What started as a trickle about the need to think about power has suddenly become a flood. In the past two months, it seems everyone has begun talking about low power.

There’s a reason for this, of course, and it shows just how central the electronics industry is to almost everything else. As the economy begins digging out from the longest downturn in decades—although at least this one isn’t racked by the inventory problems of the 2001 downturn—low power has become a prerequisite for doing business. It’s now a competitive weapon, in fact. Netbook makers are advertising battery life on the screens of new devices. Appliances are brandishing stickers with average yearly operating costs. And car makers are under pressure, if they can survive, to boost their mileage above 35 MPG.

Customers of cloud computing companies are demanding efficient server utilization before they sign on the dotted line, and cloud hosting companies are squeezing computer equipment makers to lower power, not necessarily boost performance. There are far fewer applications these days that need faster performance, other than games, and most of them really need to figure out how to write code more efficiently that can take advantage of multiple cores, or a single core plus a hardware accelerator.

For semiconductor designers, all of this pressure flows downhill. What hasn’t happened, however, is for this pressure to flow to the very front end of some designs. Low power can’t be added into designs. It has to be part of the design from the very conception, including how to minimize power usage, how many sleep states to include, how to utilize power the most efficiently and even what the power source will be.

This is a big change for semiconductor architects, and it requires a different way of looking at problems. But it’s also something that’s not ever going away. Battery technology will remain limited, no matter how much it improves, because there will always be new functions to take advantage of extra power. An iPhone may last days in sleep mode, and it can play music all day long. But play a game or video and it will burn through the battery in a few hours.

Some of the changes will be in software. Some will be in IP. And some will be in hardware. But they all have to move to the very front of the design cycle, and for most companies that’s a long way from where they are today.

–Ed Sperling

For the better part of the past decade, any progress in lowering power was dictated by portability. In the mobile market, battery life has defined some market leaders, and it continues to be a selling point even if Apple seems to have subsumed the bulk of the MP3 player and an increasing share of the smart phone market. At the very least, power conservaton needs to be good enough.

Netbooks advertise battery life of nine hours or more using the Atom processor, and for some applications it’s not unheard of to go an entire day or more without a recharge. Companies like ARM, MIPS, and some of the FPGA vendors like Actel, are working to gain a foothold in this market, as well, which could significantly increase battery life well beyond that nine hours.

But there is equal emphasis these days from the plug-in side of electronics on saving power. If consumers aren’t quite there—most of us can’t tell the power differential from one EnergyStar-qualified flat panel to the next because we don’t have smart meters in our homes—large corporations definitely are. When you run a huge data center, it’s a lot easier to see how much energy it takes to cool racks of servers and how many millions of dollars you can save by servers into deep sleep mode when they’re not in use.

The same type of thinking will be applied to the automotive industry, when car companies finally begin selling cars again. There are massive savings to be had in mileage when you can substitute an electronic microcontroller or processor—even a programmable device—in the place of a mechanical one. Individually, the savings are minimal. In aggregate, they can be significant.

There also is a growing awareness that green is good for the environment, with governments emphasizing power savings and the innovation and jobs that can bring—not to mention less reliance on outside energy—which will new ways of tackling old problems. For example, while solar panels are effective at producing energy locally, using mirrors and solar collectors to heat water that can turn a turbine can generate electricity that can be stored and transported long distances over existing infrastructure.

We are just at the beginning of these changes. Coming out of the downturn, you can expect to see lots more. And from all indications, that wait won’t be long.

–Ed Sperling

For all the loud complaints about the U.S. government sticking its hands into private industry—most notably in the realm of power efficiency—this appears to be more business as usual than unusual.

The federal government has always had a stake in technology development in electronics, dating all the way back to World War II and the Cold War after that. The only difference is that most of the money went into defense first, and then it made its way into the commercial market—usually by way of companies involved in defense contracts in one way or another.

The classic example of that was Bell Labs, which was part of a government-regulated telephone utility. It was funded by phone rates on one side, which were propped up by government regulations, and then subsidized on the other by direct government infusion of capital for defense research. When AT&T finally got broken apart in the early 1980s, Bell Labs was spun off as part of Lucent, which eventually got bought by Alcatel—a French company that made headlines when it turned around and sued U.S. companies for patent infringement.

>For the next two decades, most of the U.S. government’s capital infusion went directly in defense or aerospace, which is at least partly defense-related. The fact that public funding is now being applied to things like electronic medical records (a massive power efficiency move from a holistic standpoint), better battery technology, the power grid and automobiles in general is only different in terms of the markets to which that money is being applied. And there is at least a good argument that much of that is defense-related, as well, by removing the U.S. from its dependence on foreign oil.

So what’s changed? Probably just the openness of the funding process, the scale of the overall effort, and the names of the companies. But in relative dollar terms, a lot of this looks very much the same.

–Ed Sperling

Intel’s pending acquisition of Wind River goes to the heart of low-power design—customized software designed specifically for cores that most applications can’t use.

There’s been a lot of speculation about exactly what this means, but much of it misses the point. Intel wants to keep adding more cores into its chips because it’s too expensive and entirely too limiting to try to build chips with fewer cores. The cores hotter at each new process node, and solutions like Turbo Mode—basically supercharging a single core—or accelerators are one-time gains.

Intel needs a longer road map than one node. It needs to be looking out at least three nodes at any point in time, because its strategy is to create overlapping development teams. The only way it can get there is to figure out how to more effectively use the cores it’s planning on putting into the chips, which means it has to take over some of the software development.

And not just any software. What Wind River is best known for is real-time operating systems (RTOSes). It’s custom mapped directly into whatever cores Intel makes, and those cores have to be the right size to limit power consumption. To a large extent, this is the modern-day extension to an integrated device manufacturer. At future nodes, if you really want to innovate on a platform, you don’t just have to own the fab and the process. And if you really want to reduce power, you have to own the software, too.

The basic premise is that a general-purpose operating system has many uses, but it’s not nearly as efficient as an RTOS. Considering most cores in a chip are asleep most of the time, waking up and going back into sleep mode, or deep sleep mode, should draw almost no power. That allows more power to be diverted to the primary applications, which at least for now will probably continue utilizing the general-purpose OS, whether it’s Windows or Linux or OSX.

This also has significant implications for Intel’s new Atom processor. Sources inside Intel tell us that Atom definitely is on a multicore path. And considering Atom’s biggest market limitation right now is the fact that it’s base power consumption is 2 watts—or more like 5 watts when all the I/O layers are added in—customized software will be able to significantly reduce power consumption and open up more markets.

Embedding Wind River’s software into its cores with much more exacting specs means that Intel will be able to target solutions—based upon both power consumption and performance—for specific markets that it could never reach with previous versions of its x86 processor.

It will take several years for this strategy to completely unfold, but Intel is all about road maps and a statement of direction. In this case, that direction one now taken by some of the largest names in the embedded processing world.

–Ed Sperling