(1)

‘server with one or two processor sockets’ means a server containing one or two interfaces designed for the installation of a processor. For multi-node servers, this term refers to a server having one or two processor sockets in each server node;

(2)

‘Input/Output (I/O) device’ means a device, which provides data input and output capability between a server or a data storage product and other devices. An I/O device may be integral to the server motherboard or may be connected to the motherboard via expansion slots (such as Peripheral Component Interconnect, or Peripheral Component Interconnect Express);

(3)

‘motherboard’ means the main circuit board of the server. For purposes of this regulation, the motherboard includes connectors for attaching additional boards and typically includes the following components: processor, memory, BIOS, and expansion slots;

(4)

‘processor’ means the logic circuitry that responds to and processes the basic instructions that drive a server. For purposes of this regulation, the processor is the CPU of the server. A typical CPU is a physical package to be installed on the server motherboard via a socket or direct solder attachment. The CPU package may include one or more processor cores;

(5)

‘memory’ means a part of a server external to the processor in which information is stored for immediate use by the processor, expressed in gigabyte (GB);

(6)

‘expansion card’ means an internal component connected by an edge connection over a common/standard interface such as Peripheral Component Interconnect Express providing additional functionality;

(7)

‘graphics card’ means an expansion card containing one or more graphics processing units with a local memory controller interface and local graphics-specific memory;

(8)

‘buffered double data rate (DDR) channel’ means a channel or memory port connecting a memory controller to a defined number of memory devices in a server. A typical server may contain multiple memory controllers, which may in turn support one or more buffered DDR channels. As such, each buffered DDR channel serves only a fraction of the total addressable memory space in a server;

(9)

‘blade server’ means a server that is designed for use in a blade chassis. A blade server is a high-density device that functions as an independent server and includes at least one processor and system memory, but is dependent upon shared blade chassis resources (e.g., power supply units, cooling) for operation. A processor or memory module will not be considered a blade server when the technical documentation for the product does not indicate that it scales up a standalone server;

(10)

‘blade chassis’ means an enclosure that contains shared resources for the operation of blade servers, blade storage, and other blade form-factor devices. Shared resources provided by a blade chassis may include power supply units, data storage, and hardware for direct current power distribution, thermal management, system management, and network services;

(11)

‘High Performance Computing (HPC) server’ means a server which is designed and optimized to execute highly parallel applications, for higher performance computing or deep learning artificial intelligence applications. HPC servers must meet all the following criteria:

(12)

‘server product family’ means a high-level description referring to a group of servers sharing one chassis and motherboard combination that may contain more hardware and software configurations. All configurations within a server product family must share the following common attributes:

(13)

‘power supply unit’ (PSU) means a device that converts alternate current (AC) or direct current (DC) input power to one or more DC power outputs for the purpose of powering a server or a data storage product. A server or data storage product PSU must be self-contained and physically separable from the motherboard and must connect to the system via a removable or hard-wired electrical connection;

(14)

‘power factor’ means the ratio of the real power consumed in Watts to the apparent power drawn in Volt Amperes;

(15)

‘single output PSU’ means a PSU designed to deliver the majority of its rated output power to one primary DC output for the purpose of powering a server or a data storage product. Single output PSUs may offer one or more standby outputs that remain active whenever connected to an input power source. The total rated power output from any additional PSU outputs that are not primary and standby outputs shall be no greater than 20 Watts. PSUs that offer multiple outputs at the same voltage as the primary output are considered single-output PSUs unless those outputs:

(16)

‘multi output PSU’ means a PSU designed to deliver the majority of its rated output power to more than one primary DC output for the purpose of powering a server or a data storage product. Multi output PSUs may offer one or more standby outputs that remain active whenever connected to an input power source. The total rated power output from any additional PSU outputs that are not primary and standby outputs shall be no greater than or equal to 20 Watts;

(17)

‘direct current server’ means a server that is designed solely to operate on a DC power source;

(18)

‘direct current data storage product’ means a data storage product that is designed solely to operate on a DC power source;

(19)

‘idle state’ means the operational state in which the OS and other software have completed loading, the server is capable of completing workload transactions, but no active workload transactions are requested or pending by the system (i.e., the server is operational, but not performing any useful work). For servers where Advanced Configuration and Power Interface standards are applicable, idle state corresponds only to System Level S0;

