These days, many server vendors offer you more efficient power supply options, at a slight premium compared to less efficient base versions. While everyone wants to do the right thing for the planet, your boss probably wants you to justify any increased costs with real ROI. This is fairly easy to do.
Start by visiting the 80plus.org site and familiarizing yourself with their power supply ratings:
The 80plus.org web site has a convenient search tool that lets you search and compare power supplies by vendor, efficiency, wattage, and other factors.
I’ve talked a lot recently about the HP ProLiant SL390s server. The SL390s doesn’t have its own power supply, it draws from the shared power supplies in the SL6500 chassis. By sharing the SL6500 power supplies across up to 8 SL390s servers, you gain both efficiency and redundancy vs standalone servers with their own power supplies. Most GPU equipped SL6500 configurations will use HP’s 1200 watt power supply. HP sells both a Silver-rated (91.75% efficient at 50% load) 1200 watt power supply as well as a Platinum (94.34% efficient at 50% load) 1200 watt power supply. If you configure and buy an SL6500 on HP’s web site, the configuration will default to a Platinum power supply. Switching to HP’s 1200 watt Silver power supply will save you $50 (list price) per power supply. Not the right thing to do for the planet, but lets see if it saves your boss any money.
If you work with HP’s HPC Competency Center to custom design your own HPC configuration, we will also configured your system with Platinum power supplies by default. In fact, if a sales team asks the HPC Competency Center for a configuration using less efficient power supplies, they will get a call from yours truly asking them to explain why. Let me show you the math.
The US Department of Energy tracks regional and state electric power costs. If you don’t know what your organization pays for power, you can get an idea by looking up your state average. According to the DOE, in January 2011, the average commercial power cost in New England was 14.61 cents per kilowatt hour.
From there, a simple spreadsheet calculation will let you track the cost savings of HP’s Platinum power supply. The following is a simple 1:1 comparison and doesn’t take into account savings from the SL390’s shared power supply vs standalone servers. Here is a basic formula to calculate 3 year ROI assuming full time use (common in most HPC and Web data centers) and 50% load factor on the power supply.
Savings = Cost/KW/hour * power supply wattage * load factor * 24 hours * 365 days * 3 years / 1000 watts/kilowatt * (higher efficiency – lower efficiency)
Plugging in HP’s Silver and Platinum specs you get:
Power Savings = .1461 * 1200 * .5 * 24 * 365 * 3 / 1000 * (.9434 – .9175) = $59.66
Congratulations, you just showed that even at a list price difference of $50, HP’s Platinum power supply will more than pay for itself in less than 3 years.
Of course, the cost savings can be a lot greater compared with other vendor’s power supplies. I found one large systems vendor with a 950 watt Bronze power supply rated at 87.04% efficient at load. Plugging that into my formula and normalizing for 1200 watts by multiplying by 1200/950 at the end I get:
Power Savings = .1461 * 1200 * .5 * 24 * 365 * 3 / 1000 * (.9434 – .8704) * 1200 / 950 = $185.07.
Would you pay $50 more for a power supply to save $185.07 over the next 3 years?