ServerWatch ran an article late last week with its look at some of the big happenings in datacenters last year, and its guess at the big attention getters for 2010. The biggest issue of 2009 in their mind was related to power, from the growth of SSD usage and sighting in areas where the TVA will let companies get cheap power, to letting things get a little hotter in the datacenter
Computers are a capital expense that depreciates over time. That’s a fact everyone accepts since there’s nothing they can do about it. The power bill is an entirely different matter. For a multitude of reasons, ranging from reducing operational costs to a desire to be more green tech-minded, bringing down the power draw was job one for everyone during 2009.
Intel and AMD did their part, introducing new server processors that stayed within the power thermals of older chips, or in some cases, drew less power. Both vendors shifted from DDR2 to DDR3 memory, which also requires less power. Additionally, with the advent of Nehalem, Intel was able to retire its power-draining FBDIMM memory.
ServerWatch’s thoughts on the year ahead? More spending to upgrade server infrastructure, (even) more focus on energy efficient computing, containers (doubt it), a slow roll for clouds, and GPUs
GPUs, represent!: nVidia has been trumpeting its graphics processors as high-performance compute engines for a while, so its executives had to have been left a little red-faced when the first supercomputer with GPUs were powered by rival ATI’s graphics processors. They won’t let that stand for very long.
In any event, the Top500 list of the fastest supercomputers will show an increasing number of GPU-powered systems, and the performance bar will go up very fast thanks to these massive math co-processors.