Skip to main content

Business · Hosting · GPU servers

IP 4U-E12G

4U rackmount server for HPC and AI with up to 12 GPUs, AMD EPYC 7003 and 4,192 GB DDR4 memory. PCIe 4.0 throughout — built for heavy compute.

Up to 12 GPUs
6 quad-slot (x16) or 12 dual-slot (x8) PCIe 4.0
AMD EPYC 7003
Milan architecture, socket SP3, PCIe 4.0 throughout
Up to 4,192 GB RAM
16 DIMM slots, DDR4 RDIMM 2DPC
4,800W redundant power
2+1 redundant 1,600W Platinum PSUs, hot-swap

Technical specifications

What you get

Form factor4U rackmount, 19 inch
Dimensions940 × 440 × 176,5 mm
ProcessorSingle socket SP3 — AMD EPYC 7003 series (Milan)
GPU support (max)6 × quad-slot PCIe 4.0 x16 OR 12 × dual-slot PCIe 4.0 x8
Memory16 DIMM (2DPC), up to 4,192 GB DDR4 RDIMM
NVMe storage4 × hot-swap U.2 NVMe (optional) + 2 × M.2 PCIe 4.0 x4
Networking2 × 1GbE RJ45 (Intel i350-AM2) + 1 × OCP 3.0 PCIe 4.0 x16
Power supply4,800W (2+1 redundant, 1,600W Platinum), AC front/rear
Fans6 × 7,200 RPM high static pressure, hot-swap
ManagementIPMI

Options & components

À la carte

Prices excl. VAT. Components marked "Quote" — contact us for current pricing.

GPU accelerators

ModelVRAMSlot widthCoolingNotePrice excl. VAT
NVIDIA H200 NVL141 GB HBM3eDual-slotPassivQuote
NVIDIA RTX PRO 6000 SE96 GB GDDR7Dual-slotPassivServer EditionQuote
NVIDIA RTX PRO 6000 WE96 GB GDDR7Dual-slotPassivWorkstation EditionQuote
NVIDIA H100Dual-slotPassivQuote
NVIDIA L40S48 GB GDDR6Dual-slotPassivQuote
NVIDIA RTX 6000 Ada48 GB GDDR6Dual-slotFläktQuote
IP RTX 5090 AI32 GB GDDR7Dual-slotPassiv3 års garantiQuote
NVIDIA RTX A550024 GB GDDR6Dual-slotFläktQuote
NVIDIA RTX 4090 (48 GB)48 GB GDDR6XDual-slotAktiv3 års garantiQuote
IP RTX 4090 AI24 GB GDDR6Dual-slotPassiv3 års garantiQuote
NVIDIA RTX A450020 GB GDDR6Dual-slotFläktQuote
NVIDIA RTX A500024 GB GDDR6Dual-slotFläktRefurbished, 1 årQuote
NVIDIA RTX A400016 GB GDDR6Single-slotFläktRefurbished, 1 årQuote

CPU (socket SP3)

ModelCoresBase freq.L3Price excl. VAT
AMD EPYC 7282162,80 GHz64 MBQuote
AMD EPYC 7313163,00 GHz128 MBQuote
AMD EPYC 7343163,20 GHz128 MBQuote
AMD EPYC 7373X163,05 GHz768 MBQuote
AMD EPYC 73F3163,50 GHz256 MBQuote
AMD EPYC 7352242,30 GHz128 MBQuote
AMD EPYC 7413242,65 GHz128 MBQuote
AMD EPYC 7443242,85 GHz128 MBQuote
AMD EPYC 7473X242,80 GHz768 MBQuote
AMD EPYC 74F3243,20 GHz256 MBQuote
AMD EPYC 7453282,75 GHz64 MBQuote
AMD EPYC 7452322,35 GHz128 MBQuote
AMD EPYC 7513322,60 GHz128 MBQuote
AMD EPYC 7543322,80 GHz256 MBQuote
AMD EPYC 7573X322,80 GHz768 MBQuote
AMD EPYC 75F3322,95 GHz256 MBQuote
AMD EPYC 7643482,30 GHz256 MBQuote
AMD EPYC 7663562,00 GHz256 MBQuote
AMD EPYC 7702642,00 GHz256 MBQuote
AMD EPYC 7713642,00 GHz256 MBQuote
AMD EPYC 7742642,25 GHz256 MBQuote
AMD EPYC 7763642,45 GHz256 MBQuote
AMD EPYC 7773X642,20 GHz768 MBQuote
AMD EPYC 7C13642,00 GHz256 MBQuote
AMD EPYC 7H12642,60 GHz256 MBQuote

Memory (per DIMM)

CapacitySpeedTypePrice excl. VAT
16 GB3200 MHzDDR4 ECC RegQuote
32 GB3200 MHzDDR4 ECC RegQuote
64 GB3200 MHzDDR4 ECC RegQuote
128 GB3200 MHzDDR4 ECC RegQuote
256 GB3200 MHzDDR4 RDIMM/LRDIMM-3DSQuote

Use cases

What it is used for

AI training & fine-tuning

Train or fine-tune LLMs and diffusion models on your own data — no per-token or per-GPU-hour cloud costs.

HPC & simulation

CFD, FEA, climate modelling and scientific compute that demand high memory bandwidth and PCIe 4.0 throughput.

Parallel inference

Run up to 12 GPUs in parallel to serve multiple large models simultaneously — ideal for internal AI platforms.

Rendering & VFX

GPU rendering of 3D scenes and VFX pipelines with full PCIe 4.0 x16 bandwidth per card in the 6-GPU config.

Delivery model

Choose how to own and run it

We deliver the server as a purchase, hosted in a Swedish datacenter, or as colocation. The quote always specifies exactly what is included.

One-off cost

Buy the hardware

One-off investment. You own the server and place it wherever you like — on-premise, in our datacenter or third-party colocation.

Monthly

Hosted in our datacenter

We host the server in our Swedish datacenter — power, cooling, redundant network and monitoring included. Fixed monthly cost.

Flexible

Colocation

You buy the server and rent space in any datacenter. You pay for rack space, power and network — we deliver the hardware.

Ready to configure your server?

Tell us which GPU, how much memory and which delivery model you want — we will reply with a quote within one business day.

Request a quote

Frequently asked questions

Which GPUs fit in the IP 4U-E12G?

The server supports up to 6 quad-slot cards (e.g. NVIDIA H100, A100 SXM or RTX 6000 Ada) at PCIe 4.0 x16, or up to 12 dual-slot cards (e.g. RTX 4090, L40S) at x8. Contact us to select the right GPU for your use case.

How much does the server cost?

The price depends on CPU model, GPU selection, RAM and storage. The base unit starts around 36,000 SEK excl. VAT. Request a quote and we will spec it exactly for your requirements.

What is the lead time?

Contact us for lead time. We can normally deliver within 1–2 weeks depending on configuration. GPU card availability may affect lead time.

Can you install and manage the server for us?

Yes. We deliver as pure hardware sales, hosted in our datacenter, or as a fully managed solution. Describe your needs in the quote request and we will tailor the delivery model.

Help Center

Most Popular Articles

Loading...

Latest Articles

Loading...

Categories

Loading...

IP 4U-E12G — up to 12 GPUs, AMD EPYC | Internetport