Mobius Loop Promises Cooler, Quieter and Cheaper Data Centers

Typical chips generate timing signals by pulsing, which takes a lot of power and generates too much heat. Both are problematic, as power is expensive to use, and heat is expensive and noisy to remove.

That one-two punch forces network managers to adjust accordingly. Especially those overseeing, say, 10,000 mail servers, like Hotmail’s Phil Smoot. He notes that heat management is a crucial part of stack design for Hotmail engineers:

We spend a lot of time analyzing the specific SKU for the application we need — in case, for example, we overbuild some sort of X box for which we thought the application requirement was, say, two processors, and it turns out it was one processor. Those processors suck a lot of juice and require a lot of cooling and floor space and therefore limit how much can be deployed over time.

Juice-sucking processors have also played a major role in swapping the fortunes of AMD and Intel, reputedly forcing Intel to shelve an entire line of blisteringly hot Pentium CPUs.

RotaryWave Traps Power in an Endless Mobius Loop

That’s why Will Strauss, Principal Analyst at market research shop, Forward Concepts, insists that existing clock designs must change. “The importance of low-power clocks cannot be overstated,” he says, “It is critical to lower the power burden imposed by existing clock designs.”

Multigig Technology thinks it has the timely answer in its new RotaryWave technology. The design uses Mobius termination to, in effect, recycle a charge, which allows RotaryWave to extract more work for the money. Enough, Multigig claims, to reduce power required by around 75%. [Mobius animation]

As an added bonus, Multigig claims that no manufacturing changes are required to implement RotaryWave clocks into a design.

As chipmakers implement RotaryWave, power, cooling and square footage requirements for datacenters will shrink. Some cool and quiet servers will migrate from server rooms to desktops. And silent PCs might truly become silent, without sacrificing capabilities.

Strauss says RotaryWave proves “what creative thinking can do to address the issue and enable true low-power, high-performance chips using existing design methodologies and process technologies.”

So when can you start designing around the new paradigm? Optimists estimate two years, pessimists push it out past five.

On the other hand, with guys like Microsoft’s Smoot ready to pounce on muscular cost savings, Intel may be tempted to jump on RotaryWave early… Or perhaps AMD will outmanuever them again.

Like Smoot says, “A penny per user over hundreds of millions of users gets expensive fast.”

Same logic applies to his servers, but in this case, the savings will be bigger than a penny per.

Email Battles Backgrounder:

• A Conversation with Phil Smoot; Ben Fried; Systems of Scale; ACM Queue; December 2005.