At its 2012 Financial Analyst Day, AMD mentioned a conscious shift towards being a more agile company. Wanting a bigger piece of the high margin server market, AMD is willing to consider alternate architectures (largely believed to be ARM based) to go after disruptive holes in Intel's strategy. AMD also wants to leverage its GPU advantage in servers, although outside of the HPC market that's still a while away. Two days ago AMD gave the first example of its more aggressive, agile pursuit of server market share with an announcement of the intent to acquire SeaMicro for $334M.

We first met SeaMicro two years ago when it introduced the SM10000. Built out of a massive array of 512 Intel Atom CPUs, the original SM10000 was designed to deliver great performance at much lower power than traditional servers. Similar to a trend we've seen in client computing, not all server workloads demand the absolute latest innovations in microprocessors. There's a category of server applications that can be better served by a lower class of good enough computing, delivering much better power efficiency. Content web servers, similar to what we use at AnandTech, don't present a hugely complex workload but they do see lots of threads and have largely variable performance requirements. SeaMicro's technology reduces power consumption by using lower power CPUs and highly power optimized motherboards. Both SATA and Ethernet are virtualized and handled by SM's custom ASIC, as is connection between all of the servers in the chassis (we discussed the architecture in greater detail here). You can think of the SM10000 as a single box cloud server for content sites. As power consumption in the datacenter is one of the greatest expenses, matching hardware to software workloads is very important.

Intel already announced intentions to pursue the micro-server market in 2012 with a special sub-10W Xeon SKU, likely based on an Atom derived architecture. AMD already demonstrated its Bobcat core could be easily competitive with anything Atom based, making micro-servers an easy target for AMD in the near term. Intel expects the micro-server market to grow to be ~10% of the total server market by 2015. With the right product play, AMD could stand to achieve a bigger share of that market than it has in the overall server space.

The acquisition of SeaMicro makes sense but it does mark a significant shift in the way AMD has behaved as a player in the microprocessor space. By acquiring a server company AMD is taking a step towards  vertical integration, something which has definitely become more popular in recent years. Obviously AMD isn't going to turn into a server manufacturer, but a targeted acquisition like this is indicative of the strategy AMD laid out last month.

SeaMicro currently ships three server designs, all of which use Intel CPUs (two use Atoms, one uses low-voltage SNB Xeons). You can expect AMD to eventually transition these systems to Opterons and perhaps introduce a Bobcat based version at some point in the future.

Source: Marketwatch

Comments Locked

5 Comments

View All Comments

  • Beenthere - Friday, March 2, 2012 - link

    Looks like a good acquisition to me. Opteron and Trinity APU sales should jump nicely.
  • IlllI - Friday, March 2, 2012 - link

    welcome to 3 days ago.
  • bigboxes - Friday, March 2, 2012 - link

    Wow. A whole three days. Did you get the scoop?
  • silverblue - Saturday, March 3, 2012 - link

    You mean, one of the rare occasions when Toms have beaten AT to press on a news article? :)
  • dealcorn - Saturday, March 3, 2012 - link

    Clearly a bold move ripe with opportunity that cuts many ways. That SeaMicro is promiscuous, serving many masters, has strong appeal to AMD as mating SeaMicro ip with ARM cores just sounds right. From a design perspective, SeaMicro is already working with AMD cores and a saleable product(s) is(are) close. That is why this may be a dumb idea.

    Today the playing field is level regarding SeaMicro ip and Intel does not want them with their current, ahem, business model. If Intel bought them and then shut off ip access to competitors those wonderful guys and gals in the the Antitrust Division of the Justice Department would have a party day. Intel's sole competitive option is to build their own but why bother as long as the playing field is level. Rory just spent $300 odd million to tell Intel it is important that they obsolete the SeaMicro product rapidly. Rory has the reputation he executes well. I wonder how this will work out for him once he gets past the immediate short term benefits.

Log in

Don't have an account? Sign up now