Home / GPU servers

GPUs

GPU servers and GPU systems for AI and HPC

Data centre GPU servers, rack scale GPU systems and PCIe GPUs, configured to your specification and quoted in writing. The cooling to match is scoped alongside, so the systems and the cooling can be supplied against one written quote.

An unbranded open frame server rack holding stacked 2U servers, with neatly dressed cables, in a bright, quiet room

01

What it is

Multi GPU servers are the building block of most AI and high performance computing deployments. We scope the GPU count, host CPU, memory, storage and interconnect with you, then source the configuration and quote it in writing.

We quote for multi GPU rack servers, rack scale GPU systems, and PCIe GPUs and workstations, along with the memory, storage and interconnect for those systems. Servers, storage and networking will follow as separate categories.

02

Who it is for

  • AI training clusters, inference and edge, and enterprise private AI.
  • Research labs, engineering and simulation, data centres and colocation.
  • Organisations building AI and high performance computing capacity that want GPU systems and the cooling to run them planned together.
03

Categories

  • 01
    An unbranded multi GPU server, lid removed

    GPU servers

    Multi GPU servers for training and inference, configured to your specification.

    SuppliedView category
  • 02
    An integrated GPU rack with its front door open

    Rack scale GPU systems

    Integrated racks for larger clusters, scoped with you and your site.

    SuppliedView category
  • 03
    Two unbranded data centre GPU cards on a plain surface

    PCIe GPUs and workstations

    Data centre and workstation GPUs for existing servers and development teams.

    SuppliedView category

Coming later

  • 04

    Servers

    General purpose rack servers to sit alongside GPU systems.

    PlannedView category
  • 05

    Storage

    NVMe and capacity storage for training data and checkpoints.

    PlannedView category
  • 06

    Networking

    Switching, adapters, optics and cabling for clusters.

    PlannedView category

Images are illustrative and do not show specific products.

04

Classes we quote for

RefClassTypical configurationOrder basis
SOS-01Multi GPU rack serversRack-mount node with four or eight GPUs, dual CPU, large memory, high speed fabricPer node or per rack
SOS-02Rack scale GPU systemsIntegrated racks of nodes with networking, scoped with your sitePer rack
SOS-03PCIe GPUs and workstationsData centre and workstation cards, single or multi card, with or without a hostPer card
SOS-04Memory, storage and interconnect for GPU systemsNode memory, NVMe storage, adapters, optics and cabling for the systems aboveWith a system or separately

Generic classes only. No models, vendors or prices are published.

05

An unbranded multi GPU server, lid removed

Key considerations for GPU servers

  • GPU count per node. Four or eight accelerators per chassis, and what that means for power and cooling.
  • Interconnect. Ethernet or fabric between nodes, and the switching to match.
  • Memory. Host memory per node, sized for the workload rather than the catalogue.
  • Power and cooling. Per rack draw and the facility's ability to remove the heat. See Cooling.
  • Rack space. Units per node, rail kits and the weight your floor can take.

Each quote states the full configuration, quantity, price, validity, and payment and delivery terms, so there is nothing to interpret later.

Specifications are provided with your quote. Stock and availability on application. Cannot find what you need? Request a quote.

Images are illustrative and do not show specific products.

06

Cooling to match

In general terms. The right answer depends on the configuration; we confirm it in the quote.

RefClassCooling to matchRead more
SOS-01Multi GPU rack serversAir cooled or liquid cooled, depending on the configuration. Dense nodes usually need liquid at the chip or at the rack door.Direct-to-chip, Rear door
SOS-02Rack scale GPU systemsUsually liquid cooled, with a CDU in the rack or the row. Depends on the configuration; we confirm this in the quote.CDUs, Direct-to-chip
SOS-03PCIe GPUs and workstationsUsually air cooled. Either, when installed in a dense server.Cooling hub
SOS-04Memory, storage and interconnectFollows the system it goes into.Cooling hub

Air cooled or liquid cooled?

Qualitative only. Both are valid; the site usually decides.

QuestionAir cooledLiquid cooled
What it needs from the roomEnough cold air supply and hot air return for the rack, and a room that can carry the heat away.A coolant loop: facility water, or a liquid-to-air unit where there is none. Pipework, a CDU and leak management.
Where it fitsExisting rooms, lower density racks, smaller deployments, edge sites.Dense GPU racks, new builds, rooms that already have water, rack scale systems.
What to ask usWhether the configuration you want is available air cooled, and what that means for rack count.Which approach suits your site, what the quote should include, and what the facility must provide.
07

Checks we carry out

Sanctions and export control

Before any quote is finalised, we screen the buyer, its owners and the end user against UK sanctions and export control lists. Where a licence is required, no order proceeds until it is in place.

Know your customer

On a first order we verify the buyer's identity, registration and beneficial ownership, and we may ask for supporting documents.

Payment

Payment terms are agreed with each order and set out in the quote.

End use and destination

A declaration of end use and destination country is required for every order. We may ask for an end user undertaking.

The same checks apply to cooling equipment and to combined quotes. A quote may be declined on compliance grounds, and we may decline or not respond to any enquiry.

08

How we work

Five steps, whether the request is for GPU systems, cooling, or both. Each enquiry carries a reference from receipt, and nothing is agreed until a written quote is accepted.

  1. 01

    Scoping

    We confirm workload, configuration, quantity and site constraints before anything else.

  2. 02

    Configuration and validation

    We check the bill of materials, power, cooling and interconnect fit before quoting, and check the system against the cooling and power you have or plan.

  3. 03

    Sourcing

    We source the configuration and confirm details in the written quote.

  4. 04

    Compliance and checks

    Sanctions, export control, end use and know your customer checks before any quote is final.

  5. 05

    Logistics and delivery

    We arrange delivery to your site and confirm the terms in writing.

The five steps in detail

09

Related questions

C02Do GPU servers need liquid cooling?

Not always. Many configurations are available air cooled. Denser nodes and rack scale systems are increasingly designed for liquid, and some are only offered that way. It depends on the configuration; we confirm this in the quote.

Q07Do you supply hardware other than GPUs?

Yes. Alongside GPU systems we now supply cooling equipment for GPU deployments: direct-to-chip kits, coolant distribution units and rear door heat exchangers. Servers, storage and networking are planned as later categories, and we will say so on the GPUs page when they are available. See Cooling.

Q14Can you quote GPU systems and cooling together?

Yes. That is the point of offering both. Choose "GPU systems and cooling" on the contact form, tell us about the site, and one written quote covers the systems and the cooling equipment. See Cooling.

Q02What should I include in a quote request?

Your name, work email, company and a short message are enough to start. For a quote, add the platform or system class, quantity range, interconnect, memory per node, timeline, delivery region and end use country under Add project details. Part numbers help.

Q03Can I send a bill of materials or part list?

Yes. Paste it into the Message field. For longer lists, send the form first, then email the file to sales@duvaracompute.com quoting your reference.

Q04Do you publish prices?

No. We do not publish prices or stock. If we can help, we respond to a request with a written quote.

All questions

10
  • Liquid cooling

    Direct-to-chip, CDUs and rear door heat exchangers, for dense racks.

  • AI infrastructure

    Training clusters, inference and edge, enterprise private AI.

  • HPC

    Research labs, engineering and simulation, data centres and colocation.

  • Colocation

    Colocation space in mainland UK and Europe, sourced with your systems.

Request a quote

Ask a question or request a quote for GPU systems, cooling, or both. We reply to enquiries as soon as we can.