Skip to content
Memory and storage supply stays tight in 2026 — server DRAM contract prices climbed steeply in Q1. (TrendForce, Jun 2026)
Home›Buying guides›How a refurbished server is tested

Buying guide ·

How a refurbished server is tested before it ships, and what to ask for.

Between the previous owner's rack and yours, a refurbished server passes across somebody's bench. What happens there decides much of what you are buying, and it is the part of a listing that is easiest to inflate. This is what our bench does to every unit, what it does only when a project asks for it, what it does not do at all, and what to ask any seller to put in writing.

What every unit gets

Every order goes through one bench procedure, our Standard QC, and it is the same for servers, storage and networking. It has four parts:

  • The configuration is checked against the specification you signed, and the unit is checked for completeness.
  • A visual inspection: the chassis and any damage to it, the ports and connectors, the packaging.
  • A basic functional test: the unit is powered on and passes POST, it boots, and its core components come up.
  • Photographs, a capture of the logs and a record of the serial numbers.

What a power-on test shows, and what it does not

A unit that powers on, passes POST and boots has shown that it starts and that its core components come up. That is worth knowing, and it is all the basic test claims. It is not a burn-in or a stress test, and Standard QC does not include either of them, or a firmware update.

Deeper testing exists as two add-on levels, Engineering and Expert. Both are quoted per project and billed on top of Standard, and the quality control page describes what each one adds. Nothing from those levels is done to a unit unless the quote includes it.

What the bench does not do

For networking gear, our bench does not test each port, does not verify PoE and does not compare the firmware with the vendor's current release. If a switch has to be proven on every port before it goes into service, plan that test for the day it arrives.

Testing is not the grade, and neither is the warranty

The test says the unit started, booted and brought up its core components when it left; the grade says what condition it is in. Our Grade A is written down in one place — what Grade A means — and passing Standard QC is one of its four points, next to the state of the chassis, the mounting hardware and the labels.

A test cannot catch a fault that has not happened yet, which is what the warranty is for. Ours is twelve months on refurbished and new alike. On new, it comes with the vendor’s own — on servers, usually 3 years; on refurbished, cover is extendable up to 3 years on request in your quote. The terms are on the warranty page.

Ask for the record, not the adjective

"Tested", "fully tested" and "certified" do not say what was done. Ask any seller for two things before the order rather than after a fault:

  • The procedure, as a list of steps, with what is not in it. A seller who cannot say what the test leaves out has not written the test down.
  • The record for the specific unit: its photos, its logs and its serial numbers. A record that exists for the batch but not for the unit is a description of a process, not evidence about your machine.
  • If the machine will run under sustained load from the first day — a virtualisation host, a database, anything that runs around the clock — ask for the deeper levels by name, with the number of units and the kind of equipment. The level, the quantity and the kind of equipment decide the quote.

Which logs are worth asking for

Dell, HPE and Lenovo servers keep their own history in the management controller. It is the most useful record a seller can send, because the seller did not write it:

  • Dell: the iDRAC Lifecycle log, which records system, storage and network events, firmware updates and configuration changes, or a SupportAssist collection, which packages it with the hardware inventory and the storage controller logs. Clearing the system event log leaves an entry in the Lifecycle log with the user name and the IP address it was cleared from.
  • HPE: the Integrated Management Log, which records health and status events, firmware updates and POST codes, and the Active Health System log. An IML that has been cleared is still available in the AHS log.
  • Lenovo: the XClarity Controller event log of hardware and management events. It holds up to 1024 records and then overwrites the oldest, so on a busy machine it is recent history rather than the whole of it.
  • Drives: the health data itself, not a single "health" figure. For an NVMe drive that means the percentage of rated life used, the power-on hours and the count of media and data integrity errors, which the drive keeps over its whole life.

Where to go from here

The test is one of seven questions to put to a refurbished listing; what to check before you buy refurbished covers the other six, and how to check a hardware seller covers the seller behind them.

A specific machine in mind?Send the model, the generation or a full parts list. Enterprise hardware reprices daily, so the answer is a quote for today with a lead time, rather than a number on a page.

Send the model

Every model · new & refurbished

Same 1-yr warranty, same QC