NVIDIA: How to Do an Epic Keynote

    Slide Show

    Top 10 IT Performance Management Predictions for 2014

    One of the things that generally has me scratching my head is how technology company CEOs will get up and do a keynote without any effort to showcase their technology. Even though they have the capability to create an epic event, or to fill their keynote with stories and examples of the things their customers are doing, they often choose to instead have static slides and talk technology plumbing. IBM stands out as a better example in that it fills its presentations with customer testimonials and examples. But what makes NVIDIA more interesting is that it is a graphics company and its keynote is typically full of powerful graphics and images that drive home the point that they create magic. This year was no exception. The high-energy opening video was visually powerful, with examples of how GPUs are used from things ranging from advanced analytics to coupling the brain to a system and using games to correct mental problems.

    Start off Right

    Often, CEOs start with financial performance, which comes across as an effort to pat themselves on the back for work well done. CEOs are measured by financial performance, but even in an audience of financial analysts, this isn’t very interesting. They already know these numbers and technical folks don’t really care that much. NVIDIA opens with an announcement of a new technology, in this case, NVLINK, or a very high-speed data link between the CPU and GPU that results in a significant jump in system performance because this link was one of the largest system bottlenecks.

    This announcement showcasing not only how this link eliminates the bottlenecks between the GPU and the CPU but between GPUs was followed by the announcement of 3D packaging or building one chip on top of another with inter-chip routing on the wafer. This is expected to allow memory bandwidth to go vertical when it comes to market this year. It’s a critical part of NVIDIA’s next-generation CPU code named Pascal. It is expected to nearly double the performance of the current generation.


    The presentation moved to application next because it does no good to massively increase performance if there is no need for that future performance. This is especially intriguing with the increased focus on artificial intelligence as applied to an increasing range of problems from Big Data analytics to self-driving cars. Some of these parts will use neural networking to emulate brain neurons and how brains work. A description of the Google Brain, a massive multi-million-dollar computer used for brain emulation, helped put the advancement in perspective. With Pascal, this capability could be mirrored in a small $33,000 system. That is an increase in cost/performance of more than three magnitudes.

    To showcase machine learning, NVIDIA had the audience tweet pictures of their dogs and showcased how a computer running advanced GPU technology could accurately name the breed of the dog. This brought the audience into the pitch. I’m sure most were anticipating seeing their dogs on the big screen. Unfortunately, they chose poorly because my dogs weren’t selected.


    Next came current products and the $3,000 GTX Titan Z with 300X more efficiency and 400X lower cost for similar performance to other technology. A Transformers-like video visually morphed the last-generation card into this one and the audience was entranced. For perspective, using the old technology, it took 48 hours to emulate water and 250 hours to render each frame in the Academy Award-winning film for best graphics last year. The new card enables this same capability in real time. I think that defines amazing. This was followed by real-time demonstrations of water, smoke and fire that were photorealistic (the fire was so realistic that I actually was surprised I didn’t feel the heat). Of course, the most amazing was a demonstration of Unreal Engine 4. Using this technology, the real-time rendering was hard to tell from movie quality. It was incredibly impressive.


    The business applications presentation included a new type of system: the Iray VCA or the world’s first scalable rendering appliance, according to NVIDIA. Rendering photorealistic virtual images for design and architecture is the purpose for this system and it is fully stackable and scalable. With this class of solution, the technology isn’t viable unless the leading tools support it and apparently the leading tools do. The demonstration was of a new Honda car and the real-time rendered car was indistinguishable from a photograph. They could then slice the car apart and showcase that all internal elements are inside the simulation, including engine transmission, interior and electronics. A designer can now design a complete car in a day, though of course, getting approvals for the design, creating a solid prototype and testing it takes a bit longer. The cost to do this was half a billion dollars. With Iray, it is $50,000.

    Moving from design to enterprise technology, VMware’s CTO came on stage to talk about NVIDIA Grid and how closely the two companies were cooperating to host graphics applications in the cloud. The importance of a recognized enterprise vendor both blessing and advocating the technology is important because NVIDIA isn’t seen as an enterprise provider by itself. Grid is NVIDIA’s high-performance server technology for the cloud.


    Every technology CEO is being pounded to do mobile better, even the market leaders in mobile. The part NVIDIA has for this segment was launched in January at the Consumer Electronics Show. It is the Tegra K1 and the first merger of desktop and mobile requirements into the same part. The demonstration was computer vision as it would be used in cars for accident avoidance or self-driving. What followed was a discussion of tools that developers could use to create amazing mobile applications and of Erista, the next generation of mobile processors. This was followed by an on-stage presentation by Audi, which arguably leads in self-driving technology using Tegra. The Tegra K1 is analyzing over 100,000 images a second. An Audi A7 (side note: NVIDIA loaned me one of these a few years back and it is a wonderful car) drove itself onto the stage and they showed how the technology needed to allow this used to fill the trunk and now is about the size of three paperback books.

    Wrapping Up: Giving an Epic Presentation

    The elements of a stellar keynote presentation are clear: Know your audience and what gets them excited. Engage them during the talk. Make the audience lust after what you are offering. Knock their socks off with demonstrations that hit them where they live. Oh, and think about having one more thing. It was always a signature Steve Jobs mechanism to hold the audience to the end. For NVIDIA, it was an update on its Shield platform. The announcement was that every attendee would get a free Shield. Wow!

    Rob Enderle
    Rob Enderle
    As President and Principal Analyst of the Enderle Group, Rob provides regional and global companies with guidance in how to create credible dialogue with the market, target customer needs, create new business opportunities, anticipate technology changes, select vendors and products, and practice zero dollar marketing. For over 20 years Rob has worked for and with companies like Microsoft, HP, IBM, Dell, Toshiba, Gateway, Sony, USAA, Texas Instruments, AMD, Intel, Credit Suisse First Boston, ROLM, and Siemens.

    Latest Articles