GPU HGX-1 system

Extremely powerful and high density GPU server for HPC applications.

GPU HGX-1 system

This solution consist of 1U rackmountable Stingray server and 4U HGX-1 GPU Expansion box. This highly flexible HGX-1 Expansion Box supports configurations of eight high-performance SXM2 GPGPUs or eight PCIe-based GPGPUs or up to thirty-two low-powered GPGPU accelerators for challenging demands in a variety of HPC applications including: oil and gas exploration, financial services, cloud computing and deep learning.
The HGX-1 connects directly the Stingray server through the latest PCIe x16 Gen3 technology. Powered by three 1600W redundant power supplies, the HGX-1 GPU Expansion Box is truly designed for hyperscale AI workloads.

Specification

CPU type / platformx86 Intel
Form Factor5U Rackmount
CPU typeIntel® Xeon® Scalable
Storage4 x M.2 drives
Server typeGPU HGX-1 system
Availabilityavailable
Dimm slots24
CPU No. Max2
CPU Watt Max150
PSU standard1300W + 1600W redundant
Hot-swappable PSUy
SATA / SAS storage bays - front-
U.2 storage bays - front-
ChipsetC620
M.2 SATA slots4
PCIe x166
OCP1
On Board LAN2 x 1GbE
GPU8
FANS6
Hot-swappable fansy
Dimensions (w x h x d)432 x 43 x 949 + 441 x 172 x 947
Front panelBase server:
1 x Power button
4 x PCIe slot LED

GPU expansion:
1 x Power button with LED
1 x UID LED
1 x Attention LED
1 x RJ45 1GbE chassis management port
1 x RJ45 rack management port
Rear panelBase server:
1 x UART Port
2 x RJ45 1GbE ports (one for management)
2 x USB3.0 rear IO ports
1 x VGA port OCP card ports (SFP or QSFP)

GPU expansion:
1 x RJ45 port for rack management
3 x Mini-SAS HD connector (reserved)
NV Linkyes
GPU Cards support8 x high-performance SXM2 GPGPU or
8 x PCIe-based GPGPU or
32 x low-powered GPGPU accellerator
MS Windows Server 2008n
MS Windows 2012 R2n
MS Windows 2016n
Novell® SUSE® Linux Enterprise Servern
Red Hat® Enterprise Linuxy
CentOSy
Canonical Ubuntu Servery
VMware vSphere® ESXi™n
See full specification

Tell us about your needs