A server power supply label is a compact record of what the module is and the electrical conditions under which it is rated. Read it in three passes: capture the complete identity, separate AC input from DC output, then interpret any efficiency or safety marks. That reading can reject an obviously unsuitable spare. It does not, by itself, approve a replacement for a particular server.
Start with the identifiers, including every suffix
Photograph the entire nameplate before removing a failed module, and make a second image of any separate service sticker. A rear bay can hide part of a label; a hurried transcription can turn similar characters into different part numbers. Record the manufacturer, complete model designation, output rating, OEM or service part number, serial number, and any revision exactly as printed. Keep punctuation and suffixes. A base model may identify a product family while a suffix identifies a particular build or electrical variant.
These fields answer different questions. A manufacturer’s model identifies its design in that manufacturer’s documentation. An OEM part number can identify the version a server vendor supplies or supports. A serial number identifies an individual unit, generally for traceability, rather than a design that can be ordered interchangeably. A hardware revision may matter when a platform vendor documents a change. Do not assume that two labels with the same wattage and similar housings represent the same service part.
Where a server management interface reports a PSU part number, compare that record with the physical label and the service documentation. The displayed string may include elements that are presented differently on the sticker. For example, Dell explains one PowerEdge iDRAC record, 0FR0KXA02, as the label part number 0FR0KX followed by hardware revision A02; FR0KX is a separate internal tracking form. That is Dell’s explanation of this example, not a universal code for reading other brands. If a seller quotes only a shortened number, ask for a clear label photograph and the complete OEM identifier.
Read the input line as a condition of operation
Find the line marked Input, AC input, or an equivalent designation. It may specify AC or DC, a voltage range, frequency, and input current. These values describe the power arriving at the module. They are not the DC rail values delivered to the server board. Check the voltage available at the rack or facility and the server’s specified power cord and inlet arrangement against the exact PSU and platform documentation.
A voltage range is more informative than a wattage badge. Dell documents a particular 700 W PowerEdge supply, DPN 3D09N, with a 200–240 V AC input range and states that it does not support 110 V AC. The official Dell high line PSU bulletin makes the practical consequence explicit: at a low line such as 110 V AC, that unit’s LED is off and the server does not power on. This is a specific part, not a rule about every 700 W supply. A buyer who records only “700 W” would miss the decisive input condition.
Some labels present more than one operating condition. The module may have different output ratings at different input ranges, or the manufacturer may publish those distinctions in a datasheet rather than on the sticker. Read each condition with its paired rating. Do not carry a high line power figure over to a low line supply assumption without an explicit manufacturer statement. Similarly, an input current figure is a supply-side rating useful for matching the specified infrastructure; it is not the expected continuous draw of the server. Actual current changes with load, conversion efficiency, input voltage, and operating state.
The frequency line should be read literally. A value printed beside AC input belongs to the incoming source, not to an output rail. A DC-input model calls for a different source and connection scheme even if its housing resembles an AC model. Labels may use compact symbols, so confirm an ambiguous AC/DC mark against the exact datasheet rather than inferring it from a connector’s appearance. The nameplate also does not describe the complete rack power path: cord rating, inlet rating, upstream protection, and site wiring are separate checks.
Separate each DC rail from the total output rating
The output area may list a main rail, one or more auxiliary rails, and a standby output. Read each voltage with its associated current and polarity where shown. An output such as 12 V at 90 A represents a nominal rail power of 1,080 W by multiplication. It does not establish that the entire module can simultaneously deliver that amount plus every other printed rail maximum. The manufacturer may impose a combined output limit, input-dependent rating, or operating conditions that govern simultaneous loading.

Standby power is easy to overlook because its current is often much lower than the main output. It can support management and wake functions when the main output is not active, depending on the platform design. A prospective spare with the right main voltage and headline wattage may still differ in standby capability or interface. The label tells you what rail is rated; it usually does not reveal when the platform enables it, how pins are assigned, or whether the server accepts that particular module.
Watch the units. Volts describe electrical potential, amperes describe current, and watts describe power. The familiar voltage-times-current calculation applies to a specified DC rail, under its stated rating. The printed wattage is a rated output capability under defined conditions, not a measurement of what the server is consuming now. If a redundant chassis contains two supplies, adding their nameplate wattages does not automatically give the usable load under a redundancy requirement. The server’s power architecture and its supported operating mode determine the available capacity after one module fails.
A useful way to transcribe a dense label is to preserve the pairing: “Input: [type, range, frequency, current]” on one line and “Output: [each rail and current]; combined limit: [if stated]” on another. Do not rewrite an unreadable number from memory. A partial photograph or illegible nameplate warrants checking the exact model documentation. The task is to retain the conditions attached to a number, not merely to collect the largest number visible.
