• September 21, 2016

No Moore: What’s After Moore’s Law?

This transistor density law is stalling; how do we respond?

It’s the end of a golden age. For half a century, Moore’s Law essentially powered the information technology revolution. But faced with physical limitations and advances in technology, the law seems to falling by the wayside. What’s next?

Formulated by tech guru Gordon Moore, Moore’s Law states that the number of transistors on a microprocessor doubles every 12 to 24 months. The Intel co-founder made his seminal observation in a 1965 paper titled Cramming More Components onto Integrated Circuits. From that point, transistor density escalated, doubling every 24 months since 1975.

Density Breakthroughs

What is the result of Moore’s Law in practice? Transistor density is a rough measure of processing power. As density increased, processors shrunk even as their clock speeds increased, propelling a whole new class of computers every 10 years. In the law’s wake, mainframes gave way to minicomputers, which gave way to personal computers, which are being overtaken by laptops, smartphones, tablets, and embedded processing. Microprocessor-driven devices grew ever faster, more efficient and reliable even as they became cheaper to produce.

Moore predicted groundbreaking innovations based on his observation including home computers, digital watches, cars embedded with processors, and portable communications devices—cell phones. By the early 1980s, there was an explosion in demand for digital technology expressed by things like Apple computers, Atari game consoles, Sony compact disc players, and Hewlett Packard handheld calculators. Moore’s Law translated into nearly 50 years of uninterrupted technological growth and innovation.

Breakthrough Brake

Yet by the early 2000s, Moore’s Law began a steady stall, slamming headlong into the stubborn reality of physics. Challenges surrounding power consumption, heat dissipation, and scaling mounted. Advancements in photolithography—the process that transfers chip patterns to silicon wafers—halted at wavelengths of 193 nanometers, slowing increases in transistor density.

Overcoming these difficulties heightened fabrication complexity as well as costs, colliding with a corollary to Moore’s Law: Rock’s Law. Rock’s Law states that the cost of chip fabrication plants doubles every four years. Today, building a fabrication plant capable of producing the newest processors rings in at an estimated $8.5 billion.

The march of ever-increasing processor density is running up against other formidable limits as well. As transistors get tinier and tinier, electron behavior descends into the quantum realm, a space ruled by weird uncertainties. Performance becomes erratic and unreliable. From a materials standpoint, there’s no viable successor to current silicon technology and its limitations. Result: processor clock speeds have largely ground to a standstill since 2004.

Recapturing Moore Magic

Engineers have responded by redesigning circuitry, creating chips with two, four, or more processors. But exploiting additional processors can be difficult, as many algorithms can’t break tasks down into pieces for distribution to multiple processors.

There are a number of strategies in play to surmount Moore’s Law roadblocks. One is 3-D architecture, which stacks thin layers of etched silicon into skyscraper-like structures, instead of etching circuits onto a flat silicon surface. Another is photonic computing, which replaces streams of electrons with photon streams, boosting speeds while alleviating heat. Microfluidic cooling, which uses microdrops of water circulated through channels in silicon chips, eases destructive heat buildup associated with intense transistor density. Other developments, such as The Machine, would rely on memory-driven computing rather than processor-driven computing.

Yet perhaps the solution to the Moore’s Law dilemma lies in shifting focus from sheer processor power and speed to sophistication and refinement. Computing has increasingly shifted to mobility and the edge, with tasks relegated to mobile devices and sensors connected to servers in the cloud. The need for ever-faster processing in stationary devices is receding.

Today, computing is less about doing a few tasks at ever-accelerating speeds than it is about doing a greater number of varied tasks at once with greater integration and efficiency. As Moore’s Law comes to an end, computing technology is entering a new era.

Like this article? Check out an infographic on why you can’t ignore the mobile revolution.