For decades, ever-increasing silicon chip density has fuelled faster and better computing.
Researchers from Rockefeller University shed new light on Moore’s Law, perhaps the most well-known technological prediction in the world. It predicted that chip density (or the number of integrated circuit components) would double every two years.
PLOS One published a study that reveals a subtle wave pattern in historical data about the rise of transistor density in silicon chips, which makes computers faster and more powerful.
According to the paper “Moore’s Law Rediscovered through Intel Chip Density,” there have been six waves since 1959. Each wave saw transistor density per chip increase at least 10-fold during these six years. This paper expands on an earlier study that used DRAM chips for modeling technological evolution.
This new research clarified the arcs in the wave pattern by adopting an innovative perspective on chip density. It factors out the changing sizes of chips used in Fairchild Semiconductor International and Intel Processors beginning in 1959.
According to Jesse Ausubel of Rockefeller University in New York, three years of negligible growth they occurred after each six-year growth episode.
They believe that the next boom in transistor miniaturization is coming soon.
The demand will pull it for, e.g., Data-hungry artificial intelligence technologies such as facial recognition, 5G mobile networks and equipment, self-driving vehicles, and other high-tech innovations that require more incredible processing speed and computing power.
Cerebras, a startup, boasted that the Wafer Scale Engine was the largest chip ever made. It is 56 times larger than the largest graphical processor unit (GPU) and has been a dominant computing platform for AI and machine learning.
“The wafer-scale chip contains 1.2 trillion transistors, embeds 400,000 AI optimized cores (78x more than the largest GPU), and has 3,000 times greater in-chip memory.
They say that the end of the silicon chip era is near. There are only a few silicon pulses left before it becomes exponentially more difficult to make further advances due to economic and physical limitations.
The future growth of the computer industry depends on miniaturized innovations like nano-transistors, single-atom transistors, and quantum computing.
The paper notes that in 2019, Google’s parent company, Alphabet, made a quantum computing breakthrough with a programmable supercomputing process called “Sycamore,” which uses programmable superconducting qubits.
The published benchmarking example showed that Sycamore completed the task in 200 seconds. This feat would have taken a supercomputer of current technology about 10,000 years to complete.
Director of the PHE, Mr. Ausubel, says: “We have climbed six times into higher valleys of silicon and similar substrates, but we may be leaving the silicon valleys to explore landscapes of other materials or processes.”
“Qubit Gardens might await at the top of the current climb.”
The paper’s title refers to Gordon Moore’s 1965 observation that microchips have an exponential growth rate – double every 12-24 months (Moore’s Law).
The analysis of transistor density revealed a more complicated pattern of sequential waves of growth. Each phase lasted approximately nine years before reaching saturation and being replaced with a new one.
Dr. Burg is also affiliated with Tel Hai College in Israel. He says the new work uncovers important subtleties within a technological phenomenon that has fueled world progress for two decades.
He says that the work was based on models used for studying growth with complex feedback, which can lead to limitations in density previously used in such research. It also shows their ability to illuminate complex machinery’s evolution.
