When Intel got into the server market the average cost of a server was $58,000. Now it is $3,800. When the Pentium Pro server was introduced in 1995 everyone thought the demand would not exceed its capacity until 2000. However, since 1995 Intel delivered about 2.7 times the expected growth, up to sixteen million chips a year, from a market that had been a couple thousand. Once again people are saying the market has entered a period of slow growth, but Intel is saying business that doubled in the last ten years will double again in the next five. At the same time advancement in servers will result in virtual servers being used 75% of the time. By 2018 the US Government will have a supercomputer that is 125 times more powerful, but at only twice the power consumption. This trend toward faster and cheaper computer power is expected to continue.
Meanwhile earthquake alerts by Twitter users and WiMax effectiveness during disasters have been recognized as beneficial developments. And those evolving supercomputers allow better hurricane prediction as well. In years past, a hurricane target received a two-day notice for a 300 mile wide target. Today the time is still two days, but the target area is 100 miles. Soon the zone will be ten miles across, making for easier disaster planning and faster evacuations. Analytics of this type allow more timely medicine. Just a few years back a CT scan took 2.5 hours at a million dollar cost. Now real-time scans can now be done in minutes and in a few years the time will be seconds. The same rapid change in speed and cost has occurred with genome sequencing.
Intel does not agree that power and heat will limit Moore's Law, bringing a slow halt to the expanding web. Four billion connected devices currently exist. That number will increase to fifteen billion by 2015 and fifty billion by 2020. A major reason for this is the change from two to three dimensional transistors. Previously the 2D transistor was shrunk down to a very tiny square. The 3D transistor will put skyscrapers on those tiny squares. The squares will continue to shrink, too, from the present 32nm to 22nm next year, a goal of fourteen next year with even smaller possibilities. The use of multiple cores adds even greater potential. All of these will allow for more accurate, timely and affordable developments.
Kirk Skaugen is corporate vice president and general manager of the Datacenter and Connected Systems Group for Intel Corporation. In this role Skaugen leads the business P&L’s, strategy, and product development for Intel’s datacenter, cloud computing, communications infrastructure, and intelligent connected device platforms. His product responsibilities consist of Intel® Xeon® and Itanium® processors, Intel® Atom™ processor-based system-on-chip design, server chipsets, Intel’s wired networking for Ethernet and Thunderbolt™, server motherboards, and software solutions. Skaugen also manages cross-Intel Architecture technology path-finding spanning client to server.
During his Intel career, Skaugen was located in Asia as general manager of the Asia Pacific Solutions Group, with responsibility for the development and deployment of optimized software and solutions for Intel’s client and server platforms. He has also served as general manager of the Enterprise Platforms and Services Division with responsibility for Intel’s server board and system business. Skaugen spent seven years in a number of positions within Intel’s sales and marketing organization, including microprocessor product management, distribution business management and enterprise and client global account management.
Skaugen received his bachelor’s degree in electrical engineering from Purdue University. He joined Intel in 1992 and lives in Portland, Oregon with his wife and son.
This free podcast is from our Web 2.0 Conference series.
Photo: TechPulse 360