More Server Power For Lower Prices
Server prices will be flat or down over the next year, but performance is expected to go up with increased power efficiency to boot. From dual cores to quad cores and beyond, here's what to look for on the server horizon.
As with weather forecasts, you need to start with the big picture when forecasting technology. For small and midsize businesses, the server market big picture includes 30% more power for about the same, or even lower, prices. But like the weather, there are microclimates; when you zoom in, things unfold at different rates and in different directions, but that doesn't change the overall forecast for more power.
For pricing, the news is especially good. Server prices are, on average, trending downward rather than upward, even when accounting for the newest technology. In many, but not all, cases, you can actually expect to pay less for more.
According to Dataquest predictions released in September, the average selling price for servers, worldwide, will fall about 5% during the coming year. In 2008, the average price of a server was $5,938.78; in 2009, that's expected to fall to $5,647.14. In the U.S. market, server prices are projected to decline 4% from $5,249.80 this year to $5,041.45 in 2009.
But those are averages. The pricing is less enticing when you isolate prices for x86 servers -- the lower-end models core to the small and midsize business markets. Dataquest projects prices for x86 servers to rise, albeit slightly. In 2008, the average price was $3,383.51, and you can expect to pay an average of $12, or one-third of 1%, more next year.
There's downward price pressure at the high end of the small and midsize server market, dominated by the Intel Itanium processor chip. In 2008, Itanium pricing averaged $44,039.72; next year that price is expected to fall by a little more than 2%, to $42,973.80. Pricing for enterprise-class servers is plummeting even faster and is the primary driver for the forecasted 5% overall server pricing decline.
Processing The Cores
For the foreseeable future, the computer industry will abide by Moore's Law, which holds that the power of computer chips doubles every other year. But it's not quite that simple. Until a few years ago, vendors kept pace with Moore's Law by producing circuitry with faster and faster clock speeds. But after hitting the 3-gigahertz level, they found they could not go much faster given that at higher speeds, chips tended to overheat.
To solve the heat issue, vendors began using multiple cores on each CPU chip. A core is, basically, the circuitry that would form a discrete processor. By putting multiple processors on a chip, the vendors can offer more power at that same clock speed.
Multiple cores increase a computer's power by dividing tasks among the cores and running them simultaneously. Multitasking and multithreading operations can immediately benefit. Virtualization can benefit because virtual machines can be divided among the cores. And though applications relying on one core will not automatically show improvement, vendors are adding other chip improvements to improve the performance of the cores themselves.
Dual-core CPUs are now the norm, and the industry is moving to four and six cores, with talk of eventually offering scores of cores.