(20)

‘idle state power’ (Pidle) is the power demand, in Watts, in idle state;

(21)

‘low-end performance configuration’ of a server product family means the combination of two data storage devices, processor with the lowest product of core count and frequency (in GHz) and memory capacity (in GB) that is at least equal to the product of the number of memory channels and the lowest capacity dual in-line memory module (DIMM) (in GB) offered on the server that represents the lowest performance product model within the server product family. All memory channels shall be populated with the same DIMM raw card design and capacity;

(22)

‘high-end performance configuration’ of a server product family means the combination of two data storage devices, processor with the highest product of core count and frequency and memory capacity (in GB) equal to or greater than 3 times the product of the number of CPUs, cores and hardware threads that represents the highest performance product model within the product family. All memory channels shall be populated with the same DIMM raw card design and capacity;

(23)

‘hardware thread’: means the hardware resources in a CPU core to execute a stream of software instructions. A CPU core may have the resources to execute more than one thread simultaneously;

(24)

‘active state efficiency’ (Effserver) means the numerical value for server efficiency as measured and calculated according to Annex III point 3;

(25)

‘active state’ means the operational state in which the server is carrying out work in response to prior or concurrent external requests (e.g., instruction over the network). Active state includes both active processing and data seeking/retrieval from memory, cache, or internal/external storage while awaiting further input over the network;

(26)

‘server performance’ means the number of transactions per unit of time performed by the server under standardised testing of discrete system components (e.g. processors, memory and storage) and subsystems (e.g. RAM and CPU);

(27)

‘maximum power’ (Pmax) means the highest power, in Watts, recorded on the eleven worklet scores according to the standard;

(28)

‘CPU performance (PerfCPU)’ means the number of transactions per unit of time performed by the server under standardised testing of the CPU subsystem;

(29)

‘Auxiliary Processing Accelerator’ (APA) means a specialized processor and associated subsystem that provide an increase in computing capacity such as graphical processing units or field programmable gate arrays. An APA cannot operate in a server without a CPU. APAs can be installed in a server either on Graphics or Extension add-in cards installed in general-purpose add-in expansion slots or integrated into a server component such as the motherboard;

(30)

‘Expansion APA’ means an APA that is on an add-in card installed in an add-in expansion slot. An expansion APA add-in card may include one or more APAs and/or separate, dedicated removable switches;

(31)

‘Integrated APA’ means an APA that is integrated into the motherboard or CPU package;

(32)

‘product type’ means the design of the server or of the data storage product including the chassis (rack, tower or blade), the number of sockets and, for servers, whether it is a resilient server, a blade server, a multi node server, a HPC server, a server with integrated APA, a direct current server or none of the previous categories;

(33)

‘disassembly’ means a process whereby an item is taken apart in such a way that it can subsequently be reassembled and made operational;

(34)

‘firmware’ means system, hardware, component, or peripheral programming provided with the product to provide basic instructions for hardware to function inclusive of all applicable programming and hardware updates;

(35)

‘secure data deletion’ means the effective erasure of all traces of existing data from a data storage device, overwriting the data completely in such a way that access to the original data, or parts of them, becomes infeasible for a given level of effort.

(a)

data storage devices;

(b)

memory;

(c)

processor (CPU);

(d)

motherboard;

(e)

expansion card/graphic card;

(f)

PSU;

(g)

chassis;

(h)

batteries.

(a)

product type;

(b)

manufacturer's name, registered trade name and registered trade address at which they can be contacted;

(c)

product model number, and if applicable the low-end performance configuration and the high-end performance configuration model numbers;

(d)

year of manufacture;

(e)

PSU efficiency at 10 % (if applicable), 20 %, 50 % and 100 % of rated output power, with the exception of direct current servers, expressed in % and rounded to the first decimal place;

(f)

power factor at 50 % of the rated load level, with the exception of direct current servers, rounded to three decimal places;

(g)

PSU rated power output (Watts), rounded to the nearest integer. If a product model is part of a server product family, all PSUs offered in a server product family shall be reported with the information specified in (e) and (f);

(h)

idle state power, expressed in Watts and rounded to the first decimal place;

(i)

list of all components for additional idle power allowances, if any (additional PSU, HDDs or SSDs, additional memory, additional buffered DDR channels, additional I/O devices);

(j)

maximum power, expressed in Watts and rounded to the first decimal place;

(k)

declared operating condition class, as detailed in Table 6;

(l)

idle state power (Watts) at the higher boundary temperature of the declared operating condition class;

(m)

the active state efficiency and the performance in active state of the server;

(n)

information on the secure data deletion functionality referred to in point 1.2.2 of this Annex, including instructions on how to use the functionality, the techniques used and the supported secure data deletion standard(s), if any;

(o)

for blade servers, a list of recommended combinations with compatible chassis;

(p)

if a product model is part of a server product family, a list of all model configurations that are represented by the model shall be supplied.

