The price of gasoline at the pump is creeping up again. In fact, it’s almost 50% higher than a year ago, when the economy was in a major slump.

While this may make the average consumer cringe, it’s enough to start revving the engines of change for low-power engineering again. Higher oil prices will reinvigorate interest in energy-saving automobiles and consumer devices that can get better mileage and better battery life, respectively. They also will feed into a slew of industrial products aimed at lowering power bills and saving the planet. (Well, maybe save the planet, depending on just how strict the rules become in developing economies.)

Some of the momentum toward these changes faded during the past 12 months as oil prices dropped and the economy hit bottom, but with growth returning that should change. Companies still need to cut the power consumption of their data centers, which now consume so much power that the cost of cooling machines actually exceeds the cost of the hardware. And cars that can run on electricity and/or some other type of fuel are expected to start rolling out the door in 2010.

The hurdle that needs to be overcome is price, and that’s where economies of scale come in. It requires widespread adoption to knock prices down to what’s required for mass-market adoption, which is something that never happens in a downturn. But in an upturn, given the rising cost of energy and the fact that many companies and consumers have held off on all but the most necessary purchases, things can change dramatically.

That includes everything from new types of light to variable-speed motors, research into stronger and lighter materials and re-engineering just about everything on the planet for lower energy consumption. All of this is poised to change in the next 12 months, which like all upturns has a remarkable brainwashing effect when it comes to memories of a downturn. Fasten your seatbelt.

–Ed Sperling

ARM’s ecosystem has spoken. What it’s saying isn’t always quite so clear, but there’s a lot of talk about low power and some mumbling about performance.

At this week’s TechCon3, aka the ARM Developer Conference, there was plenty of talk about ecosystems and industry support for ARM. What most of the supporters danced around, however, was the war between ARM and Intel. No matter how you slice it, some of the tools made by ARM’s most ardent supporters are also used by ARM’s biggest competitor. This is marketing diplomacy at its best.

But the message on ARM’s side, at least for the moment, is that low power rules. The company is working on boosting the performance of its multicore processors, but it hasn’t slacked off one iota on power issues. In fact, even if Intel cuts the power consumption of its Atom processor in half it won’t keep up with the active power utilization of an ARM processor.

In some applications, power is more important than performance. In others, no matter how much ARM improves its performance it won’t be able to displace Intel. But there is also that developing netherworld, somewhere between the computer and the mobile phone, where sales are likely to explode because it’s cheaper, the battery lasts longer and it provides decent functionality. It’s the market where you don’t need a full computer but a smart phone isn’t quite enough, but where you also can type like a normal person instead of some chicken-fingered mutant.

This is fertile ground for both companies, and given the growth of these devices in business applications as well as in price-sensitive emerging markets, there may be room for both. In fact, there may be billions of opportunities that so far have never been tapped—just in case you’re wondering what all the fuss is about.

–Ed Sperling

Until now, most of the pain felt in low-power engineering was borne by companies making advanced chips that were highly dependent upon battery life and which were produced in sufficient volume to make the cost tolerable. As we emerge from the worst recession in memory, things will be different. Everyone will be feeling that pain.

There are several reasons for this. First, while the global economy was down, oil prices were up. They’re still up, although not nearly as high as they were when oil topped $140 a barrel. They’re now hovering around $75 a barrel and trending upward. That has provided great momentum for politicians to push a green initiative, and it has spurred consumers to seek out energy-efficient alternatives to just about everything.

Those who predicted several years ago that green was here to stay were right. In the future, given the alternative, buyers will opt for lower energy costs over the life of a product. And given the jump in energy costs versus interest at the bank, energy efficiency is probably a better long-term investment.

Second, mainstream is no longer 180nm. More and more companies are now pushing down to 65nm and beyond, which is where the concern about power budgets really begins. And it only gets more critical at each additional node. For the companies working at 32/28nm, power is at least as critical as area, and probably more important than performance for most applications.

This trend will only continue. The tools are there and the processes are mature enough to warrant the shift, but that doesn’t mean architects and design engineers won’t have to wrestle with some of the same issues in low-power design that the cell phone chip makers have been dealing with for a couple of years.

