
Dell PowerEdge R740xd Refurbished 2U Server
Storage-dense 14th-generation 2U server with up to 32 drive bays, built to order from tested parts

Tested used Xeon and EPYC processors, pulled DDR4 and DDR5 RDIMMs, refurbished Dell and HPE servers and used data-centre GPUs

Storage-dense 14th-generation 2U server with up to 32 drive bays, built to order from tested parts

15th-generation 2U server with Ice Lake Xeon Scalable CPUs, PCIe Gen4 and up to 24 NVMe bays
Mainstream 2U HPE server with Xeon Scalable CPUs, up to 30 SFF bays and Smart Array P408i-a
Dense 1U two-socket HPE server with 24 DIMM slots and 8 SFF, 10 NVMe or 4 LFF chassis

2U Ultra platform on the X11DPU board with 24 DIMM slots and 24 hot-swap 2.5-inch bays

First-generation SR650 with Xeon Scalable CPUs, 24 DIMM slots and up to 24 front 2.5-inch bays

Pulled Ampere accelerator with 80 GB HBM2e, tested under load and supplied after end-user verification

Complete used 4U system with eight A100 SXM4 GPUs on an NVSwitch baseboard, 640 GB of HBM2e

Hopper accelerator with FP8 Transformer Engine and 80 GB HBM2e, pulled from cluster service and tested

Ampere workstation card with 48 GB of ECC GDDR6, pulled from render and visualisation systems and tested

Ada Lovelace flagship with 16,384 CUDA cores and 24 GB GDDR6X, tested partner cards from used systems
20-core Skylake-SP processor for LGA 3647 servers, removed from decommissioned systems and tested
24-core Cascade Lake Refresh processor with a 3.0 GHz base clock, pulled from decommissioned LGA 3647 servers

32-core Ice Lake-SP processor with DDR4-3200 and PCIe 4.0, removed from off-lease LGA 4189 servers
24-core Cascade Lake Platinum processor for two- to eight-socket servers, pulled and tested

32-core Ice Lake-SP cloud-provider SKU with a 2.9 GHz base clock, pulled from retired hyperscale servers

56-core Sapphire Rapids processor with DDR5-4800, PCIe 5.0 and AMX, pulled from retired AI and cloud servers
14-core Broadwell-EP processor for LGA 2011-3 servers, the standard upgrade for Gen9 and 13G fleets

64-core Milan flagship with 256 MB L3 and 128 PCIe 4.0 lanes, removed from decommissioned SP3 servers

64-core OEM-only Milan processor built for Google Cloud, 240 W, pulled from retired hyperscale nodes

Dual-rank 32 GB registered DIMM for Skylake and Cascade Lake servers, full-pass tested after removal

Dual-rank 64 GB registered DIMM at the native speed of Ice Lake and EPYC Milan, pulled and full-pass tested
32 GB registered DIMM in 2Rx4 or 2Rx8 organisation for Ice Lake, Rome and Milan servers, pulled and tested
Quad-rank 64 GB load-reduced DIMM for high-capacity Cascade Lake servers, pulled and full-pass tested

Dual-rank 64 GB DDR5 registered DIMM for Sapphire Rapids and EPYC Genoa servers, pulled and full-pass tested
Hyperscalers and large enterprises refresh hardware every three to five years, and the processors, memory and GPUs they pull are the most cost-effective way to build or expand compute for hosting, rendering, AI experimentation and lab environments. This category lists the pulled and refurbished parts that move in volume: Intel Xeon Gold and Platinum, AMD EPYC Milan, DDR4 and DDR5 registered memory, complete Dell PowerEdge and HPE ProLiant servers and data-centre GPUs such as the NVIDIA A100.
Buyers are hosting and colocation providers, system integrators building budget servers, universities, mining and rendering farms, and resellers in the CIS, Africa and South Asia. Every item is tested before shipment: processors are run in a reference platform, memory passes a full-pass test, servers are burned in and GPUs are stress-tested, with results available on request.
Refurbished and pulled hardware is supplied through licensed partners of InTechnologies FZCO who issue the contract, invoice and the partner warranty. We coordinate sourcing from decommissioning programmes and verified brokers, consolidation in Dubai Airport Freezone and DDP or DDU delivery. Used data-centre GPUs remain subject to US export regulations and end-user verification.
Pulled parts were removed from working servers during a fleet refresh, not repaired. Each part is tested and supplied with a limited warranty from our partner, typically 90 days to 12 months depending on the item, stated in the quotation.
For processors we can filter by stepping and for memory by manufacturer and speed grade. Production dates and exact batches are confirmed per lot before shipment.
Used A100, H100 and similar data-centre GPUs are controlled under US export regulations regardless of condition. Our partner verifies the end user and destination before confirming the order.