Table of Contents
A server power supply fails without warning, and the replacement either seats on the first attempt or costs you another week of downtime. This guide removes the guesswork. It walks you through every check that decides fit: the OEM spare part number, wattage headroom, form factor, connector keying, hot-swap support, and redundancy rules. You get the same sequence our sourcing team runs before we quote a part.
We wrote this guide for IT administrators, data center technicians, procurement teams, and business owners who buy hardware rather than specify it from a catalog. It also serves resellers and MSPs who support mixed fleets of Dell PowerEdge, HPE ProLiant, HP, Lenovo ThinkSystem, Cisco, and Supermicro equipment. If you manage servers for a government agency, hospital, university, or colocation footprint, you will find the brand-by-brand compatibility notes and the sourcing guidance for end-of-life platforms especially useful. Read the four steps in order, then use the brand sections to pick a verified module for your exact chassis.
Why Power Supply Compatibility Fails?
Most failed swaps trace back to three mistakes.
- The first is buying on wattage alone. A 750W unit from the right brand still fails if the chassis expects a different module depth or a different edge connector.
- The second is mixing generations. OEMs revise the power distribution board between platform refreshes, so a module that looks identical can carry a different pinout.
- The third is mixing wattage inside a redundant pair, which many systems reject outright or flag as a degraded configuration.
Run the four power supply checks below in order. Each one eliminates a class of error before you commit a purchase order.
Step 1: Identify the Unit You Already Have
Pull the part number from the label on the module itself. That label carries the OEM spare part number, the output rating, and the efficiency class. Record all three, because resellers and OEM catalogs index stock by spare part number rather than by server model.
- Read the service tag or serial number from the chassis, then look up the factory configuration in the OEM support portal to see which module shipped originally.
- Check the management controller. iDRAC, iLO, and XClarity all report installed PSU wattage, firmware revision, and current draw under system inventory.
- Count the bays. A single-bay chassis and a dual-bay chassis often take different module lengths even within one product family.
- Note the input type. Some platforms offer both AC and 48V DC variants, and the two never interchange.
If the original module already failed and the label is unreadable, the service tag lookup remains the fastest reliable path to the correct spare part number.
Step 2: Size the Wattage Correctly
A server PSU wattage calculator gives you a starting figure, but you get a better number by totaling the real components. Add the values below, then add 20 to 30 percent of headroom so the unit runs inside its efficiency band instead of at the top of its curve.
| Component | Typical Draw | Sizing Note |
| CPU (per socket) | 105W to 350W | Use the rated TDP of the exact SKU, not the family average. |
| DDR4 or DDR5 DIMM | 3W to 12W each | Populate count matters more than capacity. |
| SAS or SATA HDD | 7W to 12W each | Spin-up surge runs higher than steady state. |
| NVMe SSD | 8W to 25W each | U.2 and E3.S drives sit at the top of this range. |
| GPU or accelerator | 70W to 400W each | Check auxiliary connector count before sizing. |
| Fans, HBA, and board | 80W to 150W | Rises with high-airflow fan kits. |
Worked Example: A dual-socket 1U node with two 205W CPUs, sixteen DIMMs, eight NVMe drives, and a single HBA draws roughly 700W under sustained load. Add headroom and a 1100W module fits comfortably, while a 750W module leaves nothing for a later drive or GPU expansion.
Oversizing has a cost too. A unit loaded at 20 percent of capacity runs below its peak efficiency point, so you burn watts you never use. Target 40 to 70 percent sustained load.
Step 3: Match Form Factor and Connector Types
Server PSU connector types vary by platform far more than desktop hardware does. Enterprise chassis use a blind-mate edge connector on the module that docks into the power distribution board, so the pin count and keying decide fit.
- Common 1U and 2U Rack Modules: Proprietary OEM hot-plug modules with 12-pin to 24-pin blind-mate edges. These are model-locked and not cross-brand.
- CRPS (Common Redundant Power Supply): An industry standard footprint used across many white-box and OEM builds, which makes cross-vendor sourcing realistic.
- Flex ATX and 1U ATX: You find these in short-depth edge nodes and some storage heads, usually with a fixed cable harness.
- Auxiliary Output: Confirm the module supplies the 6-pin, 8-pin, or 12VHPWR outputs your GPU riser expects.
Measure module length, width, and handle orientation against the existing unit. A server power supply pinout diagram from the OEM service manual confirms the rest. If the pin map differs by even one position, the board will not negotiate power and the system stays dark.
