The last major chip design that Mooly Eden convinced Intel to adopt was a failure. As he prepared the design in the late 1990s that would become the Pentium M, he knew he couldn't afford to miss a second opportunity.
Eden's team had come up with a novel design that emphasised power management and interoperability over raw performance, a concept at the time as foreign to Intel marketing executives as the shores of the Mediterranean Sea, where Eden's office at Intel's design facility in Israel was located.
The Israelis on Eden's team were a different bunch from the US Intel engineers who produced most of its prominent chip designs of the era. They were a manager's nightmare, Eden recalled over a recent dinner in California.
"It's easier to manage 50 Americans than five Israelis. [The Israelis] are undisciplined, they challenge everything, they are almost more Intel than American employees," Eden said, reminiscing over a glass of wine. He was alluding to the demanding engineering culture founded by Robert Noyce, Gordon Moore and Andy Grove that he believes has been lost in a land of political correctness and polite management.
The design that emerged from that group of irascible engineers was code-named Banias, taking the name of an ancient Israeli city known for its historical significance to three religions. As Eden prepared in 2000 to update a group of Intel executives in California on the progress of the project, prayer seemed like a good idea.
Banias was everything that the Pentium 4, Intel's new desktop processor, was not. Banias was designed to consume as little power as possible, while the Pentium 4 was designed to let Intel steadily increase its clock speed over the next five years or so.
Turning up the heat
But as the clock speeds of the Pentium 4 accelerated, so did the amount of power it consumed. And in the late 1990s, Eden foresaw that if Intel kept shrinking transistors at the pace it was on during those years, it would soon reach a point where that power would be able to escape from the chip and produce unacceptable amounts of heat.
"We knew we'd hit a power wall before 2000, and we were aggressive in selling it to management," Eden said.
Eden can be an intimidating presence for a slight, bespectacled man. Every time he opens his mouth, whether it's to explain a complicated bit of technology or order dinner, he speaks with a conviction and resolve that stem from his service as a major in the Israeli Defense Force.
In order to force Intel into what he called a "right-hand turn," to avoid that wall, Eden's engineers would have to bring a maniacal focus on reducing power consumption to every aspect of the Banias design, including the CPU, chipset and memory interface, he said.
At the time, Intel tended to design the chipset and processor separately, in part because of Eden's greatest failure at Intel.
After rising to prominence within the company for his multimedia extensions enhancements to the x86 architecture, Eden worked on a project called Timna.
The Timna processor was designed with integrated components for low-cost PC vendors that wanted a complete product that was easy to integrate into a PC chassis; Intel promoted it as a chip for PCs in emerging markets. Timna never had a chance after Intel chose to use Rambus' RDRAM memory standard across all its products. Timna's integrated nature meant that it was tied to specific technologies, such as the memory standard. When it became apparent that RDRAM was not going to be the future of computer memory, Intel dropped its support for the standard and Timna was never released as a product.
But this time, an integrated approach was crucial to the Banias design, Eden said.
The efficiencies within the chip could only be fully exploited if the processor and chipset were built to the same exacting standards for power consumption, he said.
In Eden's mind, the chipset for Banias was just as important as the processor.
However, there was a major problem. Banias ran at slower clock speeds than existing Intel processors.
Over the past decade, Intel spent hundreds of millions of dollars educating a generation of PC users that faster clock speeds meant higher performance. The company knew there was much more to the performance equation than the speed at which a chip executes instructions, but pipeline architecture and cache memory are a lot harder to understand for regular PC users than the notion that two is more than one.
So Eden wasn't just asking Intel to take a chance on Banias as a new notebook processor. He was asking the company to discard its current design philosophy, overhaul its speed-addicted marketing strategy and promote the concept of the whole system over the concept of the individual processor.
It would have been easy for Intel to pass on the processor. The company was fresh off a disastrous year in which it swallowed its pride and admitted that its Rambus decision would not work out, as well as deal with the aftermath of a delayed Pentium 4 launch due to manufacturing glitches.
Change of direction
But as any gambler knows, the largest rewards usually come to those willing to take the biggest risk. With current president and COO, Paul Otellini, solidly behind Eden's vision, Intel signed off on Banias.
The result is a processor design that is expected to form the backbone of Intel's strategy for the next several years, as well as a new focus on designing features that promote usability and system management, rather than pure performance.
Intel has now completed Eden's "right-hand turn", reorganising its operating divisions around specific technologies such as the Mobility Group, The Digital Home Group and the Digital Health Group, as opposed to its previous organisations focused on processor architectures.
However, a recent suggestion that Eden is responsible for changing the direction of the world's largest chipmaker prompted nothing but a slightly annoyed look.
"A lot of people should take credit for the project," he protested. "No one person can take credit for a project that large."
However, it's not hard to imagine Eden pushing his team with just the right mixture of constructive criticism and humour. He laughs more easily and boisterously than the average technology executive, and few Intel executives could be found showing a humorous commercial promoting tourism in Israel to most of an upscale restaurant in the middle of dinner. On a Centrino laptop, of course.