Design, deliver, and run enterprise blockchain workloads quickly and easily.
Moore's Law is ending. Now what?
In 1965, Gordon Moore predicted that the number of transistors per square inch on integrated circuits would double every other year for the foreseeable future. His prediction, now known as Moore's Law, has proved remarkably durable but is nearing the end of its useful life.
As a result, the world will need new technologies to keep pace with exploding data volumes and insatiable demand for the insights that data contains. "Can we continue to kick the can down the road?” asked Kirk Bresniker, chief architect for Hewlett Packard Labs, at the company’s recent Discover conference in Las Vegas. “No, we really need to have this conversation now."
Hewlett Packard Enterprise has invested significant resources in developing three primary alternatives to traditional computing architecture: neuromorphic computing, photonic computing, and its keystone technology, Memory-Driven Computing (MDC). All three technologies have been successfully tested in prototype devices, but MDC is at center stage.
Machines that think
At a presentation on the end of Moore’s Law at Discover, Labs research scientist John Paul Strachan discussed the Dot Product Engine (DPE) system, a Labs approach to neuromorphic computing. This technology addresses the Moore's Law challenge because it's scalable in a way that traditional integrated circuits are not.
DPE uses arrays of memristors—a type of circuit that can "remember" its previous charge state—to model human brain function in three ways. It mimics the architecture of neurons and synapses, it does all of its computation within the arrays, and it reproduces key brain operations called "matrix operations."
“Matrix operations lie at the heart of all machine learning,” said Strachan. In the case of image recognition, for example, the matrix operation consists of multiplying one set of images by another. That's how a machine learning program learns to distinguish a cat from a carrot.
Computing with light
Photonic computing has the potential to push computing past the end of Moore's Law because it processes information with little or no energy loss.
Bresniker noted that HPE's X1 optical engine allows data to travel quickly between node enclosures. Using this technology, it costs no more energy to send information down 1,000 meters of fiber than it does to send it down 10 centimeters.
"At the right scale, the wave properties of the photon emerge," he said. "Where there are waves, we begin to get interactions that can be used to compute."
HPE engineer Thomas Van Vaerenbergh demonstrated a photonic logic gate, a device that creates the 0s and 1s (closed/open) that computing requires using controlled light interference. Photonics enables complex operations like encryption and decryption to happen literally at the speed of light.
Meet the Machine
HPE's Memory-Driven Computing architecture combines photonic data transmission with non-volatile memory that retains information even when it isn’t drawing power, and systems on a chip that package processors and memory to greatly speed data processing.
As part of an ongoing research program called The Machine, HPE recently announced what it claims is “the world’s largest single-memory computer,” a system with 160 terabytes of memory that was on display at Discover.
Running a genomic analysis program for the German Center for Neurodegenerative Diseases (DZNE), the system reduced the time needed to run a gene assembly pipeline from almost six days down to two and a half minutes.
The tech industry after Moore's Law
Technology vendors will soon have to face the reality that they cannot offer better machines every year or two based on the steady improvement in processing power described by Moore's Law. Customers may benefit from the end of Moore's Law because they won't need to update their hardware so often. However, Strachan noted that tech products based on new computing architectures will initially be more expensive.
For enterprises, the hope is that innovations like MDC will enable efficiencies that offset the expense of implementing new computing technologies. This involves risk, because companies will need to experiment more with different technologies and not assume that the tech they've used for decades will be viable going forward.
"They are not used to that, at least not the bigger ones,” said Van Vaerenbergh. He predicted that many companies will need support from governments to navigate the transition to new technological paradigms.
He pointed to research by HPE Senior Fellow Stanley Williams that resulted in the White House’s Nanotechnology-Inspired Grand Challenge for Future Computing. The nanotech challenge encouraged scientists and companies to address fears about nanotech, to manage expectations, and to create working technology with practical applications. It succeeded on all fronts.
The lesson? “Those who don’t want to do long-term research are doomed,” said Van Vaerenbergh. On the other hand, companies that embrace the opportunities created by the end of Moore's Law will have many opportunities to profit.
The End of Moore's Law. What's Next? by R. Stanley Williams, Hewlett Packard Labs.
This article/content was written by the individual writer identified and does not necessarily reflect the view of Hewlett Packard Enterprise Company.