For those of you who attended the International Test Conference in Atlantic City this year, I hope you got up early on Tuesday, Sept. 28, to catch the keynote address. It was presented by Patrick Gelsinger, vice president and general manager of the Desktop Products Group at Intel. Mr. Gelsinger knows a little bit about microprocessors as he was general manager of the division responsible for the 486, DX2, and PentiumPro?? microprocessor families. He also was design manager and chief architect of the I486 microprocessor.
Early on in his address, Mr. Gelsinger laid some pretty impressive statistics on the audience. He said Intel sees a world marching toward a billion computers interconnected over high-performance networks that access millions of servers, which are performing billions of transactions involving trillions of dollars on a daily and yearly basis.
This is the world of the internet. The total number of users worldwide being connected to the internet is growing at a 36% cumulative annual growth rate. In the United States today, there are approximately 50 million users, and by 2002, Intel expects the number of users to grow to 140 million. According to a recent market survey conducted by Intel, the number-one reason people are buying computers is to access the internet.
Not only is the demand for computers growing, but so is the speed at which they operate. As for microprocessor speed, Mr. Gelsinger said that in the Desktop Products Group, the slowest product sold today is 433 MHz. The fastest one is running at 700 MHz. Intel has told the industry that in the second half of 2000, microprocessors will be available that run at 1 GHz.
Now with all these mind-boggling numbers sprayed out before an audience of test professionals, Mr. Gelsinger got down to the meat of his keynote speech. What are the test challenges that must be addressed to meet these demands? Among the four that he spoke about in-depth, I want to touch on a couple of them here. He said ATE capabilities are falling further behind the characteristics of the devices to be tested, and power delivery is a significant issue.
On power delivery, many of today??s devices are operating with a Vcc of less than 2 V. We have sub 2-V devices that are switching 20, 30, 50, and even 100 A. And these devices can be switching at gigahertz rates. This means that power must be delivered to these devices at a rate of tens of amps per nanosecond. This is not a trivial situation.
As for ATE capabilities, Mr. Gelsinger noted that as devices push down into the 0.18-micron level and below, a new type of fault occurs which is not detected by traditional fault-detection techniques. Stuck-at faults and opens and shorts are not the problem. But resistive faults are. These are classified as resistive opens and resistive shorts and are more prevalent for deep submicron devices.
Of course, Mr. Gelsinger drove home the need to reduce the cost of test. Perform more testing during burn-in, implement more built-in self-test which can reduce the complexity and cost of the ATE, standardize testers to allow their reuse across multiple generations of devices, and standardize test software so it can be used on any vendor's ATE.
As summed up by Mr. Gelsinger, “As we look at the test industry today, we think a fundamental transformation, a fundamental change in behavior is required to meet these challenges.” These are lofty goals. But the alternative is not acceptable.
Paul Milo
Editor
e-mail: [email protected]
Copyright 1999 Nelson Publishing Inc.