E281-4U
Banner showing Puget Server E281-4U

Puget Server E281-4U MGX

This NVIDIA MGX™ server uses a pair of AMD’s EPYC 9005 processors to enable up to eight NVIDIA H200 NVL GPUs in a 4U chassis, offering tremendous performance for AI training and fine-tuning.

Overview

This GPU server uses a pair of AMD EPYC™ processors to support eight dual-width GPUs in a 4U rackmount chassis. Configurable with up to 8x NVIDIA H200 NVL GPUs sporting a combined total of over 1TB of VRAM, this server is the ultimate solution for machine learning, AI, and rendering workloads.

NVIDIA Certified Badge Logo

Not sure what you need?

and one of our experts will reply within 1 business day to help configure the right server for your workflow. If you don’t see what you are looking for, check out our custom servers page for more options.

System Core


2x

Intel E810 Dual 25/10G SFP28 PCIe x8 [add $253.05 each]
NVIDIA Mellanox Dual 25G SFP28 PCIe x8 [add $416.35 each]
Intel E810 Dual 100G QSFP28 PCIe x16 [add $506.45 each]
NVIDIA Mellanox ConnectX-7 400G OSFP PCIe x16 [add $1910.20 each]
This system includes onboard dual 10Gb RJ45 ports as well as a dedicated remote management port. Up to 5 additional network adapters can be selected as well.

Storage


Chassis & Cooling


This system requires four 200-240V power connections. For redundancy to be functional, the total power consumption of the system must be lower than the maximum output of three of the PSU modules.

Software


Also available with Windows Server 2022 or 2025. Ask your consultant or contact us for licensing costs.

Accessories


For H200 NVL only, please select an NVLink option matching the number of GPUs

Additional Information

Help us help you! We review each configuration to ensure you’re getting the right hardware. Any info you can provide about your workflow and software will help us provide you with a better experience.


System Cost

Loading…

per unit

Typically ships in 1-2 weeks

Contact us for lead times

Contact us for quotes for more than 100 units