NVISION 08

Last month, NVIDIA hosted their first visual computing conference in San Jose, California. It was a three-day celebration of the power of the graphics processing unit (GPU) and all it can do. There were tracks for gamers, investors, developers, and interesting performances and eye-candy for the general public and lookieloos. There was a trade-show floor, and a couple of razzle-dazzle keynote sessions. And, the key message from NVIDIA seemed to be that their main product, the GPU, is evolving from "graphics engine" to "compute engine" and that their impact is expanding from cool graphics to life-changing computation. Of course, this would be a good thing for NVIDIA given their status as "last-man-standing" in the dedicated graphics chip business, but there are reasons to believe they can pull off this transformation, and I'll touch on some below.

Day One -- They kicked off NVISION 08 by inviting gamers from around the world to participate in a Guinness World Record attempt to sustain the longest LAN party (40 hours and 15 minutes) ever held. Hundreds heard the call, and filled darkened cavernous rooms at the San Jose Convention Center with their PCs and bodies to game away for days on end. Some of the world's best gaming teams came as this was also the stage for the super-bowl of video gaming, the Electronic Sports World Cup. The environment was buzzing, and occasionally near riotous as team members would shout out instructions to others on their team who were about to be ambushed and slaughtered in a virtual world.

Day one also included a keynote from NVIDIA's CEO Jen-Hsun Huang in which he laid out many examples of how NVIDIA technology is going beyond creating fantastic visuals, to also enabling transformational problem solving through ubiquitous massively parallel computing. He led off with describing foldingathome which is a down-loadable software application that helps scientists find out how proteins fold, and in turn, help understand and find cures for critical diseases. By using GPUs, the process runs many times faster, potentially speeding medical advances. And, when you multiply those gains by millions of NVIDIA GPUs already installed in PC's, the impact is potentially huge. He also spoke to more traditional examples such as virtual product prototyping, digital image processing, and mapping applications such as Google Earth.

Perhaps the most entertaining part of his talk though, was about virtual worlds. He drove the point home with an on-screen avatar of himself that was decidedly hip -- making fantastic break-dancing moves, and apparently winning the affections of at least one attractive avatar lady friend. It's less clear how this is a revolutionary application of graphics processing power that will improve lives, but it was fun eye-candy, and kept the crowd entertained.

Road-Mapping Technology -- After the keynote, I attended some of the technical sessions, and took away a very high level view of trends in this space -- namely that graphics processors are improving faster than Moore's Law which predicts 2x performance every 18-12 months. There are also three different levels of performance that track each other by a lag of approximately 5 years. The highest level of performance is what you see in "cinematics" which are the super-high-end graphics that go into feature films and the promotional videos for new games. These are generated in days, and played in minutes but look stunning. The next level of performance is real-time video-game graphics. What's interesting is that even though it doesn't look as good as the cinematics, it will in about 5 years. Just imagine game play that looks like as good as today's movies! The last level of performance is graphics for mobile devices, and this lags desktop games by another 5 years, so 5 years from now, we'll be seeing immersive high-end graphics on our cell phones.

There were also a range of sessions dedicated to new companies in the graphics space, or companies using their graphics processors for interesting compute problems. These sessions were sometimes inspiring, sometimes bewildering, but most often interesting and engaging. I love to see people trying to create something new, and these sessions had creativity and collaboration throughout.

On the whole, NVIDIA did a fantastic job for their first year, and I'm sure they learned a lot. I think they got the message out that "it's not just about graphics anymore." They've staked claim to a very broad space, and have a lead in pursuing it. It will be interesting to see how things evolve, and where they are in a year when I assume they'll be doing NVISION 09.

For their grand finale, they brought in the Mythbusters and did an amazing demo, but I'll put that in the next post along with some thoughts about what I took away regarding NVIDIA's long-term prospects.

For a complete set of photos of the event, click here.

Remote DVR Decision

Yesterday the 2nd Circuit Court of Appeals announced a finding that a lower court erred in ruling that Cablevision's plans to introduce a remote-storage DVR system would violate copyright laws. This potentially opens the door to cable, telco and Internet video service providers offering virtual-PVR functionality without having to place and manage set top boxes with hard drives in the home. The decision will most probably be appealed, but if it stands, it will have dramatic implications.

One of the obvious implications is that it further de-values the already challenged business model of pay-per-view video on-demand since it makes large quantities of free, time-shifted content available on-demand. Though this isn't a major revenue stream for studios, it's something they'd rather not lose.

Another implication is that the trend toward time-shifted viewing will accelerate from its already rapid growth. According to e-Marketer, between May of 2007 and May of 2008, average time spent watching time-shifted programming per month among unique US users grew from 3 hours and 44 minutes to 5 hours and 50 minutes -- an increase of 56%. As this grows, the significance of scheduled programming time slots, and the value of in-line advertising decline, while the value of carousel broadcasting of unique shows, and non-traditional advertising (e.g. product placements) will grow.

It will be interesting to see how rapidly Cablevision moves to take advantage of this decision, and how fast the content owners move to challenge it. In time, I expect more and more content will be served directly from content aggregators to end-users through the Internet or on-demand channels from service providers, but for many years, the PVR model will dominate. One can only imagine how tempting it will be, if this decision stands, for the virtual PVR provider to argue that they only need to keep one copy of every show they offer time-shifted, and how adamantly the content owners will argue that doing anything more than offering the end-user a virtual space to shift in would violate their rights. It's a fascinating time.