The 55th anniversary of Moore’s Law is a time for reflection and to ask how relevant the theory remains. The answer is it depends on who you talk to.
But first, a little background. In 1965, Intel’s co-founder, Gordon Moore, predicted that chip improvements would double processor speeds and that overall processing performance would double every two years.
That theory stuck and Moore’s Law became something of a guideline for computer processor manufacturing. Thanks to Moore’s Law, there have been steady improvements made to iPhones, Samsung Galaxy smartphones, and various other devices.
SEE: 10 tips for planning a data center hardware refresh (TechRepublic)
Now, some industry experts believe Moore’s Law is no longer applicable. “It’s over. This year that became really clear,” said Charles Leiserson, a computer scientist at MIT and a pioneer of parallel computing, told MIT Technology Review in February. Moore’s Law, Leiserson said, was always about the rate of progress, and “we’re no longer on that rate.”
In 2019, Nvidia CEO Jensen Huang declared that Moore’s Law is dead and now it’s more expensive and more technically difficult to double the number of transistors driving the processing power. That sentiment was also proclaimed a year earlier by Mike Muller, chief technology officer at chip designer Arm.
Others say not so fast.
It may be slowing down, but “the trend is still there,” said Karen Panetta, an IEEE fellow and dean of graduate engineering at Tufts University. Many people are holding to the true definition that it has to be transistors on silicon and they have to double every two years, and in that case, that is not happening, she acknowledged.
“The name of the game now is the technology may not be traditional silicon transistors; now it may be quantum computing, which is a different structure and nano-biotechnology, which consists of proteins and enzymes that are organic,” Panetta said.
That means the essence of Moore’s Law will likely change given that with quantum computing, “you may end up with exponentially more than” processing power doubling every two years, as well as the use of different materials, she said. “It may not be transistors and a constant number of two…every couple of years; it may be raised to the power of more, depending on what the technology is.”
Mario Morales, a program vice president at IDC, said he also believes the law is still relevant, in theory.
“If you look at what Moore’s Law has enabled, we’re seeing an explosion of more computing across the entire landscape,'” Morales said. “It used to be computing was centered around mainframes and then it became clients and now edge and endpoints, but they’re getting more intelligent, and now they’re doing AI inferencing, and you need computing to do that. So Moore’s Law has been able to continue to really push computing to the outer edge.”
Echoing Panetta, Morales said there’s been a shift in how we think about Moore’s Law. “We’re going beyond it now when we think about incremental improvements of software,” he said. Because so much of computing now includes AI and machine learning, changes are happening much faster than the previous 18- to 24-month period, Morales said.
Moore’s Law will probably be replaced within the next five years—or maybe upgraded based on what comes out of nanobiology or quantum computing, Panetta said.
Morales doesn’t think it will be replaced, but rather, augmented. “Moore’s Law has been in place for 55 years and it’s still going,” he said. “It gets more and more challenging to push the envelope in processing technology—it’s still happening, but it’s getting more expensive.”
Over time, Moore’s Law will be augmented by the amount of innovation in design and AI support, he said. Being able to run artificial intelligence is accelerating how we think about computing, Morales said.
“At the end of the day, you have to abide by what Moore’s Law continues to drive in the cadence of processor technology and the amount of innovation that continues,” he said.