High Energy Costs Drive Processor Innovations; Retail Electricity Regulation Stifles Them

Lynne Kiesling

Have you noticed those AMD processor ads in airports, magazines, etc., that say things like “this processor saves enough power to steam the milk for 1,000 lattes”? I am fascinated by the fact that AMD sees value in a general marketing campaign focused on energy saving features of processors! We would see more such innovations, and more ubiquitous use of such innovations, if we abandoned the false notion that retail electricity regulation keeps energy costs low.

Why am I fascinated?


Because like electricity, computer processors are not things that most people think about, even though their use is ubiquitous. I can only think that the target audience here are corporate executives (CIO, CFO level) who care about maximum performance at lowest cost, where lowest cost doesn’t just mean cheapest processor.

I am also fascinated because for the past couple of decades, energy costs have been so low that the amount of electricity that devices use has been a very low-priority issue. But two things have changed to realign those priorities: computing scale has increased by so many orders of magnitude that energy and waste heat management in places like data centers is now important (data centers are hot, and the money spent cooling them has become substantial), and energy costs have risen (see, for example, this AMD ad from May on Slashdot, going right to their target market).

AMD’s marketing materials use what I think is a great metric to get out in general use: performance-per-watt. Look, for example, at this press release from March 2006 about their dual-core Opteron processors:

Performance-per-watt leadership
In addition to offering native dual-core technology, the AMD Opteron processor helps deliver a significant savings for enterprise datacenters. Dual-Core AMD Opteron processors offer significant performance gains while operating in the same power and thermal infrastructure as single-core AMD Opteron processors. This means fewer servers can do the job of many, to lower overall operating costs. AMD PowerNow!™ technology with Optimized Power Management can increase those savings even further because it enables servers and workstations to dynamically power down processors, based on usage. AMD PowerNow! technology can reduce CPU power consumption by up to 75 percent during idle time, to help decrease the strain on datacenter cooling and ventilation systems and helping minimize overall power consumption for enterprise IT and workstation customers.

How cool (no pun intended!) is that? I look forward to a time when all devices, not just enterprise-focused ones, tell you performance-per-watt. For end-use consumer devices, the best way to enable this kind of interest and conservation is through retail competition and market prices in electric power.

A third driver that I find fascinating is the role of competition in bringing this energy saving capability about. AMD and Intel are fierce competitors in the processor market (an excellent example of a competitive duopoly). They have both been developing dual-core technology, and while it’s not as well-phrased as “performance -per-watt”, Intel’s description of their dual-core chip focuses on energy efficiency:

Its enhanced voltage efficiency supports cooler and quieter system designs as compared to traditional desktop and laptop PCs. And thanks to the innovative energy efficient technologies built-in, the Intel® Core™ Duo processor is able to transfer power only to those areas of the processor that need it, thereby enabling laptops to save power and desktops to have thinner, sleeker designs.

Note also that Intel’s focus in this language is on desktops and laptops, not enterprise servers (although they are certainly active in both, and what I’m about to say applies to servers too). And that brings up a really important point: energy cost is one driver of this innovation, but so is consumer experience. More intense processors has meant bigger and louder fans and more waste heat, which has meant bigger and hotter devices. As far as I know the story, one of the biggest drivers for Apple’s shift to Intel’s Core Duo processor was the waste heat under the hands of laptop users (I can vouch for that heat!). I was at the Apple store on Monday and I can attest that the new laptops are much cooler under the hand. And, as the Intel language suggests, lower cooling requirements mean slimmer devices.

Thus lowering energy costs and improving the consumer experience are aligned.

Back to AMD-Intel competition: this fascinating article from last week highlights Intel’s current dominance in the performance-per-watt space, their continuing innovation to maintain that dominance, and AMD’s innovation to overtake that dominance.

If AMD has its way, then Intel’s performance-per-watt leadership claim may be short lived: The first quad-core Opterons, due in 2007 will consume the same power as today’s dual-core processors, while AMD expects Intel to increase the power envelope when adding more cores.

Intel disagrees with that claim, as the article goes on to say. But the point is this: the competition between them to reduce energy costs while enhancing performance is driving these innovations. Such technological change has an enormous role to play in bringing about conservation and energy efficiency. But until we have robust retail electricity markets, the ability and incentive to use such innovation to drive conservation and efficiency is limited by the artificial political restraints placed on retail electricity prices. This misguided approach is counterproductive, because it keeps costs higher for longer. Retail rate regulation does us no favors, because it reduces our incentives to implement more innovations like these. It perpetuates the false perception that we are keeping costs low now, but by doing so we don’t adopt the technological innovations that drive the true, long-term cost reductions.


4 thoughts on “High Energy Costs Drive Processor Innovations; Retail Electricity Regulation Stifles Them

  1. Robust: Here is a term that used to mean ‘gangly and myriad’ (so for example ‘robust data’ would be worthless) and has come (through coffee ads, I suspect) to mean ‘{some combination of} richly flavored or endowed, tolerant, tough and vibrant.’ What’s your defining etymology for that, anyhow?

    Because we’re probably in for a bunch of it; utilities are making a hideous hamhanded show of annual tree-trimming by lines and lethality of their rights-of-way, while economic awareness is telling us that the meter is much at game. Generation is still up there as far as skill level, but not running the Research reinvestment level of Semiconductors by far; simply selling up the training ladder a bit will probably help a lot.

    Do we even want to use the old rights of way and provisioning domains?

  2. Profits: Apple and Intel Going Opposite Directions

    Lynne Kiesling Earlier this week Apple announced its quarterly profits for Q2; profits increased by 48 percent. Unexpectedly continued strong iPod sales combined with a 61 percent increase in sales of laptop computers. What the WaPo article doesn’t say…

  3. Profits: Apple and Intel Going Opposite Directions

    Lynne Kiesling Earlier this week Apple announced its quarterly profits for Q2; profits increased by 48 percent. Unexpectedly continued strong iPod sales combined with a 61 percent increase in sales of laptop computers. What the WaPo article doesn’t say…

Comments are closed.