The fact is, computing hardware is so close to 100% inefficient that converting actual power in Watts directly to heat (at 3.4 SBTU per Watt) is extremely close to reality. And since the actual theoretical conversion is 3.412 SBTU per Watt, a small amount of "unconverted power" might be considered to have already been taken into account by using the more commonly used 3.4 conversion factor.
The first law of thermodynamics, known as conservation of energy, states that energy can neither be created nor destroyed – it can only change from one state to another. Incoming electric power is electromagnetic energy. The Ethernet or optical fiber output from a server or data switch is also of electromagnetic form, but the amount of actual energy in those signals is so miniscule compared with the electrical input to the compute device as to be negligible. If the device contains disk drives, there will also be some electrical power converted to mechanical kinetic energy in the spinning discs. But, again, the amount of mechanical energy is so small compared with the heat produced in the process of overcoming the inertia of the discs and maintaining their velocities at the very high rotational speeds at which they operate, that it is simply not worth taking into account in making practical computations of the kind we are discussing. Further, outside of a sophisticated laboratory, we have no realistic way of measuring the amount of power that is converted to something other than heat (or, conversely, to measure both the incoming power and the heat release with such accuracy as to be able to calculate the conversion differential).
If a manufacturer follows ASHRAE's recommended guidelines for specifying power consumption and heat release of its devices (as set forth in ASHRAE Publication Thermal Guidelines for Data Processing Environments, Chapter 5, published in 2004), then these small energy conversion differences may have been accounted for, and the heat release numbers given for the configuration employed should certainly be used. However, even with this more accurate information, the total heat loads you compute must still be considered "estimates," and should be qualified as such. There are two reasons: First, your actual configuration is probably slightly different than the closest thing the manufacturer provides in the power and heat data; and second, these figures are all taken at specific line voltages, which are probably slightly different in your facility. Since actual current varies inversely with voltage, power efficiencies and, therefore, heat release characteristics, will also change slightly. None of these variations, however, are likely to be large enough to make any difference in room designs or total load requirements. The entire process of air conditioning and creating air movement is so much less exact than the differentials we are discussing as to make them meaningless outside of the laboratory.
In short, unless you have accurate manufacturer's data on heat release, you are best using total conversion of actual power consumption to heat. Obtaining reasonably accurate numbers for actual power consumption is difficult enough without becoming further concerned about minute differences in energy conversion.
Regarding the 1.08 factor in the denominator of the equation:
CFM = BTU / 1.08 x TD
The value 1.08 is a weight of air correction factor. It comes from the specific heat of air (0.24), and the volume of one pound of air under standard conditions (13.34 cu ft). Air conditioning effectiveness actually depends on the weight of air, but since calculations are done in cubic feet per minute (CFM), a conversion is necessary:
60 Minutes / 13.34 cu ft per Pound = 4.5 Pounds of Air Per Hour.
4.5 Pounds of Air Per Hour x 0.24 (Specific Heat) = 1.08
Obviously, at higher altitudes, where the air is less dense, and "standard conditions" no longer prevail, this correction factor must be further corrected for accurate calculations on large systems. For most situations in data centers, however, the 1.08 factor is just fine and, in many cases, can even be ignored without real consequence.
This was first published in September 2005