SPONSORED BY

eInfochips

Get Ready to Rumble

Sally Ward-Foxton

Ladies and gentlemen, in Nvidia’s corner, from Oak Ridge, Tennessee, world champion since June 2018, weighing in at Linpack performance of 148.8 petaflops: Summit! In Arm’s corner, now topping the list of world’s most powerful computers, from Kobe, Japan, weighing in at 415.5 petaflops and six years in the making: Fugaku!

This week, Fugaku unveiled performance figures that bested previous world champ Summit by a factor of 2.8. This supercomputing heavyweight, with physical size measured in warehouses (396 racks), runs on Arm-based A64FX SoCs from system builder Fujitsu. At 152,064 compute nodes and more than 7 million compute cores, it doesn’t pull its punches.

Image: Fujitsu

Fugaku is the undisputed champion, holding TOP500, HPGC (High-Performance Conjugate Gradient) and HPL-AI (High-Performance Linpack AI) titles. Fugaku also retains the Green500 top spot as the world’s most energy-efficient supercomputer.

This is the first time an Arm-based supercomputer has taken the TOP500 title. Summit uses six Nvidia Tesla V100 GPUs in each of its 4,356 compute nodes, alongside two IBM Power9 CPUs (total 2M+ cores). While the systems in its corner may have lost the overall title, Nvidia still came out swinging in this round. It accelerates eight of the 10 fastest supercomputers on the list and two-thirds of the TOP500 overall.

How long will Fugaku hold the title? Hard to say. New challenger Frontier is due to come online sometime in 2021. Frontier, built by Cray and co-located with Summit at the Oak Ridge National Laboratory, is expected to top out at 1.5 exaflops (Fugaku’s peak performance is about 1.0 exaflops by comparison). It will use AMD Epyc CPUs and AMD Radeon Instinct GPUs.

1… 2… 3… it’s a knockout! Fugaku will be fully operational by next April.

Enjoyed this article? Subscribe now to continue receiving EE Times Weekend.

Life or Death on Remote Control

Junko Yoshida

Missy Cummings, professor at Duke University, is something of a celebrity. She is an expert in the field of robotics and human interaction. She speaks plainly and clearly. She doesn’t mince her words when she sees hypocrisy, hubris or hype (Tesla’s Autopilot and Elon Musk come to mind). But above all, I pay attention to those who have lived their lives so close to death.

Cummings was a Navy fighter pilot flying A4s and F/A-18s.

I knew that already. But I hadn’t heard her speak so eloquently of her career change until I watched NTSB’s webcasted seminar.

Cummings began her speech:

I knew I needed to find another job when I transitioned from the A4 to the Hornet.

One of the most striking things to me about the Hornet was, when you're on the catapult getting ready to be shot off, you have to put your hands up in the air and show everyone that you are not touching anything.

It's only when everyone's assured that you're not touching anything, they will actually fire you off the front of the catapult. [This is] because in fly by wire systems where the computer is controlling everything, you're so close to the stall threshold when you're being shot off the catapult. If a pilot touches the stick, it'll set up a set of pilot-induced oscillations.

Cummings noted that it has led to multiple deaths off the front of the carrier. She witnessed firsthand how computerized and automated weapons work.

I flew Hornets for three years. For those three years I flew, one person a month died. That's 36 people I knew died flying a Hornet. They all died not in a military mission. It was always human error and it was always in some training envelope.

That’s when Cummings realized that the systems were becoming “a lot more complicated.” Automation or not, Cummings concluded: “We were also being set up for some danger in areas that we didn't really know what was happening.”

Is a self-driving car safer than a human driver? This is a popular social media meme. I avoid most of that discussion. However, I bookmark some of Cummings’ tweets. She is no empty vessel in discussing “unsafe autonomous systems.”

Enjoyed this article? Subscribe now to continue receiving EE Times Weekend.

Traveling Though Wormholes

Maurizio Di Paolo Emilio

Wormholes connect two points in the universe. Entrance from one end of the wormhole would almost instantly lead to the exit at the other end, even though these two points may be billions of light-years apart. But it’s not as simple as that. The problem lies in not knowing how to get through them.

These “tunnels” first appeared as a solution to Einstein’s field equations. They are also called Einstein-Rosen bridges, named after the famous German scientist Albert Einstein and his then-American assistant, Nathan Rosen.

The wormhole could be built in such a way that one end could remain almost motionless, while the other could move almost at the speed of light. Entering this wormhole would allow coming back in time. A wormhole is a shortcut through spacetime, so sending a light pulse into it could allow faster communication than light.

Also, according to theory, the wormhole connects at least two points to a single gorge, through which we could transfer the matter to another point in spacetime or even to another dimension or universe, according to some scholars, but this needs further research both in theory and experiments.

We can imagine spacetime as something “pierced”: a hypothesis that involves the possibility of passing from one area of spacetime to another, connecting parallel universes.

This type of event could be very useful to control spacetime. We can only live in the present, stepping every second into the future but slowly and without the slightest chance of returning to our past.

