It’s rather fortuitous for the semiconductor industry that global warming became a raging issue over the past half decade. Without it, new chips would be a much harder sell.

Classical scaling ended at 90nm—meaning that for every new node on the Moore’s Law road map you no longer got an associated boost in performance—because it was simply too hard to continue making complex SoCs with a single processor core. They ran too hot when they were on, and they continued to run even when they were supposed to be off.

This problem wasn’t just confined to devices like smart phones, either. Inside of enterprise data centers, heat costs money. It requires additional cooling to make sure data doesn’t get corrupted. The solution for chipmakers like Intel and IBM was to break a chip down into multiple cores, each running slower than a single core. But without the advantage of parallel application software—even games still use only one or two cores—that offers no great advantage.

Enter the global green movement—and just in the nick of time. Without it, there would be no compelling reason to buy new chips.

Virtualization has taken care of what to do with multiple cores in data centers. Server utilization is up as much as 25 times, which is much more efficient use of power and cooling resources. And it’s about to have the same impact in handheld devices, where it can be used to simplify low-power designs.

But that wouldn’t actually matter if people weren’t suddenly focused on energy. If consumers were only concerned about performance, there would be no impetus to move to the next process node. Different functions could still be built into chips on the same board instead of inside the same chip, and battery life could be extended by simply using a bigger battery.

Our progress in the past half decade owes a lot to consumer awareness about the need to conserve power and rising power prices because of the difficulty and expense of building new power plants. Fifty years ago, if power ran out companies would simply build another plant. Now they raise prices.

Put in perspective, this is all good. It’s good for companies making more efficient electronics and it’s good for the people consuming it. But when you really consider which came first, it seems that it wasn’t so much the electronics industry saving the planet with increased power efficiency. The compass needle points in favor of the planet saving the electronics industry.

–Ed Sperling

Buried deep within Intel’s new Xeon announcement is an interesting piece of irony. You can now save power by consolidating machines and you can boost performance with new machines, but if you’re just buying a new server you can’t necessarily get big gains in performance and slash power.

Granted, by reducing the number of servers through virtualization you can always save power. But the speed of the clock—basically how fast the entire machine will process data—is the determining factor as to how much power a Xeon server will actually use. At 3.46GHz—a frequency that most of us never expected to see again—the chip will draw up to 130 watts. At 2.66GHz, it will draw 80 watts. And there is a vague promise of future versions that will run as low as 40 watts.

What makes this intriguing is the series of tradeoffs being made at the enterprise computing level. Now that virtualization has pretty much guaranteed more cores can be utilized—rather than waiting for all applications to be parallelized or threaded to use more cores—performance guarantees are no longer included with each new rev of a chip. You can run more applications on more virtual machines, but you don’t necessarily get more performance with more cores. If software is parallelized, six cores could mean six times the performance—or at least close to that number. With virtual machines, six cores means six applications running on the same machine at the same time, and often at the same speed.

That puts processor makers back on the fast track again—Intel and all of its rivals, including AMD in the enterprise, and ARM and Apple in the portable device world. If they can’t sell lower power—there are always some gains at each new process node, of course—then they have to sell more performance. And they have to figure out ways to save energy elsewhere in the machine—everything from software to the overall package, board layout and in the multiple on/off and sleep states.

Intel bought Wind River at least partly for this reason. You have to wonder who’s next on the acquisition list—and just how far into the electronics supply chain the company and its rivals are willing to reach.

–Ed Sperling

The introduction of multicore processors in a slew of battery-powered devices is an interesting development. The ARM processor now comes in quad-core configurations, and the Intel Atom processor is now shipping in a dual-core configurations.

We can only assume there will be more cores at each new process node, and probably more cores added at each rev of existing process nodes. But what do we actually do with all those cores.

Aside from threading applications onto two or even four cores, the extra cores are largely wasted. As Freescale’s Lisa Su pointed out, there’s a big difference between adding eight cores and getting eight times the performance. Or in battery-powered devices, maybe it’s a question of achieving higher performance and longer time between charges.

The glaring disconnect in consumer electronics engineering is that we know how to create the cores—and that’s no small feat—but we don’t know how to effectively utilize them. In the plug-in server world the answer has been virtualization, because very few applications are parallel enough to take advantage of multiple cores natively. Databases and some graphics applications are the exception.

In the consumer world, the number of applications that can utilize multiple cores effectively is far fewer. That leaves end-device manufacturers scrambling to find uses for those cores. In some cases, the solution has been assigning specific cores for specific functions, and matching power and performance with application need. In most others, designers are left scratching their heads.

The addition of virtualization to consumer electronics may change all of that, but it also may change the dynamics of engineering these chips. There will be additional burdens on the software engineers to manage power, and there will be additional burdens on the hardware engineers to make sure there are no resource shortfalls caused by poor prioritization. And all of this will create more challenges for the verification engineers, whose ranks will need to grow significantly just to get chips out the door.

The engineering community is about to cross over from just creating the chips to figuring out how to use them more effectively—at the architectural and design level. And from here on, things should get very interesting.

–Ed Sperling

Green is good. Everyone says so. Large end-user companies increasingly are looking for green bills of materials and consumers are demanding green products.

But what exactly is green? While the entire semiconductor design chain is focused on reducing power, the real difference between green and non-green is more than just power consumption. If it takes more energy to generate a solar cell than it ultimately saves in electricity, then that isn’t exactly green. And if a distributed smart grid allows consumers to use more energy than they would if they had to pay for every kilowatt/hour from the power company, it’s questionable just how green that approach is, too.

Carbon footprints are even more confusing. While ecologists expound on the advantages of being carbon neutral, the reality is that basic survival puts people into the plus rather than the negative camp. Unless you live on the equator, fish for a living (with homemade hooks made from bone you carved out of wild boar that you killed with a rock) and make your clothes from plant fibers, you’re probably well into the carbon-positive world.

The reality of green is that it’s a step in the right direction. Carbon footprints will never really be neutral or negative, and nothing will ever been completely green. But cutting power, improving time between battery charges, decreasing energy costs—particularly in large, energy-intensive sectors such as enterprise data centers—are all important. Aside from the environmental impact, which is significant even though it’s the subject of fierce debate, there is the unequivocal convenience factor of fewer battery charges, longer battery life, and less money spent on powering devices of all sizes.

As a marketing term, green is very subjective. And for the semiconductor industry, it may be a moving target defined by the companies that can offer the most functions and features at the lowest power the same way, a decade ago, the benchmarks veered into a war over performance. What remains to be seen is whether consumers will really be willing to trade in performance for being even more green, or whether there will be a balance struck between being green and getting the job done at the same or greater speeds. And perhaps even more important, just how much extra they’re willing to spend to have a brighter shade of green.