Tech Talk iconTech Talk

Why Aren't Supercomputers Getting Faster Like They Used To?

Currently, the world’s most powerful supercomputers can ramp up to more than a thousand trillion operations per second, or a petaflop. But computing power is not growing as fast as it has in the past. On Monday, the June 2015 listing of the Top 500 most powerful supercomputers in the world revealed the beginnings of a plateau in performance growth.

There are a number of technical aspects and economic factors that interfere with supercomputing improvements. Experts disagree on the cause, but the result could be a slowing of the pace of improvement in some scientific fields.

Read More

Intel Hits Snag On The Way To Next-Generation Chips

Depending on what stories you’ve been reading during the last week or so, you’d either think Moore’s Law is in very deep trouble or has been rescued yet again from imminent demise.

Last week, IBM made a big splash with news of a 7-nanometer chip, which employs silicon-germanium instead of silicon, a long-awaited move toward alternate channel materials. Although not yet ready for mass production, the chip signaled we’re still on track to make smaller, cheaper, and better transistors.

Then this week, Intel CEO Brian Krzanich announced on a quarterly earnings call that his company has hit a sticking point with the manufacture of its 10-nm chips. The first 10-nm product, a chip code-named Cannonlake, will now be released in the second half of 2017. Depending on how you count, that’s a delay of at least six months. (At one point, the debut of this chip was ballparked for as early 2015.) 

If you haven’t been keeping close track, 14-nm is the smallest manufacturing generation currently in mass production. The 10-nm generation comes after 14-nm; 7-nm follows 10-nm. Each generation, or node, is supposed to have smaller features. And with any luck, those smaller transistors will also be capable of producing speedier and less power-hungry circuits than their predecessors.

One possible way of interpreting Intel’s delay and IBM’s chip is to conclude that IBM is gaining ground on Intel. But IBM has said little about when 7-nm chips would debut. Bringing those chips to maturity will likely rely on GlobalFoundries, which recently completed acquisition of IBM’s fabs. 

Another possible interpretation: Moore’s Law is faltering. Intel has long set the pace, hitting a new node every two years. But according to the earnings call transcript, Krzanich said Intel is now releasing chips with smaller transistors every two and a half years. He pinned at least part of the difficulty on the ability to print finer features.

The thing is, it’s hard to say what will happen going forward. Maybe we’ll get back to a faster pace of chip releases. Perhaps this slowdown will only continue. 

Chipmakers are battling the laws of physics, to be sure. But their timelines also depend on the economics of manufacturing—factors such as yield (the fraction of chips produced that actually work) and how many manufacturing steps it takes to transform portions of a wafer into fully-functional chips. 

But whatever happens, Moore’s Law won’t grind to a halt overnight. The cadence of Moore’s Law has changed before. And if it’s happening again, we may see other developments emerge that will pick up the slack, ones that explore what’s possible beyond simple miniaturization. The semiconductor industry has had the good fortune of being able to guide itself by a simple principle for 50 years. Now it seems the way forward is starting to get just a bit more complicated.

See our special report “50 Years of Moore’s Law” for more.

Weyl Fermions Found, a Quasiparticle That Acts Like a Massless Electron

After an 85-year hunt, scientists have detected an exotic particle, the “Weyl fermion,” which they suggest could lead to faster and more efficient electronics and to new types of quantum computing.

Electrons, protons, and neutrons belong to a class of particles known as fermions. Unlike the other major class of particles, the bosons, which include photons, fermions can collide with each other—no two fermions can share the same state at the same position at the same time.

Whereas electrons and all the other known fermions have mass, in 1929, mathematician and physicist Hermann Weyl theorized that massless fermions that carry electric charge could exist, so-called Weyl fermions. “Weyl fermions are basic building blocks; you can combine two Weyl fermions to make an electron,” says condensed matter physicist Zahid Hasan at Princeton University.

Read More

Introducing a New Material for Invisibility Cloaks

Unless you’re a teenage wizard, making things invisible involves some challenges—cloaking devices tend to be bulky and absorb some of the light they’re trying to reroute. Now a new design may lead to invisibility cloaks that are thinner and don’t lose brightness, rendering them more practical for certain uses.

Read More

A Middle East Supercomputer Makes the Top 10 List for the First Time

For the first time, a system in the Middle East earned a Top 10 spot on the list of most powerful supercomputers. Shaheen II, located at King Abdullah University of Science and Technology (KAUST), in Saudi Arabia, placed 7th in the the semi-annual competition, the results of which were announced earlier today. Shaheen II is a Cray XC40 system that cranked out 5.536 petaflops per second on the Linpack benchmark.

Read More

Google Funds University Living Lab for Internet of Things

Carnegie Mellon University’s campus could soon transform into a living laboratory for testing how Internet-connected sensors, gadgets, and buildings might change our daily life. Google has awarded half a million dollars to Carnegie Mellon and a broader university coalition to develop the technologies needed to make that vision a reality.

Read More

With Pluto Encounter a Generation Passes the Torch

Those of us now into retirement age are—among holding other distinctions—the generation that the late Carl Sagan modestly described as holding a unique place in the entire history of humanity. As children, he related, we could look up at the other planets but could only wonder what it was like on them, based on telescopic observations, or on guessing. Sixty years later, as grizzled geezers, we remember watching the entire sequence of planets one after the other being probed and measured by human devices, revealing alien worlds in awesome detail. Now Pluto—which a spacecraft called New Horizons will fly by on Tuesday—completes the set. We were the last to have to guess and imagine; all future generations will have factual data and on-site images at their fingertips.

It is an amazing lifelong parade of discoveries that we can ponder in anticipation of the Pluto “grand finale”. As a lifelong space nut —I was 12 when Sputnik launched and I was already star struck—with a career in spaceflight, I’m bouncing up and down with anticipatory glee.

But I now see that Sagan was overstating the uniqueness of our cohort’s good fortune in scheduling our birth dates. Our grandchildren need not be envious of an experience they can never share, because I’ve become persuaded that their lifetime of space exploration will exceed our own in ways we didn’t even realize.

Read More

Swiss 'Pac-Man' Satellite to Gobble Up Space Junk in 2018

Until the laser cannon on the ISS is fully armed and operational, we’re going to have to find other ways of dealing with the existing problem of space junk, or else. Earlier this year, we wrote about how the European Space Agency is experimenting with junk-snagging nets for a satellite removal mission in 2021, but EPFL’s Center for Space Engineering doens’t want to wait that long, and they're getting ready to send up “a giant Pac-Man to gobble up space debris” in 2018.

Read More

Tech Talk

IEEE Spectrum’s general technology blog, featuring news, analysis, and opinions about engineering, consumer electronics, and technology and society, from the editorial staff and freelance contributors.

Newsletter Sign Up

Sign up for the Tech Alert newsletter and receive ground-breaking technology and science news from IEEE Spectrum every Thursday.

Load More