and with what combination of hardware and software. There are a few other products out now which are GPU based and interface with Renderman, such as MachStudio Pro 2, so this may provide this type of solution, along with many of the new features in Renderman Studio 3 and Renderman Pro Server 16 due out in two months.
The rest of the things I saw at the show weren't really my cup of tea. There were a number of 3D scanners being presented, but unfortunately nothing that was really a one-stop solution. By that, I mean that I would like to see a scanner which can be operated handheld, with range adjustments to accommodate close, small subjects, as well as large, mid range subjects such as a full body person or car, and the ability to further scan far large structures via LIDAR, while at the same time taking high resolution (at least above 10 megapixel) full color images for texture purposes, all in a single wireless unit. Additionally, it would be nice for the price point to be below the five thousand dollar range for a unit like this. The same goes for 3D printing. The technology seems to be improving in some of these units, but the entry price point seems to be around ten thousand, extended up to the half million mark. As I mentioned in a previous post, the only way to get below this is to use a unit like the MakerBot where you assembly your own tool, and then have to work hard to get results which approach some of these pre-assembled units. While extremely large facilities and research institutes may be able to afford these units, smaller studios like ours find it difficult to justify the expense of these units when we simply wouldn't get all that much benefit from them, and when we do occasionally need to use them, we can just pay for the one-time service from a provider. I personally would love to bring this technology in-house, as it would save time to be able to perform these operations on set and in the studio at will. Hopefully, we see a price evolution that mirrors the other, longer established hardware, in a short amount of time, opening up this market a bit more. 

A funny thing happened later in the day. Not paying much attention, I didn't realize when the show ended and thought it would go until at least 6:00pm. I hadn't eaten any lunch, so around 2:30pm we decided to walk over to the LA Live area and try one of the new restaurants. That was quite enjoyable, and we ended up coming back around 4:15pm. On the way, I ran into some more friends, chatted briefly, then headed over to the expo floor again. I was literally shocked when I got there and the place was already half dismantled, the lights full on and multitudes on workers disassembling scaffolding and crating up equipment. I didn't realize the show ended at 4:00pm, and was simply amazed at the efficiency of the convention center staff and how rapidly they were putting the show to bed. We then headed over for one last talk which was still going, and finally called it quits just before 6:00pm.

Now back in the studio, I have my work cut out for me. I have this giant list in my head of things to go over with the other artists, research and development I want to get started on, software and hardware to look into, and a multitude of workflow improvements I think we can pursue. I took a great deal away from this year's show, and hopefully you the reader had a similar opportunity and experience. I hope you enjoyed my recap of the events and my various random thoughts on things, and I would welcome any comments or questions any of you might have at my email address listed in my profile. Until next time, good luck with your endeavors and the pursuit of the magical art that is animation and visual effects. This field is what we all make of it, and frankly it's one of the coolest things to be involved with.