Step 4: Confirm Hot-Swap and N+1 Redundancy Support
Hot-swap power supply compatibility depends on the cage and backplane, not on the module alone. A chassis wired for a single fixed unit will not accept a hot-plug module even when the module physically slides in.
For an N+1 redundancy power supply configuration, follow three rules:
- Match wattage across both bays. A 750W and a 1100W pair usually triggers a mismatch alert and drops the system to non-redundant mode.
- Match the efficiency class. Firmware on several platforms refuses to pair a Platinum module with a Titanium module.
- Feed each module from a separate PDU on a separate circuit, otherwise you have hardware redundancy with a single point of failure upstream.
Update PSU firmware through the management controller after installation. Mismatched firmware between paired modules causes intermittent redundancy warnings that look like hardware faults.
Brand Notes: Dell, HPE, and Lenovo Compatibility
Dell, HPE, and Lenovo server PSU compatibility follows the same pattern: modules interchange within a generation and family, and rarely across them.
Dell PowerEdge
Compatibility rule: PowerEdge 13G units such as the R630 and R730 use different module revisions than 14G units such as the R640 and R740, so plan every swap inside a single generation. Dell logs third-party modules in iDRAC as a warning rather than a block, though it complicates support calls.
Browse our full range of Dell PowerEdge power supplies.
1. Dell E750E-S0 750W Redundant Power Supply
Specifications
- Output: 750W hot-swap redundant module
- Efficiency: 80 Plus Platinum, 100V to 240V AC input
- Supported chassis: PowerEdge R520, R620, R720, R720XD, R820, T420
- Condition: Tested refurbished stock, warranty backed
- Price: $87.5.
Pros
- One part number covers six 12G PowerEdge platforms, so a single spare protects a mixed rack of R620, R720, and R820 nodes.
- Platinum efficiency holds waste heat down, which lowers the cooling load you carry across a fully populated cabinet.
Cons
- The module will not seat in 13G or 14G chassis, so R630 and R640 nodes need a different revision entirely.
- 750W leaves thin headroom once you add a GPU riser or populate every bay with high-draw NVMe drives.
Why Buy This In 2026?
Dell ended production on 12G platforms years ago, so tested refurbished modules are now the only reliable channel. Agencies and hospitals running validated 12G systems keep them in service with this exact part.
Check current stock and pricing for the Dell E750E-S0 750W power supply.
2. Dell 0XW8W 750W EPP 80 Plus Platinum Power Supply
Specifications
- Output: 750W hot-plug module for redundant bays
- Efficiency: 80 Plus Platinum, EPP rated
- Supported chassis: PowerEdge R530, R540, R630, R640, R730, R730XD, R740, R740XD, T430, T440, T630, T640
- Connector: Dell blind-mate edge, cage mounted
- Price: $50.00.
Pros
- Coverage spans both 13G and 14G rack and tower lines, which simplifies your spares matrix across two hardware generations.
- Platinum output pairs cleanly with a second identical module for N+1 redundancy on any supported chassis.
Cons
- It does not interchange with 12G platforms, so an R720 fleet gain nothing from stocking this revision.
- GPU-loaded R740 nodes need a 1100W or 2000W module instead, since 750W caps out under accelerator load.
Why Buy This In 2026?
R640 and R740 nodes still carry production virtualization and database workloads. Stocking this module keeps a failed bay to a ten-minute fix instead of a procurement cycle.
Check current stock and pricing for the Dell 0XW8W 750W power supply.
HPE ProLiant
Compatibility rule: ProLiant Gen9, Gen10, and Gen10 Plus each use a distinct Flex Slot module. The outer shells look alike, but the connectors are not equivalent, so match the spare part number instead of the wattage label.
Browse our full range of HPE ProLiant power supplies.
1. HPE 865414-B21 800W Flex Slot Platinum Hot-Plug Power Supply
Specifications
- Output: 800W hot-plug Flex Slot module, spare part 866730-001.
- Efficiency: 94 percent, 80 Plus Platinum certified, 100V to 240V AC input.
- Supported chassis: ProLiant DL360, DL380, DL560, DL580, ML110, and ML350 Gen10.
- Requirement: Installs in the Flex Slot cage with backplane, and both bays must carry the same spare part number.
- Price: $99.6.
Pros
- The Flex Slot footprint runs 25 percent smaller than the older Common Slot design, which frees rear chassis space for options.
- One part number covers the mainstream Gen10 rack and tower line, so your spares strategy stays simple.
Cons
- HPE blocks mixed part numbers in one server, so you cannot pair this module with a different 800W revision.
