Product catalog
Find the model you need and request a quote. We reply with an all-in price and a delivery date.
41 of 41 products
NVIDIA
AI AcceleratorsHopper-generation accelerator with the largest memory pool in its family, suited to serving and fine-tuning large language models that outgrow 80GB parts.
NVIDIA
AI AcceleratorsThe reference training accelerator for multi-GPU nodes, deployed eight to a baseboard with full NVLink bandwidth between every GPU.
NVIDIA
AI AcceleratorsA PCIe card version of Hopper with extra memory, built for LLM inference in standard air-cooled servers and bridged in pairs over NVLink.
NVIDIA
AI AcceleratorsA general-purpose data-centre GPU that covers inference, fine-tuning, rendering and virtual workstations from one PCIe card.
NVIDIA
AI AcceleratorsA low-power, low-profile card for adding inference and video capacity to servers that have no spare power or slot height.
NVIDIA
AI AcceleratorsPrevious-generation workhorse that remains a sound choice for established training pipelines and for matching an existing A100 fleet.
NVIDIA
AI AcceleratorsWorkstation-class card with display outputs and active cooling, for engineering desks that need local model development alongside visualisation.
AMD
AI AcceleratorsHigh-memory accelerator for teams standardising on ROCm, able to hold very large models on fewer devices.
Intel
AI AcceleratorsAccelerator with Ethernet networking built into every device, so clusters scale out over standard switches rather than a separate fabric.
AMD
Server CPUsTop-bin general-purpose 4th Gen EPYC for dense virtualisation and consolidation, replacing several older two-socket hosts with one.
AMD
Server CPUsA balanced 64-core part with higher clocks than the 96-core bins, a common pick for databases and per-core licensed software.
AMD
Server CPUsA 32-core option that keeps the full memory and PCIe capability of the platform at a lower power envelope.
AMD
Server CPUsDensity-optimised processor for cloud-native and containerised workloads where thread count per rack unit matters most.
AMD
Server CPUs5th Gen EPYC with markedly higher boost clocks, frequently specified as the host processor in current GPU servers.
AMD
Server CPUsThe highest core count in the 5th Gen range, aimed at scale-out services and large virtual machine estates.
Intel
Server CPUs4th Gen Xeon Scalable flagship with built-in AMX acceleration, and the host processor used in many eight-GPU reference systems.
Intel
Server CPUs5th Gen Xeon drop-in upgrade for LGA4677 platforms, bringing more cores, a much larger cache and faster memory.
Intel
Server CPUsMainstream 32-core Xeon for general virtualisation and application servers where platform cost per node is the constraint.
Intel
Server CPUsPerformance-core Xeon 6 with twelve memory channels and MRDIMM support, for memory-bandwidth-bound compute and AI host roles.
NVIDIA
AI ServersA fully integrated eight-GPU system delivered as one validated unit of hardware and software, for teams that want a supported reference platform rather than a custom build.
NVIDIA
AI ServersBlackwell-generation integrated system for organisations planning new training capacity, with power and cooling needs that should be checked against the facility first.
Supermicro
AI ServersAir-cooled 8U platform built around an eight-GPU HGX baseboard with Intel host processors, widely used as a configurable alternative to integrated systems.
Supermicro
AI ServersThe AMD-hosted counterpart of the 8U HGX platform, for estates that standardise on EPYC.
Dell
AI ServersA 6U eight-way accelerator server offered with several GPU families, fitting organisations that already run PowerEdge tooling and support.
Dell
AI ServersA 2U server for PCIe accelerators, a practical step for inference and mixed workloads before committing to eight-GPU nodes.
Lenovo
AI ServersA flexible 3U chassis that takes either PCIe cards or a four-GPU HGX board, useful where one platform has to serve several workload types.
HPE
AI ServersA GPU-dense take on the familiar 2U ProLiant, for teams that manage their fleet through iLO and want accelerators in the same operational model.
NVIDIA
NetworkingThe standard leaf and spine switch for 400Gb/s InfiniBand GPU fabrics.
NVIDIA
NetworkingHigh-radix Ethernet switch for AI fabrics built on RoCE rather than InfiniBand.
NVIDIA
NetworkingNetwork adapter that attaches servers and GPUs to 400Gb/s fabrics, available for both InfiniBand and Ethernet.
NVIDIA
NetworkingA data processing unit that moves networking, storage and security services off the host processor and onto the adapter.
Arista
NetworkingA 400G fixed switch for data-centre leaf and spine roles, running the same EOS software as the rest of an Arista estate.
Cisco
NetworkingA 400G Nexus 9000 switch for spine and aggregation layers in NX-OS or ACI fabrics.
Samsung
Memory & StorageThe most commonly specified server memory module for current DDR5 platforms, bought in matched sets to fill every channel.
Micron
Memory & StorageA non-binary capacity that sits between 64GB and 128GB modules, letting memory be sized to the workload instead of rounded up.
Samsung
Memory & StorageA read-intensive data-centre NVMe drive, a dependable default for boot, cache and general server storage.
Micron
Memory & StorageMainstream data-centre NVMe drive offered across several form factors, which simplifies standardising one model across different chassis.
KIOXIA
Memory & StoragePCIe Gen5 enterprise drive for keeping GPU servers fed with data, roughly doubling sequential throughput over Gen4 parts.
Solidigm
Memory & StorageVery high capacity QLC flash for data lakes and content stores, replacing racks of hard drives where reads dominate.
Seagate
Memory & StorageEnterprise capacity hard drive for bulk, backup and archive tiers where cost per terabyte decides the design.
Western Digital
Memory & StorageA 22TB helium drive for dense storage enclosures and object storage nodes.
Specifications are indicative and confirmed on quote. Photos are representative of the hardware type, not the listed product. Product and brand names belong to their respective owners and are used for identification only.
