Brought to you by
Brought to you by
According to the certification criteria, for a special-use building that exceeds the primary energy requirements it must be demonstrated, that this exceedance occurs despite the use of energy efficient equipment. In this case the certifier is in the position to evaluate the efficiency of the main electricity consumers. Currently no fixed criteria for this are available. The following text is a guideline elaborated by PHI for the evaluation of servers and server rooms:
As for all other Passive House components, basically the most energy-efficient and cost-effective server variant should be chosen (taking into account the system requirements). Because the requirements for different server systems may vary greatly in different environments, ranging from memory-intensive but low computation database or archiving applications to computation-intensive simulations and real-time applications, it is not possible to make a general statement as to what constitutes an efficient server. A server with a high energy consumption and high processing capacity can be just as efficient as a server with a low energy consumption and less processing capacity. In order to find out whether a service is efficient or not, the power consumption can be put into relation to the computations carried out; the more computations it is possible to carry out per watt, the better.
The SPECpower_ssj2008-Benchmarks1  database provides substantiated information for this and lists the commercially available server models with their hardware; using this list, it is possible to choose extremely efficient servers (Top 30) or build one's own server using the listed hardware.
For externally communicating simple and clearly defined criteria, the largest and most easily influenced components of the server must be examined in a way that is also easy to understand for non-experts.
The biggest, and also the most essential, energy applications in a server are the processors; a faster processor will always use more energy than a slower processor in the same generation; however, from processor generation to processor generation, energy consumption declines significantly with a uniform processing capacity.
If a processor    out of the top 25% of the current hardware generation is selected, one can be sure that the processor is designed to be as efficient as possible; if the processor is also selected to match the hardware requirements (i.e. it is neither over-dimensioned, nor under-dimensioned), then it can be ensured that the biggest energy application of the server is as well adapted to the demand for energy efficiency as possible.
The second factor which is easily influenced is the power supply unit; conversion of 230V AC to various DC levels in a computer is always subject to losses; nevertheless, there are major differences in the degree of efficiency and it is possible to achieve power consumption savings in the double-digit percentage range simply by using an extremely efficient power supply unit.
The 80 Plus initiative has specified criteria for the certification of efficient computer power supply units; a power supply unit that has achieved 80 Plus Platinum certification or, if possible, the even more ambitious 80 Plus Titanium certification  should always be used. If the two criteria for hardware are kept in mind, it can be ensured that efficiency with regard to hardware will be extremely high; if the server is also accordingly dimensioned for its tasks, it can be assumed that the system will be extremely efficient.
In order to further reduce the energy consumption of the IT system, consideration must be given to the usual construction guidelines for server rooms (warm/cold aisle encasing and optimised air flow) and if possible, reliance must be placed on open air cooling with fresh air supply (instead of a fully air-conditioned system).
However, these guidelines presuppose a separate server room, which is not always given; in contrast with this, it is always possible to implement the hardware requirements when purchasing a server.
The key points are:
If possible, consumption measurements should be carried out for all servers in idle mode as well as in operation mode under realistic conditions in order to test the effectiveness of the measures and so that a small database with “empirical values” can be created for future projects.