- Gen10 Plus and Gen11 platforms require a newer Flex Slot revision, so check the QuickSpecs before you order.
Why Buy This In 2026?
DL360 and DL380 Gen10 nodes remain the workhorse of most mid-sized data centers. This module is the standard 800W replacement and the easiest one to stock in pairs.
Check current stock and pricing for the HPE 865414-B21 800W power supply.
2. HPE 865438-B21 1600W Flex Slot Platinum Hot-Plug Power Supply
Specifications
- Output: 1600W at 200V to 240V AC, hot-plug Flex Slot module.
- Efficiency: 96 percent single-output rating, which meets the EU and UK requirement in force since January 2024.
- Supported chassis: ProLiant Gen10 servers rated for high-output Flex Slot modules, including GPU-configured DL380 and DL580 nodes.
- Best fit: Accelerator nodes, dense virtualization hosts, and fully populated NVMe chassis.
- Price: $124.8.
Pros
- Headroom at 1600W supports two or three accelerators without pushing the module past its efficiency band.
- The 96 percent rating keeps this unit compliant for buyers shipping equipment into Germany and the wider European market.
Cons
- The unit needs a 200V to 240V feed for full output, so a 120V rack circuit will derate it.
- Oversizing hurts a light node, since a 1600W module loaded at 15 percent runs below its peak efficiency point.
Why Buy This In 2026?
GPU retrofits into existing Gen10 chassis are the fastest route to added compute without buying new servers. This module supplies the power those retrofits demand.
Check current stock and pricing for the HPE 865438-B21 1600W power supply.
Lenovo ThinkSystem
Compatibility rule: ThinkSystem SR630 and SR650 share several modules within the same generation, while the older System x range does not interchange with ThinkSystem at all. XClarity reports the installed part number, which makes verification quick.
Browse our full range of Lenovo ThinkSystem power supplies.
1. Lenovo ThinkSystem 750W Platinum Hot-Swap Power Supply (7N67A00883)
Specifications
- Output: 750W hot-swap module, 115V and 230V AC input.
- Efficiency: 80 Plus Platinum certified.
- Supported chassis: ThinkSystem SR series rack nodes, including SR630 and SR650 within the same generation.
- Verification: XClarity Controller reports the installed option part number, so you can confirm the match before ordering.
- Price: $107.9.
Pros
- Dual-voltage input makes this module the practical choice for buyers deploying the same build in the US and in Germany.
- The 750W rating sits in the efficiency sweet spot for single-socket and lightly loaded dual-socket SR nodes.
Cons
- The module does not interchange with the older System x range, so mixed legacy fleets need two separate spares.
- Fully populated SR650 nodes with NVMe and a GPU will exceed this rating, so a higher-wattage ThinkSystem module suits those builds better.
Why Buy This In 2026?
SR630 and SR650 platforms sit past their initial warranty window but remain in heavy production. A tested hot-swap module restores redundancy at a fraction of a new-server cost.
Check current stock and pricing for the Lenovo ThinkSystem 750W power supply.
2. Lenovo 94Y8298 900W High-Efficiency Hot-Swap Power Supply
Specifications
- Output: 900W hot-swap module for redundant bays.
- Efficiency: 80 Plus certified, high-efficiency Platinum class.
- Supported chassis: Legacy System x rack platforms, including the x3550 M4 and x3650 M4 families.
- Condition: Tested stock for platforms that the OEM no longer supplies with new parts.
- Price: $132.6
Pros
- The 900W rating covers a fully populated System x node without the oversizing penalty of a 1400W module.
- Stocking this part keeps an accredited legacy platform in service instead of forcing an early refresh.
Cons
- The module belongs to the System x generation, so it will not seat in a ThinkSystem SR chassis.
- Supply is finite because production ended, so plan one spare per chassis type rather than ordering on failure.
Why Buy This In 2026?
Plenty of x3650 M4 nodes still run file, print, and line-of-business workloads that nobody has budget to migrate. A tested 900W module puts the second bay back in service and costs far less than replacing the node.
Check current stock and pricing for the Lenovo 94Y8298 900W power supply.
80 Plus Efficiency Ratings and Total Cost
An 80 Plus certified server PSU states its efficiency at 20, 50, and 100 percent load. The difference between tiers looks small per unit and adds up fast across a rack.
| 80 Plus Tier | Efficiency at 50% Load | Best Fit |
| Gold | 92% | General purpose and budget refresh builds |
| Platinum | 94% | Mainstream data center and virtualization nodes |
| Titanium | 96% | High-density racks and power-capped facilities |
A 1000W module running at Titanium instead of Gold saves roughly 40W of waste heat at half load. Multiply that across 40 nodes and you recover both power budget and cooling capacity.
