As late as last year, chipmakers were predicting the future was in adding more and more cores to SoCs, processors, and even FPGA. It was almost like Wall Street investment in 1928, 2000 or 2007.

To paraphrase Herbert Hoover, there was no reason that the stock market couldn’t continue to rise forever. I’d couple that with the words “paradigm shift” or “sea change.” Anytime you hear those phrases it’s time to start building a bunker and stockpiling essentials.

While cores running at lower clock speeds do indeed use less power—and smaller cores running at slower speeds use even less power than big ones—there’s a limit to how many you can use effectively for most applications. Databases, some corporate applications, and some math and scientific problem solving are the exceptions. After that you’re largely relegating separate functions on different cores, with shared memory, shared busses, cache and all the things that don’t necessarily play well together.

In fact, the picture that’s emerging of multicore designs looks less like a multiprocessing concept and far more like a multi-power domain management scheme. It’s rare that more than one core is operating at any time in most devices, and it’s even more rare that you’ll see more than two operating.

In fact, the latest approach to improving speed in devices is to add accelerator technology in one of the cores. That’s because you can’t build a chip at 45nm or 32nm and run the clock fast enough to add major performance gains. For anyone who’s old enough to remember Intel’s 486 processor with the 487 math co-processor, this is roughly the same idea. The FPGA vendors have been experimenting with this technology in their labs, and Intel’s upcoming turbo boost builds on the same concept.

And if you go far enough back, this begins to look a lot like time sharing on computers that’s been internalized, so you don’t need a sign-up sheet anymoreOnly the time management is no longer about different people using the computer. It’s about different applications or functions using the power, because the computers are now sufficiently shrunk down that they involve single cores of a chip.

That may say something about the next big thing or the next paradigm-shifting technology. We’ve made huge strides in technology, particularly in terms of shrinking designs and conserving power. But if you want to figure out what’s coming next, it might help just to look backward.

What do you think?

–Ed Sperling

One of the great things about low-power design is being able to see how what happens at the sub-micron level translate into macro changes that affect our everyday lives.

Semiconductors already have changed the way we live, the way we transact business and how we communicate—sometimes for better or worse. Low-power design continues to drive down the cost, extend battery life and keep all of us on our toes to solve some of the most challenging applied physics problems that five years ago used to be in the realm of theoretical physics.

But how all of this applies to the overall electric grid raises some questions that have never been fully resolved since the days of Thomas Edison and George Westinghouse. Edison believed firmly in direct current because it was safer. Westinghouse believed in alternating current because it worked better for long-distance transmission. Things became so contentious between the two of them that Edison created the electric chair to prove that AC is dangerous. (As with many inventions, the electric chair had different applications than the original concept.)

More than a century later, though, we’re back to wrestling with the same issues. If energy can be produced locally using solar farms or rooftop cells in residential areas, then how do you get that into the grid? And if you’re already on the grid, how much intelligence to you add further down into appliances like air conditioners to reduce their power draw during peak hours.

All of this works like the energy management features on a chip, only on a much, much bigger scale. There is voltage stepping up and down, depending upon demand, and there is dynamic current management and clock gating. But so far, the grid still falls prey to brownouts, blown transformers and surges that cause untold damage.

In short, the grid is only as smart as the various components on the grid, and so far there is no consistency to how they’re designed, matched or where they’re located. Maybe—just maybe—the architects of the nation’s power systems should take a few cues from the semiconductor world—as well as a lot more components to manage it every step of the way, from generation to delivery at the customer site. They might even find things would work a lot more smoothly if they hired some chip designers.

–Ed Sperling

There’s been a lot of marketing hype about what age we’re in. According to leading experts we’ve gone from the Information Age to the Computer Age to the Networking Age to the Internet Age.

My humble response, which would have been delivered with bowed head and a nicely handwritten note if I actually knew where to send it: Baloney.

The information age is still in its infancy, and all the other pieces are simply enablers for the movement of information. Computers allowed information to be processed faster. Networking allowed it to move from one desktop to the next and from one company to another. The Internet allowed it to move everywhere.

The next phase of the information revolution will be information that moves between machines, or between machines and people and back again. It will be in trees that communicate over a wireless network when there is a fire or a device embedded in the human body that communicates to a medical office without any conscious effort.

What isn’t obvious in all of this is just how important low-power design has become. A key part of the next phase of this revolution will be low-power wireless communication using energy scavenging, extremely long-lasting batteries and more efficient software. A second part will be much better communication that eliminates the need for some travel—essentially a human carrier of information. And a third part will be much better security, made possible by things like smart cards that draw extremely low power.

Information has always been the key driver of semiconductors. Nothing has changed, except the perception of exactly what age we’re in. My guess is this is the marketing age, but even that’s based upon the dissemination of information.

What do you think?

–Ed Sperling

Why is Apple developing its own chips?

I’ve posed that question to a lot of people in the past week, since news began leaking out that Apple was hiring system-level designers. From all the pieces that have begun to surface, the answer isn’t quite as simple as it might appear.

What appears to be solid is that Apple is exploring the possibility of creating its own low-power chips for mobile applications. It did the same back in 1989 with the PowerPC chip that it co-developed with Motorola and IBM, then its arch rival in the PC space.

The new chip development may or may not include a whole range of lower-function devices such as netbooks, but if it happens—and nothing is certain until it’s certain—it will include the iPhone and various other devices of that scale. And in typical Apple fashion, this may be a way of delivering more function for less power because this way Apple can tune the software to the hardware, and vice versa.

But that’s only one piece of the puzzle, albeit an important one. Being able to offer a high-performance device with a significantly longer battery life is a competitive direction that more companies will be taking. Virtually all the major chip companies say there are orders of magnitude improvement possible by writing software more concisely instead of having to worry about decades of backward compatibility.

Apple’s exploration has to be put in context, though. There is a fundamental restructuring of the tech world under way, and it will affect all of the chips that are designed in future years. In past downturns, companies with cash in the bank typically have bought up smaller companies and expanded their technology portfolios. What’s different this time is companies with cash are moving into different markets, driven by both convergence of technologies into a single device and the commoditization at the device level.

This is why Oracle is buying Sun, most likely to create tightly bundled application-specific hardware, and it’s why HP is challenging Cisco and why Cisco is now pushing into the enterprise server market. It’s also why over the next two years most cell phones in other parts of the world will contain GPS systems, rather than using standalone devices with specific functions, and why the current GPS makers will have to start offering other features or die a swift death.

So what will make one device stand out from another? More than likely, it won’t be functions. Most will offer the same functions, although some will still be better than others. The more compelling reason will be better battery life, better interfaces and better business context—namely cheap and plentiful applications—and where none of that matters, the differentiation will be cost.

These changes already have started to ripple down from the enterprise server market to the consumer world. It likely will infiltrate the component world, as well, where the quest for lower power and integrated software will become the next big challenge. And the next big thing won’t be a killer application or a new form factor. It will be solving that challenge better than the next company.

–Ed Sperling