One condition, impossible for now, is being able to travel through space at the speed of light: That’s the only possibility for us to slow down our clock and then, once back on Earth, find ourselves in the future. According to Kip Thorne and Michael Morris, physicists at CalTech, the “wormhole” could be used to jump from one point of the universe to another, thus exceeding the speed of light. Does a wormhole connect two distant points in the universe or does it bridge two parallel dimensions? We’ll have to find out.

Enjoyed this article? Subscribe now to continue receiving EE Times Weekend.

The Best Tech Doesn’t Always Guarantee Commercial Success

Nitin Dahad

Last weekend, I was quoted in a national U.K. newspaper about the sale of a small U.K. photonics semiconductor company, Center for Integrated Photonics (CIP), to Huawei back in 2012. The company was at the forefront of research into photonic devices, carrying on pioneering work from both BT and Corning. But it was losing money, so the rescue package to save that research involved finding a buyer willing to invest in it.

The story raises something I’ve experienced throughout my career: The best research or best tech does not always equal commercial success. Two experiences come to mind: the first with Arc International back in 1998 when we were fundraising, the second when leading delegations of U.K. tech companies to the U.S., Brazil, the Middle East, and Africa.

Image: Nitin Dahad/EE Times

In August 1998, at Arc, we spent a week visiting VCs on Sand Hill Road in California demonstrating our wonderful Arc Wizard (that’s what it was called at the time), which could configure a 32-bit RISC processor any way you wanted it. Everyone said “nice” but passed, as I’m sure they didn’t understand how this might equal commercial success and an exit.

In 2009, that experience of taking Arc to the U.S. landed me a role as U.K. government tech advisor for startups and scale-ups seeking global expansion. In working with hundred of companies, again, it was apparent that many founded by engineers and technologists hadn’t quite worked out how unique technologies translated into commercial benefit. Their pitch was invariably just about the technology.

Image: Nitin Dahad/EE Times

The big takeaway is that research and cutting-edge technology is great. But today, when I get a pitch about a great new tech breakthrough, I invariably ask about the business model. Now you know why. Having great leading-edge tech is essential, but so is having a way to sell it.

Enjoyed this article? Subscribe now to continue receiving EE Times Weekend.

Kurt Busch: ‘Following Your Passion is Horrible Advice’

Kurt Busch is the CEO of Syntiant, a California company that makes AI accelerator co-processors for low-power battery-operated systems. The chips are aimed at acceleration of neural networks for voice control in consumer devices.

Kurt’s career has afforded him extensive experience in the world of semiconductors. Before co-founding Syntiant, he was CEO of Lantronix, the provider of smart networking and communications solutions for IT and the IoT. He also served as a senior vice president and general manager of the high-performance analog business unit at Mindspeed Technologies, which was acquired by Macom.

Kurt has worked at Analog Devices, Intel, Digital Equipment Corporation, and two other startups, and he’s an Engineering Hall of Fame inductee at his alma mater, the University of California at Irvine.

Image: Syntiant

What personal projects will you be working on this weekend, Kurt?

This weekend is all about cleaning out the garage. With the Covid lockdown, gyms are closed, so we have been turning the garage into a fitness center. We now have a smart TV for online class, weights, TRX bands, and a rower. It is not nearly as fun as going to a group fitness class, but it is better than nothing.

If you could travel to any time in history, what would you do?

I really don’t want to go back in history. I truly believe we live in the best time ever to be alive. Every day is better than the last. Even though the news would have us believe otherwise, the standards of living, availability of information and entertainment, and overall levels of world peace have never been better. If anything, I would like to go 100 years into the future to see how great it will be.

How do you teach your kids about technology?

I think they teach me. While I have walked through the basics with them, it was watching my youngest son talk to his iPad many years before he could type or even spell that made me realize the next interface to technology is voice. If a toddler can use a computer with voice only, then that could be the path to bring billions of people into the technology age. My children showed me how technology can be the great equalizer.

What was your first job in the industry?

My first job was designing networking cards. I recall with some apprehension my first experience seeing a product I designed coming off the manufacturing line. Every few seconds, one came off the assembly line, and all I could think was, Wow they are making a lot of those. I sure hope I didn’t screw something up.

What advice would you give to those considering founding a tech startup?

Following your passion is horrible advice. Being passionate about something does not guarantee success. I think passion is only one leg of a four-legged stool. I would suggest to them to follow a direction that you are good at, that you like to do, that benefits society, and that someone will pay you for. If you can meet all four requirements, your chances of success are much better and you will feel better when you get there.

Kurt Busch is CEO of Syntiant

Arm's Great Monday | Mercedes-Benz Hands Keys to Nvidia

The perfect way to round up the week, EE Times On Air offers thirty minutes of audio journalism every Friday on the week's most compelling stories in electronics.

On this week's show: the top supercomputer in the world, and displacing Intel at Apple? Arm had a very, very good Monday. We talk with Tirias analyst Kevin Krewell. Also, Mercedes-Benz just handed the keys to its cars to Nvidia. What can we expect now?