Where a datasheet gives several power figures, check the headings and footnotes before reconciling them with the label. Peak, continuous, standby, and condition-specific figures are not interchangeable descriptions. If the label and current official documentation appear to disagree, check the full model and document revision and seek manufacturer clarification for a critical replacement requirement. A guessed interpretation can produce a spare that appears acceptable on paper but fails under the actual source or load.
Interpret marks within their stated scope
An efficiency tier or compliance logo adds information about a particular kind of evaluation. It is not another output rating. An 80 PLUS mark, for example, relates to an efficiency certification under program test conditions. The tier does not say how many watts the server currently draws, how much heat the installed unit releases at every operating point, or whether that unit fits a given chassis. For an exact efficiency claim, use the listed model and its applicable test record rather than extrapolating from the mark alone.
Safety marks likewise need to be read in context. A symbol, geographic code, file number, or recognized-component mark may have a specific scope. Its presence does not certify an entire assembled server or turn the module into a supported field replacement. When a mark affects procurement or regulatory review, verify the exact product and designation in the issuing organization’s database or the manufacturer’s documentation. A cropped photograph that shows a logo but omits the model and rating lines is poor evidence for a purchase decision.
Other label text can include manufacturer location, manufacturing date, environmental symbols, warning text, or handling instructions. Treat warnings as instructions for that product, not as decorative markings. Do not open or probe a supply to resolve a nameplate question. If the concern is electrical safety or physical damage, follow the server vendor’s service procedure and use qualified personnel. Nameplate interpretation is a document-reading task, not a substitute for live electrical testing.
What the nameplate cannot settle for a replacement
The label is strong evidence of identity and declared electrical ratings. It is weak evidence of mechanical and system compatibility. A complete replacement check needs the server model’s service manual or approved-parts information. The module must fit the correct bay and latch, mate with the intended connector and pin assignment, receive appropriate airflow, and operate with the platform’s control and redundancy design. Those properties generally cannot be reconstructed from a wattage line or a photograph of the outer shell.
Two supplies with apparently equal electrical ratings may differ in connector geometry or signaling. Server platforms can also have vendor-specific acceptance rules, supported alternate part numbers, and limits on mixing modules. For one concrete example, Lenovo’s ThinkSystem SR850 V4 user guide requires supplies installed in that chassis to have the same wattage, vendor, and part number or alternate part number. That is a rule for that system; it should not be projected onto every server. A technician should apply the rule in the manual for the actual chassis.
Firmware reporting is another distinct layer. A PSU can have a physically plausible label and still be unrecognized by management software, or be recognized without meeting the intended operating configuration. Management inventory is useful for comparing installed and proposed identifiers, but it does not replace the approved-parts list. Likewise, a matching outer length is no proof that the connector, handle, insertion depth, or cooling path is correct. Avoid test-fitting a questionable spare in production merely to answer a catalog question.
Label photographs are particularly valuable when procurement crosses several naming systems. Give the supplier the exact server model, original OEM spare number, complete PSU model including suffix, input and output ratings, and relevant revision. Request the proposed unit’s own label and OEM compatibility reference. If the platform permits alternate parts, identify the specific approved alternate rather than accepting “same wattage” as an equivalence claim. This preserves a traceable path from the failed module to the ordered spare.
From label reading to a defensible spare order
Imagine a failed module pulled from a redundant server. Its face says “700 W,” while the side sticker carries a longer OEM number, a manufacturer model suffix, and a narrow AC input range. A candidate replacement in a catalog also says “700 W.” The useful comparison starts with the full identifiers and input condition, then moves to the rails and the server’s approved alternate list. It does not start or end with the shared wattage. If the site feed falls outside either module’s stated range, the candidate is excluded before any question of mechanical fit.

Once electrical conditions align, consult the platform manual for supported PSU types and mixing rules. Match the proposed part number against the vendor’s approved replacement or documented alternate, including any relevant revision condition. If support documentation is incomplete, ask the OEM or module manufacturer for the missing compatibility statement. The label remains a source record: it tells you exactly which unit and ratings you are asking about. It cannot supply the platform approval that the manual or vendor must provide.
This division of evidence is the central skill in reading a server PSU label. Capture the complete nameplate, keep input and output values with their conditions, and interpret marks according to their limited scope. The resulting record can eliminate unsuitable candidates and make a replacement inquiry precise. A safe, supported spare order still rests on the exact server’s service information and the specific replacement part it names.