(a)

product type;

(b)

manufacturer's name, registered trade name and registered trade address at which they can be contacted;

(c)

product model number;

(d)

year of manufacture;

(e)

PSU efficiency at 10 % (if applicable), 20 %, 50 % and 100 % of rated output power, with the exception of direct current online data storage products, expressed in % and rounded to the first decimal place;

(f)

power factor at 50 % of the rated load level, with the exception of direct current online data storage products, rounded to three decimal places;

(g)

declared operating condition class, as detailed in Table 6; it shall also be indicated that ‘This product has been tested in order to verify that it will function within the boundaries (such as temperature and humidity) of the declared operating condition class’;

(h)

information on the data deletions tool(s) referred to in point 1.2.2 of this Annex, including instructions on how to use the functionality, the techniques used and the supported secure data deletion standard(s), if any.

(a)

indicative weight range (less than 5 g, between 5 g and 25 g, above 25 g) at component level, of the following critical raw materials:

(b)

instructions on the disassembly operations referred to in point 1.2.1 of this Annex, including, for each necessary operation and component:

(a)

Information listed in points 3.1 and 3.3, in the case of servers

(b)

Information listed in points 3.2 and 3.3, in the case of data storage products

1.

2.

All configurations offered within a server product family shall contain the same number of populated processor sockets used during testing. A server product family can be defined for a server with only partially populated sockets (e.g. one processor populated in a two socket server) as long as the configuration(s) are tested as a separate server product family, as required, and meet the same requirements for the number of populated sockets within that separate server product family.

For servers with expansion APA, the unit under test shall be tested with the expansion APA removed, when measuring the idle state power, the active state efficiency) and the server performance in active state. Where an expansion APA relies on a separate Peripheral Component Interconnect Express switch for communication between the APA and CPU, the separate Peripheral Component Interconnect Express card(s) or riser(s) shall be removed for Active State and Idle State testing of all configurations.

For multi-node servers, the unit under test shall be tested for per node power consumption in the fully-populated chassis configuration. All multi-node servers in the multi-node chassis shall share the same configuration (homogeneous).

For blade servers, the unit under test shall be tested for blade server power consumption in the half-populated chassis configuration, and the chassis shall be populated as follows:

3.

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

Effserver = exp [Wcpu × ln (Effcpu ) + WMemory × ln (EffMemory ) + WStorage × ln (EffStorage )]

where: WCPU , WMemory and WStorage are the weightings applied to the CPU, Memory and Storage worklets respectively, as follows:

and

where:

where:

where:

and

where

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.

The Member State authorities shall verify one single unit of the model or, in case the manufacturer reports on a server product family, of the model configuration. If the verification is done on the low-end performance configuration or the high-end performance configuration, the declared values shall be the values for the respective configuration. If the verification is performed on a randomly selected or ordered model configuration, the declared values shall be the values for the high-end performance configuration.

2.

The model or model configuration shall be considered to comply with the applicable requirements if:

3.

If the results referred to in points 2(a) or 2(b) are not achieved, the model and all model configurations that are covered by the same product information (according to Annex II point 3.1(p) shall be considered not to comply with this Regulation.

4.

If the result referred to in point 2(c) is not achieved:

5.

The model or model configuration shall be considered to comply with the applicable requirements if, for these three units, the arithmetical mean of the determined values complies with the respective verification tolerances given in Table 7.

6.

If the result referred to in point 4(b) is not achieved, the model and all model configurations that are covered by the same product information (according to Annex II point 3.1(p) shall be considered not to comply with this Regulation.

7.

The Member State authorities shall provide all relevant information to the authorities of the other Member States and to the Commission without delay after a decision being taken on the non-compliance of the model according to points 3 and 6.