2 Comments

Summary:

In the debate over the future of server hardware, it comes down to the need for highly efficient hardware using up a fraction of the space and a fraction of the power of legacy hardware solutions, versus the desire for more powerful options from existing manufacturers.

Stacey Higginbotham (GigaOM), Barry Evans (Calxeda), Andrew Feldman (SeaMicro),Don Newell (AMD), Omid Tahernia (Tilera) - Structure 2011

Stacey Higginbotham (GigaOM), Barry Evans (Calxeda), Andrew Feldman (SeaMicro),Don Newell (AMD), Omid Tahernia (Tilera) - Structure 2011At GigaOM’s Structure conference, hardware executives from Calxeda, SeaMicro, AMD and Tilera battled it out over how the hardware infrastructure of the future will take shape. For the most part, the debate came down to the need for highly efficient hardware that used up a fraction of the space and a fraction of the power of legacy hardware solutions, versus the desire for more powerful options from existing hardware makers.

The panel ended with a bit of “show and tell,” with Calxeda CEO Barry Evans showing off his company’s first-gen product, a wafer-thin board which he said contains the performance of a server, but uses only 20w of power. Seamicro CEO Andrew Feldman followed that up with his own product, a slightly larger blade. Unlike Calxeda’s prototype, which will be in testing later this year, Feldman talked up how his product was actually in use today. “It’s bigger, but it’s here and it’s available,” he said.

Not to be left out, Tilera CEO Omid Tahernia, while not brandishing a piece of hardware, said his company could provide more performance for the same footprint if vendors used his massively multicore chips. But he also said that it was important to be able to provide what customers wanted from a performance standpoint.

While all the others were busy comparing the size of their server blades, Don Newell, AMD VP and CTO of server products at AMD, also preached the importance of having multiple options, including more powerful hardware. He also played down the move to new server architectures based on processors from ARM and other vendors.

“People are talking as if x86 was in the past,” Newell said. “When you ask which architecture will dominate over the next few years, anyone who thinks x86 will not dominate is severely confused.”

  1. Lucian Armasu Friday, June 24, 2011

    Will x86 chips dominate “in the next few years”? Absolutely. But ARM is the future and will gradually disrupt the x86 chips market.

    Share
  2. Cost is the third (and IMO, most important) leg of this stool. Take your all-in costs (space, power, hardware cost, support cost, migration cost – e.g. you must change your code to support new architecture, etc), and make the simple choice, “which architecture gives me the necessary compute at the lowest cost”. Further, determine your variable cost vs. fixed… energy is volatile, hardware cost is fixed. If you (or your company) has a energy volatility strategy (hedge, ppa, self-gen) maybe you don’t care much about energy cost fluctuation. If you are living on the spot market maybe you just go for lowest cost solution today and roll the dice. Nevertheless, lowest power consumption for equal compute will almost always come out ahead.

    Share

Comments have been disabled for this post