Advanced Computing in the Age of AI | Saturday, January 28, 2023

Intel’s New Atom S1200 Low-Power SoC Beats ARM To the 64-Bit Arena, But Can’t Handle Facebook Datacenter Workloads 

At its announcement, Intel boasted that one of the 1200's 20 design wins even swapped out an ARM design for the new chip. The race for fledgling datacenter microserver market is on.

When Intel introduced the Atom S1200, its first low-power Server-on-Chip, it boasted 20 design wins, one of whom swapped out an ARM design for the Intel SoC. But it has not yet revealed who those customers are. What we do know is that, apparently, Facebook is not one of them.

After years of dismissing the market for small low-power chips running smartphones and tablets, Intel now sees them starting to move into microservers as a way to deliver high datacenter performance at lower prices and lower power. It has gotten the microserver religion, writes Stacey Higginbotham in an article on GigaOm. It even released a product roadmap last year in which it predicted that segment of the market could end up accounting for 10 percent of overall server market. With the Atom SC1200, Intel beat ARM chips to 64-bit designs.

The article also refers to Clayton Christensen's seminal book "The Innovator's Dilemma," which pointed out that companies with disruptive technologies are often beaten out by upstarts with newer innovations that disrupt the previous disrupters. It's hard for the older companies to abandon their legacy products. In that book, Intel was an example of a company with disruptive technology -- the microprocessor -- that destroyed the design approaches of mainframe computer companies.

Christensen pointed out that there is huge opportunity for disruptive technologies to move upstream (Intel's processors are now used in some of the biggest supercomputers in the world,) but it's almost impossible for established companies to move downstream (IBM had to go to Intel for a microprocessor when it produced its standard-setting IBM PC, but could never capture much of the compatible PC market it created, nor could it come up with a competitive microprocessor for the market.)

Now ARM chips are the ones moving upstream from smartphones and tablets into the server market, while Intel is trying to move downstream, after failing to act when weaker, low-power chips started dominating the market for smaller devices. So the question becomes: Can Intel succeed in solving the Innovator's Dilemma where IBM failed?

The GigaOm article also wondered about the financial impact of selling cheaper processors, but that has never been a long-term problem for innovators. They sell a lot more of the cheaper chips.

Then again, some of the big companies with huge datacenters don't seem to be willing to take the smaller-is-better plunge yet either. In a follow-up article, Higginbotham notes that, despite the fact that another Facebook executive was touting the new chips on stage with Intel at the announcement, Facebook spokesman Michael Kirkland said the social networking giant does not yet have plans to even test these smaller chips. "Based on how it looks on paper it doesn’t look like it meets the needs of our workloads,” said Kirkland. Ouch.

Still, Kirkland sounded as though the company is very interested in the coming Avoton chips from Intel. "We look forward to Avoton," said Kirkland. Maybe that means the new Atom chips are just placeholders until the Avotons arrive next year.

Add a Comment