I recently participated in the second annual IEEE Technology Time Machine symposium in Dresden, Germany, the chip-making center of Europe. In Hong Kong last year, the IEEE’s (Institute of Electrical and Electronics Engineers) inaugural TTM event covered chips, devices, networks and systems. This year in “Silicon Saxony,” there were discussions on everything from emerging nanotechnologies to application-layer services. Speakers included such industry leaders and academic luminaries as James Truchard, CEO of National Instruments, and UC Berkeley professor Leon Chua, who correctly postulated a fourth basic fundamental component: the memristor. Here are some highlights from the symposium.
Truchard observed that instruments have evolved from vacuum tubes to transistors and now to software. Today, hardware acquires the data — at accuracies of up to 16 bits at 10 billion samples per second and headed higher — and software does analysis and presentation. To do this in essentially real time, leveraging advanced parallel processing, National Instruments has had to avoid the weaknesses of object-oriented programming — time and concurrency — in favor of “actor-oriented” programming. This is a different paradigm than, say, Web application programming or use of Hadoop or MapReduce. This shift has enabled NI software to be used in quantum optics, tokomak plasma control and the CERN Large Hadron Collider. NI software is also being used in cold fusion research. As Truchard quipped, “We measure, we don’t judge.”
And who’s to say that cold fusion won’t pan out? The next day, Chua recapped J.B.S. Haldane’s four stages for new theories: nonsense, interesting but perverse, true but unimportant, and “I always said so.”
4G LTE not fast enough for you? Time to start thinking about 4.5G. Dresden has a 4.5G test bed running in the heart of the city. Beyond faster data rates, Gerhard Fettweis, a professor at the University of Technology, Dresden, and chair of the symposium program committee, pointed out some implications of the next wave in mobile communications — machine to machine. He expects that beyond Facebook, there will be “Thingbook,” an interaction network for devices. As devices become more ubiquitous, mobility will need to be embedded in everyday items, such as soda cans and sneakers, which, unlike your smartphone, won’t be recharged overnight. Because of this, Fettweis argues that wireless protocols will need to be redesigned for much greater energy efficiency, and batteries will need to last five years or be complemented by energy scavenging. Moreover, emerging applications, such as inter-vehicle coordination for traffic control without lights, will need to operate in millisecond timeframes.
Markus Hofmann, head of Bell Labs research, highlighted innovation around converged networks and IT. Hofmann points out that current approaches are conflicted. Networks would like to maximize utilization and throughput by smoothing load. And applications would like to minimize latency. His team is using game theory to solve the problem. Networks will tune themselves in anticipation of how selfish apps will likely respond to such tuning.
Lauri Oksanen of Nokia Siemens Networks pointed out that networks are the key to the smart cities and societies of the future. In NSN’s vision of a “liquid network architecture,” a thousand-fold increase in network capacity is feasible — leveraging multiple, optimized, self-organizing networks, greater spectral efficiency and more base stations with shorter ranges.
SAP’s Dean Jacobs observed that the software industry will need to either reinvent itself or face commoditization.
Google’s Peter Magnusson stated that Google App Engine is running 500,000 applications and two trillion data store operations each month. Magnusson pointed out that there is still substantial room for innovation in machine learning and customer service. For example, he just bought a car, and it had no information preloaded, such as his home address in the navigation system.
Oracle’s Eric Sedlar argued that elasticity is good, but homogeneity has costs, so there is value in specialization. He also observed that energy makes up an increasing percentage of data center operations costs. Therefore, the costs of over-provisioning are becoming relatively low as long as equipment can be turned off.
Power and chips
Peter Bauer, the CEO of Infineon, stated that energy demand is projected to grow 37 percent in the next 25 years. The latest technology dramatically reduces losses due to electric power transmission to a mere two percent per thousand miles. Silicon is “running out of steam” for power, and other materials, such as gallium nitride and silicon carbide, are of increasing interest.
Chip design overall is undergoing a revolution. IBM’s Paul Farrar discussed new technologies, including silicon photonics, nanowires, nanotubes, spintronics and quantum cascades.
Chua highlighted the memristor, pointing out that flash memories lose 20 percent of their data each decade, whereas titanium dioxide memristors will essentially store “forever.” Moreover, the memristor mimics the way human neurons behave, which makes it easy to build, say, neural nets with 50,000 synapses per neuron. The entire Library of Congress could be stored on a single memristor chip.
Laurent Malier, CEO of the Laboratory for Electronics and Information Technology in Grenoble, argued that 3D is the future for chips. 3D architectures, such as Intel’s Tri-Gate, are not only more space efficient, but faster. By 2020, Malier claims that we should be able to see 100 billion transistors per square centimeter, thanks to stacking 20 layers deep. Wolfgang Lehner, a professor at Dresden, plans to put one million cores into a one-liter cube.
Tom Wiegand, a professor at the Berlin Institute of Technology, talked about some emerging technologies, including image adjustments to simulate eye contact in immersive video conferencing and 3D immersion through the use of 14 stereoscopic cameras. He is conducting studies using electrodes to determine optimal compression levels based on the response of the human visual cortex.
Nokia’s Leo Kärkkäinen showed a number of emerging concepts, including pCubee, a three dimensional box that uses 2-D displays; and proposed crowdsourcing of video through “Directors Cut,” which stitches together multiple videos of the same event captured by different individuals.
Karl Leo, a professor of optoelectronics at the Technical University of Dresden, claims that 55” organic LED (OLED) TVs will be coming soon, and they have a strong potential to replace LCD TVs in the next few years. Moreover, new process technologies enable the creation of such displays on large rolls of aluminum foil and plastic.
A thousand-word post can’t do justice to spending three days with some of the smartest people in the world. Fortunately, we can all head back to the future in 2013, when the event is targeted to be held in the U.S. from May 15 to 17.
Joe Weinman is senior vice president of cloud services and strategy at Telx. Later this month, Weinman will be speaking at GigaOM’s fifth annual Structure conference. The conference will take place June 20-21 in San Francisco.