Get a Free Quote

Our representative will contact you soon.
Email
Tel/WhatsApp
Name
Company Name
Message
0/1000
tg580v3 heterogeneous computing flagship-2

AI Server

Home >  Products >  Servers >  AI Server

All Categorys

Servers
Workstations
Storage Devices
Network Switches
Infrastructure Accessories

All Small Categorys

Servers
Workstations
Storage Devices
Network Switches
Infrastructure Accessories

TG580V3
HETEROGENEOUS COMPUTING FLAGSHIP

TG580V3 is a new generation flagship 4U dual-processor 8-card GPU server product launched by Aethlumis, based on the 5th/4th Gen Intel® Xeon® Scalable processors, fully upgraded in CPU, GPU, and I/O specifications, supports various mainstream AI acceleration cards in the industry, and has excellent performance, flexible architecture, strong scalability, rich configuration, and high reliability. It is suitable for application scenarios such as artificial intelligence, high-performance computing, and data analysis.

  • Description
Is there a problem? Please contact us to serve you!

Is there a problem? Please contact us to serve you!

Inquiry

EXCELLENT PERFORMANCE PREFERRED FOR ARTIFICIAL INTELLIGENCE

Supports 2 x 5th/4th Gen Intel® Xeon® Scalable processors, 4 x UPI interconnections, supporting up to 350W TDP;

Supports 10 x 450W GPU cards, providing ultimate heterogeneous computing power;

Supports GPU to CPU x32 transmission bandwidth, which is twice the industry x16, meeting the requirements of high communication bandwidth scenarios between CPUs and GPUs;

Supports 32 x DDR5 memory, with a maximum frequency of 5600MHz and a 75% increase in memory bandwidth.

 

LEADING ARCHITECTURE WITH FLEXIBLE CONFIGURATION

Supports two configurations: 8-card direct and 10 card switch, supports multiple GPU topologies, and flexibly matches the needs of different application scenarios;

Supports GPU Direct, support storage/network RDMA/GPU P2P direct data transmission, meet cluster deployment requirements;

Ultra high scalability, supports up to 15 x standard PCIe slots, can be configured with 8 x dual width GPUs+7 x PCIe standard cards+1 x OCP 3.0 network card;

Flexible and optional storage configuration, meeting the needs of high-capacity and high performance local storage, supporting up to 24 x U.2 NVMe.

 

STABLE, RELIABLE, AND INTELLIGENT MANAGEMENT

The key components of the system are designed with redundancy and hot swapping, and support tool free disassembly and assembly to improve maintenance efficiency and system availability;

Integrate intelligent management chips, provide an open management platform, support multiple management protocols such as IPMI2.0, Redfish, SNMP, etc;

Supporting various management functions such as remote KVM, virtual media, critical component status monitoring, and abnormal alarms, achieving comprehensive remote system level intelligent management.

 

详情页.jpg