Used & Tested AI GPUs: what a meaningful test report should contain
Previously deployed hardware needs a documented condition and test scope. A useful report connects the unit's identity, operating conditions and results to the equipment you are purchasing.

Identify the actual unit
Record the model, serial number, memory variant, physical format and inspection date. Used and tested hardware has been deployed previously, then inspected and tested. Document the offered condition, included accessories and any available service history. The report should identify the checks performed on the offered unit; results for another GPU of the same model do not establish its condition.
Identify the host server as well. Firmware, drivers, power and cooling affect the test environment. A report listing the host and installed GPUs makes unexpected results easier to interpret and gives your installation team a more useful starting point when the equipment is delivered.
Specify the test scope
NVIDIA DCGM provides software-readiness, memory, compute, interconnect and stress diagnostics, with coverage depending on the platform and selected tests. NVIDIA distinguishes these checks from comprehensive hardware field diagnosis and its RMA process. A diagnostic pass is evidence for the checks performed; it does not establish universal certification or replace assessment of the complete system. NVIDIA: DCGM diagnostics and their scope ↗
Agree the required checks before ordering. Record tool versions, test selection, duration and skipped checks. Request relevant error counters and observations alongside the final status. Sustained-load temperatures, power and clock behaviour help establish whether the equipment operated within the intended conditions throughout the test.
Assess the complete server
A server requires checks beyond its GPUs. Include system memory, processors, storage health, networking, power supplies and management access in the acceptance plan. For multi-GPU work, verify the actual topology and a suitable shared workload. Supported platform configuration remains important even when individual cards pass diagnostics. NVIDIA: certified systems and supported configurations ↗
Retain the starting state and final report. Consistent identifiers and baseline results help resolve faults after installation. Record cosmetic issues and agreed exceptions too, so an expected condition is not mistaken for transport damage and the receiving team knows exactly what was accepted before shipment.
Connect testing with warranty
Our Secondary Market (Used & Tested) range carries a 18-month commercial warranty, in addition to mandatory consumer rights. The written offer identifies the equipment, documented condition, test scope and applicable claim arrangements. Include additional acceptance requirements in that offer, especially where the hardware will join an existing cluster with its own operating standards or documented qualification process.
The objective is practical confidence. Know what was tested, under which conditions, what passed and how a problem will be handled. Those details are more useful than an impressive list of test names or an unsupported claim that a used GPU is effectively new.
Sources & further reading
Manufacturer specifications and purchasing references. Exact supplied configurations require confirmation.
Sources checked October 11, 2026.
Explore relevant equipment
Discuss your requirements
Share the model, configuration and quantity you need.





