Qualcomm jumps into the increasingly crowded ARM server market

Unlike the mainstream server market, ARM servers have a number of competitors.

At its annual meeting with financial analysts in New York, Qualcomm announced it planned to pursue the ARM-based server market, a market that's rather crowded these days.

Qualcomm is wading into a field already occupied by AMD, Marvell, Texas Instruments, Cavium and AppliedMicro, which have all introduced ARM server CPUs. Granted, Qualcomm is the 800 pound gorilla in the room but those companies won't just roll over.

Barron's reports that Qualcomm CEO Steve Mollenkopf said the requirements that have nurtured the architecture of mobile device semiconductors is becoming more and more relevant to the way server chips must be designed. “The high end of the smartphone and the tablet really are starting to merge with what would be feasible in the data center,” he is quoted as saying.

To a point, he's repeating what has been said for some time now, starting with SeaMicro. When it launched, it said its servers were meant for specific tasks, like edge serving, file and print or light HTML work. A Xeon processor was far too much power to be serving up JavaScript or HTML, but an Atom, which SeaMicro used, was more than enough and saved a lot of power.

For heavy data processing like RDBMS, ERP, data warehousing, business intelligence and Big Data, you are not going to use an Atom or ARM. That's petabytes of data that need to be processed and those mobile chips simply aren't going to get the job done any time quickly.

Still, edge servers will be important, especially with the rise of the Internet of Things. Mollenkopf noted that the data center market is projected to reach more than $15 billion by 2020 and Qualcomm wants a piece of it.

One reason Intel isn't sweating the whole ARM-on-servers effort is it hasn't caught on. Part of that is due to the fact it takes a long time to make a server chip from scratch. When the idea of taking ARM processors to servers first kicked up a few years ago, ARM was a 32-bit processor with no chipsets to support things like ECC mode.

Since then, a whole lot of R&D has had to be done to make ARM server-capable, and it's expensive. Calxeda learned that the hard way. But if anyone is positioned to make it work, it's ARM. It already has a 64-bit processor and up to 8-core chips.

Next comes the software. Yes, there are Linux ports. There are Linux ports to ever CPU under the sun. Those ports are not exactly on par with the enterprise versions of Red Hat and SUSE. It could be those companies are waiting for some kind of standardized platform, since at the moment, everyone is doing their own ARM server implementation and they are not uniform the way Xeon servers are.

The saving grace will be a Windows Server port to ARM, which Microsoft apparently is doing. Bloomberg said Microsoft is unsure on releasing it, hardly a surprise after Windows 8 on ARM stiffed so badly. Then again, the tablet and server are completely different animals. If Microsoft were to come out with a full port of Server 2012 for ARM, with all the features of regular Server 2012, that might be the spark that ignites the ARM server market.

There is definitely a place for low-power servers doing simple functions like HTML serving or file and print. The big question is how much of a market there will be when the platform finally matures.

ITWorld DealPost: The best in tech deals and discounts.
Shop Tech Products at Amazon