by James Montgomery, news editor
July 13, 2010 – Innovation drives everything — and innovation spawned from the semiconductor and related industries is poised to do nothing less than change the world, or at least help everyone adapt better to it, as related by Bernie Meyerson of IBM, in SEMICON West’s opening keynote address on Tuesday morning.
The concept of innovation has significance at IBM — each year in an "Innovation Jam" it asks all employees to tell, basically, what big issues are important to them, and from that figure out what challenges can be addressed and solved with technology. Examples:
Water. In the US, 36 states foresee a water shortage (a top-of-mind theme to a show in California!). And two-thirds of cities in China are already there.
Traffic. Meyerson’s slide showed a photo of paralyzing traffic jam in just one "small business district" in Los Angeles — and how things are worse in places like Rome and Beijing — and how traffic problems lead to huge amounts of lost time, wasted gas, tons of carbon emissions, all translated into billions of dollars.
Food. A typical carrot travels 1600 miles to its final destination, other foods hundreds of miles. And $48B worth of food is simply thrown away.
Healthcare. 2.2M errors in dispensing medications occur simply because the prescription is misread.
The common theme among all these areas, Meyerson summed, is they are largely caused by too much or not enough information. Thus, he proposed, applying IT can help create a "smarter planet" to solve them. And this touches our industry in a handful of ways: semiconductors (logic, memory, storage), sensors, energy, etc. Our industry has a good track record at meeting challenges and pushing the envelope — over the last hundred years, there’s been a 100 improvement in what $1k can get in technology (battery developers are happy today with an 8× improvement!).
And the challenges are significant. Sequencing the genome in and of itself isn’t a revolutionary world-changing thing, Meyerson suggested — "but sequencing the genome for $80 — that’s revolutionary!"
Turning his attention specifically to advancements in semiconductors, he addressed the problems with Moore’s Law and scaling silicon, and how efforts to find answers have "lit up" the periodic table, which has also greatly increased complexities and costs. Some technologies are being investigated and show promise (e.g, nanowires, carbon nanotubes, and graphene), but none are ready to scale to making billions of devices.
Meeting challenges isn’t just about improving devices. "The speed of life is dreadfully slow," he said — specifically, it’s 3×108 m/s. In football field-sized data centers, it takes lots of cycles (and power too) just to fetch data from a server 100m away. 3D can change that equation to just 100μm, he noted.
Radical collaboration, system-level innovation
Since finding new ways to develop technology will involve bumping up against fundamental rules of physics, industry has come around to the need to combine intellectual as well as financial capital in "radical collaborations" — such as has been done with the IBM-led Common Platform Alliance, in which companies with the same base technology go out and compete for customers. Such a model has made it into the course teachings at Harvard Business School, Meyerson noted.) Such collaboration has to come in the face of such massive numbers vs. the past. There will be a projected 2B people on the Internet, which has been a popular flag to wave as representative of demand — but it’s the "Internet of things," devices that produce and share data, which will be the real drivers, he noted. Consider: there will be 33B RFID tags, all wanting to "talk" to IT. And overall, a trillion various types of Internet-connected devices.
One area that can take advantage of cloud-computing, application specific methodology, one that can be highly utilized for batch workloads, and prioritize jobs in real-time, is EDA. Meyerson showed how IBM has created a "design cloud" with 20,000 cores, 150+TB RAM, and 3000 users, and figured out how to intelligently manage system workload so that utilizations are maximized — a slide showed a 48-hr map where system utilization was "steadfastly" above 90%.
Addressing the aforementioned global "human" challenges (water, etc.) will require just such system-level innovations, "holistic designs," Meyerson said. One example: "stream computing," essentially real-time continuous digestion and analysis of data from numerous sources, all funneled to a person (or decision tree) to make an informed decision. Consider a commodities trader pulling information from the NYSE, SEC, video news, and weather data to calculate proper buying (or selling) of oranges, with <1msec latency. TD Bank, he said, is already doing this to identify and execute trades, with <150μm latency. For a healthcare spin, a group in Ontario is using this methodology to identify infants in various ways, ultimately able to predict onset of a particular dangerous affliction a full 24hrs before a trained ICU nurse.
Another example of an intelligent "systems of systems": transportation. Generally speaking, conventional operation is to compile transaction data, e.g. tollbooths. Beyond that, systems could determine who crossed the point of toll, where, and how fast, and from this proactive traffic models could be devised. Ultimately the system could be optimized to take that data into account and build pricing models and even citywide infrastructures. Meyerson showed a correlation of traffic volume and speed, where 10-min ahead forecasts and real-time data tracked basically into identical trend lines. Such a system is already being established in Singapore, he noted.
Such system-level activity — not discrete differentiators — will define the "winners" in just the next five years, Meyerson suggested.
Quips
An engineer joke, courtesy of Meyerson: What happens when an engineer gets his hand caught in a door? He pries it out, and then sticks his other hand in to get a second set of data.
Technology development isn’t just complex and costly, it can be downright ugly. Meyerson showed a photo of the EUV lithography tool currently operating at Albany Nanotech. Not only does it cost roughly $100M, but, he said, it looks like "someone took an electromagnet through an auto junkyard."