Another computing expert has weighed in on Moore’s Law predicted demise. Speaking at a press roundtable on Wednesday in San Francisco, Henry Samueli, co-founder, chairman and CTO of Broadcom, said that the famous observation that ushered in five decades of ever smaller, faster and cheaper chips will not hold out much longer. This is yet another in a long-line of similar claims from people with first-hand knowledge of the principles involved. IT World was first to report on the story.
Initially proposed by Intel co-founder Gordon Moore in the 1960s, Moore’s Law anticipated that the number of transistors on an integrated chip would double every 24 months. The observation became a kind of short-hand for seemingly endless generations of cheaper, more powerful processors.
There are two main issues with regard to the continued shrinking of silicon-based CMOS transistors: one is based on the limits of physics and the other is economic in nature. Most experts agree that the economic incentives for scaling transistors will be reached before the ultimate physical limitation.
Essentially Moore’s Law has been on extended life support, with manufacturers going to ever greater lengths with each generation of chip architecture. Denser chips used to be cheaper to make because of manufacturing economies of scale, but at a certain point the expense of developing and manufacturing smaller, more powerful chips cancels out the expected cost savings. Many, like Samueli, believe that that tipping point is at hand.
“The cost curves are kind of getting flat,” Samueli told reporters at the Broadcom event. Where before, chipmakers could count on faster processors, less power consumption and lower cost — now they must choose two out of three.
As process nodes approach the atomic scale, chip designers must contend with the strange behavior of the quantum world. There is still room for further miniaturization, but the challenge grows more difficult with each shrink. Samueli believes that the industry will reach a fundamental limit in another three generations or so, at the 5nm point, which is about 15 years away. At 5nm, the transistor gate is only 10 atoms wide.
“You can’t build a transistor with one atom,” Samueli said. “As of yet, we have not seen a viable replacement for the CMOS transistor as we’ve known it for the last 50 years.”