New, Refurbished, or EOL Sourcing
The right power supply choice depends on the age of the platform and how long you plan to run it.
- Buy new power supply when the platform is current and you expect three or more years of production service. You get full manufacturer warranty coverage and firmware support.
- Buy a refurbished server power supply when the platform sits mid-life or late-life. Tested units hold the same rail tolerances as new stock and typically cost 40 to 70 percent less.
- Plan EOL server power supply sourcing before you need it. Once an OEM stops production, the secondary market becomes the only channel, and stock thins out over time. Keep one spare per chassis type on the shelf.
Healthcare and government buyers running validated or accredited systems often cannot replace a platform on the OEM schedule. A qualified compatible PSU for data center servers keeps those systems in service without a forced migration.
Replacement Walkthrough
Use this server power supply replacement guide for a hot-swap module in a redundant pair. Total time is under ten minutes per node.
- Confirm in the management controller that the surviving module carries the full load and reports no faults.
- Verify the replacement part number, wattage, and efficiency class against the installed module.
- Unplug the power cord from the failed bay, then release the handle latch and slide the module straight out.
- Seat the new module until the latch clicks. Partial seating is the most common cause of a post-swap fault.
- Reconnect the cord, confirm the status LED turns solid green, and check redundancy state in iDRAC, iLO, or XClarity.
- Update PSU firmware if the two modules report different revisions.
For a non-redundant chassis, shut the node down and disconnect both cords before you open the lid. Never work on a fixed internal unit under load.
Compatibility and Upgrade Notes
A power supply upgrade rarely stands alone. When you raise capacity, check the rest of the node:
- Memory: Adding DIMMs raises draw by 3W to 12W per module. Browse the server memory category for matched kits.
- Drives: NVMe arrays draw more than the SAS arrays they replace, so recalculate before you populate every bay. See the server drives and storage category.
- GPUs and Accelerators: Confirm auxiliary connector count and total 12V rail capacity, not wattage alone.
- Rails and PDUs: A higher wattage module can change your circuit loading and the receptacle type you need at the rack.
- Fans: High-airflow kits often become mandatory once you add GPUs or high-TDP CPUs.
Why Buy from Server Evolution?
Every power supply we ship passes a bench test under load, and we record output rails, fan function, and firmware revision before the unit leaves the facility. We cross-check each module against the OEM spare part number for your server model, so you receive a part that seats and negotiates power on the first attempt.
We stock new, refurbished, and end-of-life modules for Dell, HPE, and Lenovo platforms, back every unit with a warranty, and support buyers across the US, Canada, and Germany. If a module ever arrives faulty, our RMA process and return policy give you a documented path to a replacement. If you run a fleet, our team will build a spares matrix keyed to your chassis mix so you hold the right units instead of guessing.
Final Thoughts
Most failed replacements come from one skipped check, not from bad hardware. This guide solves that problem by giving you a repeatable sequence: confirm the spare part number, size the wattage, verify the form factor and connector, then match the pair for redundancy.
When you know the part you need, browse our server power supply category, or open the Dell, HPE, and Lenovo brand pages linked above. Every listing shows the part number and supported chassis, so you can verify fit before you order.
Still unsure which module your server takes? Send us the model, service tag, or the label from the failed unit and our team will confirm the correct part, current stock, and pricing. Contact our support team and ask about volume pricing if you are sourcing spares for a full rack. Request a bulk quote.
Frequently Asked Questions
1- Can I mix two different wattage power supplies in the same server?
No. Most enterprise platforms require matched wattage in a redundant pair and will report a configuration mismatch. The system usually keeps running, but it drops out of redundant mode, so a single failure takes the node down.
2- Will a third-party PSU work in a Dell or HPE server?
It works only if the module matches the OEM footprint, pinout, and firmware expectations. Dell and HPE both log non-OEM modules in the management controller. For production systems, source an OEM-matched module, whether new or refurbished, and keep the part number on record.
3- How do I know what wattage my server needs?
Total the rated draw of your CPUs, DIMMs, drives, and accelerators, then add 20 to 30 percent of headroom. Compare that figure against the wattage options the OEM offered for your chassis, and choose the closest option at or above your total.
4- Is a refurbished power supply reliable in production?
Yes, when the unit is load-tested and documented. A tested module holds the same rail tolerances as a new one. Refurbished stock is often the only practical option for EOL platforms, and it carries a warranty when you buy from a qualified reseller.