For more than four decades, technology has allowed chip makers to place twice as many transistors into the same space every 24 months. This design capability, commonly known as Moore’s law (named after Intel Corp. co-founder Gordon Moore), has enabled the rapid advancement of innovation from personal computers to smartphones and beyond.

But the gift that keeps on giving is running out. Intel has been slowing the pace of launching new chip-making technology, and this is forcing software developers and hardware designers around the world to seek unconventional solutions that can meet growing demand for compute power, especially when it comes to the need for high-powered training models required by artificial intelligence applications.

The answer may lie in the ability to reconfigure hardware devices to gain more speed and expand system capability, a process known as scale-out.

“For me as a software person, the end of Moore’s law is a bad thing because I can’t increase the compute power any more on a single chip. The only possible way to speed up the training of those models to enable AI is to scale out, because we can’t put more cores on the chip,” said Natalia Vassilieva (pictured, left), senior research manager at Hewlett Packard Enterprise Co.

Vassilieva recently spoke with Dave Vellante (@dvellante) and Peter Burris (@plburris), co-hosts of theCUBE, SiliconANGLE Media’s mobile livestreaming studio, during the HPE Discover EU event in Madrid, Spain. She was joined by Cat Graves (pictured, right), research scientist at HPE, and they discussed the company’s work on memory-based solutions, HPE’s development of a powerful new machine, potential applications for high-performance computing, and influences in their own technology research careers. (* Disclosure below.)

This week, theCUBE spotlights Cat Graves and Natalia Vassilieva in our Women in Tech feature.

Memristor offers smaller, faster solutions

Scientists at HPE Labs are looking at two key areas to address the dilemma of a slowing Moore’s law. The first involves a new Memory-Driven Computing architecture that relies on memristor, non-volatile electrical components that can retain memory without power.

An example of what this means can be found in one of the common headaches in daily life: the computer hard-shutdown. When power goes out, everything has to be rebooted, and whatever was open needs to be restored. With memristor, everything would be right there on the screen as soon as the computer powered up. It’s a pipe filled with information that never drains until asked.

HPE Labs developed the first memristor in 2008 and has been working on integrating the technology to create smaller, faster and more energy-efficient computers. “We’re seeing this potential for a huge amount of speedup and also the potential energy savings as well,” Graves said.

The second key research area involves designing architecture that can apply task-specific accelerators to this pool of ready memory. HPE Labs has been working on this concept through its Dot Product Engine, a high-density accelerator for performing complex matrix-vector multiplication. In other words, it’s really powerful and really fast.

How powerful? HPE recently provided a peek at this next-generation computer architecture during a conference in Washington, D.C. High-end gaming systems generally have 64 billion bytes of memory. HPE’s prototype has 1,280 high-performance microprocessor cores that can access 160 trillion bytes of memory.

“It’s not just a clock speed issue; it’s thinking about what computations actually matter, which ones you’re actually doing, and how to perform them in different ways,” Graves explained. “We have an ecosystem now that is actually favoring accelerators and encouraging the development of these specialized hardware pieces that can slot into the same architecture that can scale.”

The significance of HPE Labs’ research could soon become apparent as the “internet of things” powers up and starts generating massive amounts of data. HPE’s approach is to create a neural network or central learning engine that can gather device experiences and apply AI rapidly and energy-efficiently.

Reliance on computational memory

Memristor and the Dot Product Engine offer a way to deal with the slowing of Moore’s law and still achieve the computer power and speed that will be necessary in the data-driven world.

“Even if you increased the power of a single device, then the challenge will be how to bring the data fast enough to that device,” Vassilieva said. “What the Dot Product Engine does is computations in-memory, inside, so you limit the number of data transfers between different chips.”

The market for this kind of low-power, high-performance computing is expected to be robust. A first wave of related hardware architecture can already be seen in HPE’s recent release of its Edgeline EL1000 and EL4000 solutions. The idea is to integrate compute, storage, control and data capture functions in one box that can transfer information to and from the edge as needed.

The growth of edge connectivity is driving interest in a number of fields, including space travel. “There’s a lot of interest in the automotive industry, space and robotics for more low power but still very high-performance, highly efficient computation,” Graves said.

Scientific curiosity and supportive teachers

The work of both scientists at HPE is a continuation in the lifelong pursuit of learning in math and physics. Both of Graves’ parents are scientists, and she was exposed to technical concepts at an early age.

“We always had books around the house, and I was encouraged to think and be curious,” Graves said.

Raised in the Soviet Union, Vassilieva took her first programming class in middle school and was encouraged by teachers to pursue specialized training. Women were provided more opportunity in many fields inside the Soviet state, including computer programming, because they were often required to take difficult jobs after World War II, according to Vassilieva.

“For me, it’s not the percentage that matters,” Vassilieva said. “Just don’t stand in the way of people who are interested in that, and give equal opportunity to everybody.”

Watch the complete video interview below, and be sure to check out more of SiliconANGLE’s and theCUBE’s coverage of the HPE Discover EU event. (* Disclosure: TheCUBE is a paid media partner for the HPE Discover EU event. Neither Hewlett Packard Enterprise Co., the event sponsor, nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE