Intel believes that it pioneered the microserver - or low-power server - several years ago.
According to product manager Naveen Bohra, the origins of this idea go back to Intel's first Xeon LV, which was shown in 2006 based on the company's Sossaman core; basically a 64-bit version of the very first "Core" processor for mobile devices. Of course, the idea of the microserver today is different and follows different applications trends and cannot be compared to the data center needs five years ago.
In some way, Intel has to admit that it did not exactly create the current trend toward using what are essentially tuned smartphone chips. A similar trend has happened before, in 2000, when Transmeta surprised Intel with its low-power processor.
Intel sources admitted back then that the company's move toward low-power CPUs would not have come as early and as strong as it did without the wake-up call from Transmeta. Today we know that Transmeta was too small to survive against Intel's might, and the late start did not damage Intel in the end.
Flash forward, it is not Transmeta but ARM and vendors such as SeaMicro that came up with the idea that the very low-end of processors - in terms of performance - can be used as server processors. Intel has reacted and announced that it will be releasing Atom S-branded processors for this market, which are upgraded Atom SoCs that support a 64-bit environment, ECC memory and possibly virtualization down the road.
Naveen told us that "no one really knows how the microserver market will evolve," but Intel decided to develop the Atom S as the company "does not want to be left out as the competition comes in."
It is common sense to assume that Intel wants to sell more expensive Xeon processors than cheap Atom chips to its customers, but the company says it will adopt its product line if it sees a business opportunity: "We will not artificially try to protect our main revenue with the Atom S. However, we want to provide the best product for the needs of our customers," Naveen explained. Consider it a strategy to keep ARM as a potentially powerful new rival at distance to Intel's core business as well as a way to expand an existing business into a newly evolving market segment.
So there is some uncertainty, but Intel knows very well why the Atom S can be useful and beneficial in some server environments. The SoC targets the segment below the lowest-power Xeons the company is serving today (Ivy Bridge Xeons reach down to 17 watt TDP) - a range below 10 watt TDP.
Atom S processors are squarely aimed at applications that require high I/O operations, and are positioned between the servers that are responsible for the compute-heavy data crunching in the backend and the end user. Intel envisions microservers to be used for the data delivery to the end user and primarily deliver static content, such as "Wikipedia-like pages," Naveen said. In such an environment, microservers could excel due to their performance-per-watt advantage over a virtualization scenario that relies on overpowered Xeon processors. Atom S microservers will require much less cooling, less space, and will achieve higher densities and end up to be much more efficient than traditional x86 servers.
Intel considers its manufacturing prowess as well as knowledge of the server market as its biggest advantage against ARM. In the past, there was quite a bit of saber rattling between ARM and Intel, but it is clear that both companies have respect for each other as Intel recognizes that ARM's experience and value "is in low-power," Naveen said.
However, he believes that the technology adoption will be made based on customer evaluation, which he hopes will give Intel an edge due to a consistent Intel and x86 architecture in data centers that would enable customers to keep the same software stack and not run in hybrid mode. "It is really about the entire system, and not just about power," Naveen said. If everything plays out in favor of Intel, customers would see an advantage in "flexibility" to make architectural changes fairly easily when using the Atom S approach.