Third, power is now a competitive weapon for marketing departments. If Samsung’s numbers are correct and 33% of consumers are willing to plunk down money for an extra hour of battery life, think of what they’ll do for four or six hours of extra battery—or even a couple hours more while using power-hungry applications.

This will drive an increasing number of developers to include the lowest-power techniques available. And it will mean there is finally a market for power modeling and some of the complex verification that needs to be done with multiple power islands and voltages.

All of this is good news for the providers of low power chips, tools and IP. Pain forces design engineers out of their comfort zone to learn new techniques and approaches to solving problems. What’s changed this time is that the pain relief will be much shorter lived than in the past, so while there are fewer design starts overall there will be an almost constant learning curve for everyone involved.

–Ed Sperling

It’s a known medical fact that f you can lower the temperature of the human body by one degree, chances are pretty good that a person will live an extra 10 years.

There are a number of studies to that effect, because not everyone’s body temperature is the same. Those whose body temperature is naturally lower tend to live longer and often better lives, assuming roughly the same environmental conditions. That doesn’t mean if you cool it down abnormally you’ll survive longer. You may work harder to maintain an optimal temperature and die earlier from the induced stress.

The same idea applies in electronics. If you can design a semiconductor with lower the voltage in a device it will last longer. The challenge comes in the design, though, and so far the semiconductor world hasn’t seen a whole lot of effort in power modeling with multiple power domains.

How important this is depends upon the application, and there’s a supreme bit of irony here. Typically designs where engineers are most likely to experiment with advanced ways of cutting power are the ones in the highest volume, namely consumer devices. And for the most part, no one expects those to last very long. A three-year lifespan for a smart phone is a long time. Most contracts only last two years.

Even automobile manufacturers only specify to battery makers that the batteries last as long as the lease period. They don’t want to have to change them out during that time. Beyond that, life expectancy on a car battery is questionable. It’s the same with brake light bulbs, if you’ve owned a car longer than a typical lease. If you’ve ever replaced one, you’ll notice the other one typically goes out within weeks of the first one.

So where is all this leading? First of all, low power designs are good for a lot of reasons. They save battery life, which is a convenience. What most vendors aren’t recognizing, though, is they also can last a lot longer—meaning that if you can compile the right kind of data you probably can sell the longevity factor and change the economics for the components.

The chip industry has a long history of giving away technology at the lowest possible cost with no premium for additional value. Lower-power designs are more complex, require more hours for verification and more engineering dollars to create. They require more third-party IP and more up-front design. And that should be worth something extra.

Moore’s Law is an economic equation based upon the price of transistors. There’s no reason that lower power has to mean lower cost when it adds longer life and better reliability. And there’s no reason that marketers can’t capitalize on this in a very big way.

–Ed Sperling

Silver bullets were invented to kill werewolves. Now they’re supposed to solve complex problems in business and technology. And in the semiconductor design world, they’re supposed to simplify everything from architectural design to implementation and verification.

If you believe in werewolves you might believe in silver bullets, but most of the people working on 22nm designs don’t believe in anything except more restrictive rules, incremental improvements in materials and processes, and an awful lot of difficulty in defining coverage models for verifying those designs and major headaches in getting heat out of new devices.

New materials with better insulation properties can help somewhat. Limiting shapes and placement on chips can help, too. So can exotic technologies like air gap and new substrates like fully depleted SOI. But the challenges are growing by the node. To area, power, performance there are now issues of validation and functional verification, and even then the designs are becoming so complex that reliability is in jeopardy.

Never have so many factors converged on advanced nodes. This is an industry filled with some very smart people, and given enough time and people these kinds of issues will be solved one at a time. But what’s changed is they won’t be solved at 28nm and beyond with any single solution. You can’t just change lithography or interconnect materials or gate structures or insulation technology and expect to move on to the next process node. Now you have to do all of these things and more at every node—and make even more tradeoffs about area, power and performance, manufacturing processes and materials.

We need to get rid of the notion of silver bullets once and for all. There’s no place for them at future process nodes.

–Ed Sperling