Skip to content

Dear customers! Please send orders and inquiries to [email protected] or to Telegram @servix_uz instead of calling by phone: due to the high workload, the operator cannot answer all incoming calls.

Servix

Refurbished or new server: support period, hidden costs, an honest calculation

Servix technical team · Updated

In brief. A refurbished server is justified for expanding an existing fleet, test environments, cold standby and branch offices, but not for production where downtime is costly, or for projects of five years or more. Check the end-of-support date: for HPE ProLiant Gen10, sales ended at the end of 2024, and support remains in effect until 2028–⁠2029. VMware charges for a minimum of 16 cores per processor, so an old server with 8-core CPUs costs the same in licenses as a new one with 32-core CPUs, while doing many times less useful work.

A server with the same specifications can cost half as much if it is not new. The temptation is understandable, and in many scenarios it is a sensible purchase. The problem is that the decision is usually made on one figure — the price — without looking at three others: until what year the vendor will support this model, what the licenses for its cores will cost, and how much electricity it will consume over five years. Let’s look at when the savings are real and when they are illusory.

Three different things under one word

Factory refurbishment (refurbished). The equipment has been inspected and restored by the vendor or an authorized partner and has received a new warranty, usually for a year or longer. This is the most predictable option: there is someone to file a claim with, and the history of the hardware is known.

Used, from a system integrator. The equipment has been decommissioned at a known customer and inspected, often with consumables replaced — fans, RAID controller batteries, sometimes drives. The warranty is provided by the seller, not the vendor, and its terms are worth reading carefully: the period, what exactly is covered, and whether there is a replacement pool.

Used, from a private seller. A listing on a marketplace: “everything works, tested.” Here you are buying not a server but a lottery ticket: neither the operating conditions, nor the remaining life of the drives, nor the firmware history is known. For a test bench it is acceptable; for anything the company’s operations depend on, it is not.

The main question: until what year will it be supported

Every server platform has two dates. The first is the end of sales (End of Life), the second is the end of service support (End of Service Life), after which the vendor stops releasing microcode updates and security patches, and service contracts are no longer issued. Usually the second date comes about five years after the first.

A practical guideline: for HPE ProLiant Gen10, the end of sales came at the end of 2024, and support continues until 2028–⁠2029, depending on the model. That is, a refurbished Gen10 bought today means several more years of normal life with updates. For platforms a generation older, however, the margin may run out within the next year or two, and then the server will be left without patches — even though vulnerabilities in microcode and BMC are found regularly.

Exact dates for a specific model are always listed on the vendor’s website, and you need to check them before the purchase, not after.

What to check before buying

  • Serial number and warranty status. Both HPE and Dell let you check the warranty by serial number on their portal in a minute. How this works is covered in detail in the article on warranty and support for HPE and Dell. If the seller will not provide the serial number before payment, that is your answer.
  • Completeness. Rails, a cable management arm, bay blanks, power cables, drive trays — all of this is purchased separately and suddenly adds a noticeable amount to the budget. A server without drive trays means you will not install the drives until the package arrives.
  • Management licenses. iLO Advanced from HPE and iDRAC Enterprise from Dell are paid. Without them, there is no remote console, no image mounting and only part of the telemetry, so the machine will have to be administered physically.
  • Drive condition. Check the power-on hours and SMART counters, and for SSDs, the consumed write endurance. The drives in a used server are often the very component that fails first.
  • Power supplies and the controller battery. The RAID controller cache battery has its own service life; a depleted one turns a fast write cache into a disabled one.

Hidden costs that erode the savings

Licenses are counted by cores, not by the age of the server. This is the most underestimated item. Since VMware moved to subscriptions, there has been a minimum of 16 cores per socket (the 72-core minimum order introduced in spring 2025 was later canceled) — an old two-socket server with 8-core processors will cost exactly as much in licenses as a modern one with 32-core processors, but will do many times less useful work. More on this in the article on choosing processors for virtualization.

Electricity and cooling. The difference in energy efficiency between generations is significant: what used to require three servers is now done by one. Over five years, the electricity bill and the cooling load can eat up the entire price difference.

Cost of downtime. Calculate not the price of the server but the cost of an hour when it is not working. If that hour is expensive and a spare part takes a week to arrive, the savings on the purchase turn into a loss after a single incident.

When a refurbished server is the right choice

Expanding an existing fleet: adding an identical node to a working cluster is cheaper and simpler than introducing a new platform with different drivers and firmware. Test and training environments, where functional compatibility with production matters, not performance. Cold standby — a machine that waits for its moment in case the primary one fails. A branch office or warehouse with an undemanding workload: file service, domain controller, video surveillance.

When not to buy one

Production workloads where downtime is costly — here you need a vendor service contract with a guaranteed response time, and it cannot be arranged for equipment without support. Projects with a horizon of five years or more: the platform will go out of support before it pays for itself. New types of workloads — accelerators, NVMe, DDR5 memory requirements: the old generation simply cannot handle them, and an upgrade will not solve that.

How to calculate honestly

Add up the cost of the equipment, licenses for the required number of cores, the cost of support for the entire service life, and an estimate of power consumption over the same years. Then do the same for a new server, but keep in mind that the new one will need fewer nodes for the same useful workload. It often turns out that the gap in the final total is much smaller than the gap in the price tag — and sometimes the new option comes out cheaper.

For help with this calculation and configuration selection, write to us. The catalog has both current generations and proven platforms from previous years: HPE ProLiant, Dell PowerEdge, and the whole section is Servers.

Which tasks refurbished hardware can handle and which it cannot, and how much you actually save with Gen10 versus Gen11 based on our price list, is covered in the article new or refurbished server: where the savings are justified.

Related collections and tools: Dell PowerEdge, HPE ProLiant Gen11, Server configurator.

Questions and answers

Can vendor support be arranged for a used server?

Sometimes yes: vendors have programs for putting equipment under service after an inspection, but this is paid, and for equipment past the end of the platform’s service life, no contract will be issued at all. This needs to be clarified before the purchase, for the specific serial number.

How much cheaper is refurbished equipment?

The spread is wide and depends on the generation, condition and configuration, so the question can be answered in figures only for a specific item. Focus not on the discount percentage but on the total cost of ownership over the planned service life.

What if the server has already been bought from a private seller?

Check the warranty status by serial number, update the microcode and BMC firmware to the latest available versions, read the SMART data from all drives and replace those nearing the end of their life, and check the RAID controller battery. And set aside spare components: for equipment without support, your own stock of spare parts is the only way to recover quickly.

Related reading