Division 3 The data to calculate the active state efficiency (Effserver) and the idle power (Pidle ) shall be measured during the same test according to the relevant standard, where the idle power can be measured either before or after running part of the test for the active state efficiency.

The active state efficiency (Effserver) of servers shall be calculated as:

Eff_server = exp [W_cpu × ln (Eff_cpu ) + W_Memory × ln (Eff_Memory ) + W_Storage × ln (Eff_Storage )]

where: W_CPU , W_Memory and W_Storage are the weightings applied to the CPU, Memory and Storage worklets respectively, as follows:

• W_CPU is the weighting assigned to the CPU worklets = 0,65

• W_Memory is the weighting assigned to the Memory worklets = 0,30

• W_Storage is the weighting assigned to the Storage worklets = 0,05

and

$${\mathrm{Eff}}_{\mathrm{CPU}}={\left(7i=1{\mathrm{Eff}}_i\right)}^{1∕7}$$

where:

• i = 1 for workletCompress;

• i = 2 for workletLU;

• i = 3 for workletSOR;

• i = 4 for workletCrypto;

• i = 5 for workletSort;

• i = 6 for workletSHA256;

• i = 7 for workletHybrid SSJ;

$${\mathrm{Eff}}_{\mathrm{Memory}}={\left(2i=1{\mathrm{Eff}}_i\right)}^{1∕2}$$

where:

• i = 1 for workletFlood3;

• i = 2 for workletCapacity3;

$${\mathrm{Eff}}_{\mathrm{Storage}}={\left(2i=1{\mathrm{Eff}}_i\right)}^{1∕2}$$

where:

• i = 1 for workletSequential;

• i = 2 for workletRandom;

and

$${\mathrm{Eff}}_i=1000\frac{{\mathrm{Perf}}_i}{{\mathrm{Pwr}}_i}$$

where

— Perf_i

Geometric mean of the normalized interval performance measurements;

— Pwr_i

Geometric mean of the measured interval power values;

In order to create a single energy efficiency metric for a server the interval efficiency values for all the different worklets shall be combined using the following procedure:

1. (a)

   combining the interval efficiency values for the individual worklets using the geometric mean to obtain individual worklet Efficiency values for the worklet;

2. (b)

   combining worklet efficiency scores using the geometric mean function by workload type (CPU, Memory, Storage) to obtain a workload type value;

3. (c)

   combining the three workload types using a weighted geometric mean function to obtain a single, total server efficiency value.