TG780V3
LARGE MODEL INFERENCE AI SERVER

NEW PLATFORM, EXCEPTIONAL PERFORMANCE

Utilizes the Intel 4th/5th Gen Xeon Scalable processor platform, featuring comprehensive upgrades in CPU compute power, PCIe speed,and memory bandwidth to fully unleash GPU performance.

Supports 8x 600W GPUs; each card utilizes PCIe 5.0 x16, delivering efficient heterogeneous computing power.

CPU-GPU direct-connect design ensures high efficiency and low latency, significantly improving data transfer efficiency compared to PCIe Switch architectures.

COMPATIBLE WITH FAN AND BLOWER CARDS

Superior compatibility, optionally supports either fan-cooled (active) or blower-style (turbine) GPU cards.

Supports 8x four-slot fan-cooled cards or 8x dual-slot blower-style cards.

FLEXIBLE CONFIGURATION & ON-DEMAND SELECTION

Supports up to 11 standard PCIe 5.0 slots; multiple PCIe configurations available.

Optional 1 x OCP 3.0 NIC with multiple speed options available.

Supports 8 x 3.5"/2.5" SAS/SATA HDDs/SSDs; optionally supports 2/4 x U.2 NVMe SSDs, balancing high capacity and high-performance local storage.

STABLE, RELIABLE & INTELLIGENT MANAGEMENT

Key system components feature redundant and hot-swappable designs, complemented by tool-less maintenance, enhancing fault recovery efficiency and system availability.

Integrated intelligent management chip provides an open management platform supporting multiple protocols like IPMI 2.0, Redfish, and SNMP.

Supports various management functions including KVM, Virtual Media, status monitoring of key components, anomaly alerts, and more, offering comprehensive remote system-level intelligent management capabilities.

Beijing Puhua Haotian Technology Co.