My colleague Matt Robben from the Windows Server performance team has produced a very comprehensive white paper and accompanying blog post which goes into great detail about how you can save energy with Windows Server 2008 power management, including new options for increasing savings even further (up to 10% more than the default configuration, based on the systems the performance team tested). Even if you aren't running Windows Server 2008, the white paper is chock full of useful advice on how to save energy through your hardware configuration and the components you use. If you operate, spec or buy servers, this is a must read.
The paper correctly points out that the cost to power and cool the server is the same if not more than the cost of buying it when amortized over three years. With this in mind, you’re probably thinking that any investment in power efficient hardware is worth the potentially extra cost. Generally it is, but there are some exceptions.
For instance, based on some analysis I did a few months back of SPEC power benchmarks, it turns out that buying low power processors might not make much financial sense - depending on the price you pay. I calculated that with a $200 spread between two equally performant processors with different TDPs (Thermal Design Power), the ROI is about 7 years – obviously way beyond the life of the machine. However, the $200 difference in question was based on 'list' prices. On Amazon.com (they really do sell everything!), the difference was only $30 – which had a payback of about 7 months for a single processor.
This obviously applies to other components as well (e.g. price per GB for 2.5” vs 3.5” disk drives). If more efficient components are more expensive, then it might be worth doing some rudimentary math to see if the power savings will ever exceed the increased cost of the components before the machine is retired.