- 2U rack chassis for high-performance GPU server applications
- Dual-socket Intel® Xeon® scalable processor
- 8x PCIe x16 Gen3 slots for GPU and expansion cards
- 2x RJ-45 Ethernet ports with Intel® i350 Gigabit Controller
- Up to 6x SATA and 4x U.2 NVMe for flexible storage options
Description
SHARK A.I. 4-Slot Dual GPU server – High-performance AI and HPC platform
The SHARK A.I. 4-Slot Dual GPU server is a universal carrier-grade platform specifically designed for edge computing, artificial intelligence (AI) and high-performance computing (HPC). By combining a powerful dual-socket server node with PCIe expansion and PCIe switch technology, the system supports up to four NVIDIA® GPUs. Thanks to its modular architecture, the server adapts flexibly to a wide range of workloads and enables efficient GPU and CPU communication via configurable topologies and bandwidths. InfiniBand support also ensures seamless scaling across multiple GPU clusters.
Dual-socket architecture for maximum computing power
With two powerful CPU sockets, the SHARK A.I. offers exceptional compute performance and memory bandwidth optimized for data-intensive deep learning, machine learning and scientific computing applications. The supported single- and dual-root complexes enable individual configuration of the computing load. Single-root setups bundle all GPU resources for deep learning training, while dual-root configurations optimize CPU load for more efficient data processing.
Flexible GPU management and PCIe optimization
The SHARK A.I. 4-Slot Dual GPU Server offers dynamic PCIe lane allocation, allowing GPUs to be assigned to specific I/O interfaces or CPU cores. This improves the information flow and efficiency of multiple virtualized applications without the need for multiple systems. This intelligent resource allocation reduces the need for additional hardware, minimizing energy consumption, cooling and space requirements.
As a powerful Edge High Performance Computing (EHPC) solution, the SHARK A.I. enables agile and flexible integration into cloud infrastructures. The server offers the option of implementing different head nodes and freely adapting the number of GPUs per virtual machine (VM). This makes it the ideal solution for GPU-accelerated AI models, scientific simulations, remote virtualization and industrial automation.
Reliability and easy maintenance for continuous use
The high-quality hardware architecture of the SHARK A.I. was developed for uninterrupted operation. Hot-swappable components, redundant fan modules and redundant power supplies ensure maximum reliability. The GPU cards can be quickly installed or replaced by simply removing the top cover, enabling a maintenance-friendly configuration without long downtimes.
Specifications
| Product: SHARK A.I. 4-Slot Dual GPU server | |
|---|---|
| SYSTEM | |
| CPU | Dual-socket Intel® Xeon® scalable processor, max up to 165W TDP |
| Chipset | Intel® C621PCH |
| GPU | Supports dual slot GPUs |
| RAM | Up to 512GB DDR4 2666MHz (16x RDIMM slots) |
| Storage | 1x M.2 slot 6x 2.5″ SATA SSD/HDD 4x U.2 2.5″ NVMe + 2x 2.5″ SATA SSD/HDD (Built-to-Order) |
| Expansion | 8x PCIe x16 Gen3 slots(4x @ x16 or 8x @ x8 link) 1x PCIe x16 (x16 Gen3 link) and 1x PCIe x8 (HHHL) Supports a maximum of 4 dual-slot GPU cards |
| TPM | Yes |
| INTERFACE | |
| Ethernet | MAC/PHY supports 2x RJ-45 Ethernet ports Intel® i350 Gigabit controller |
| USB | 2x USB 3.0 (front) 1x USB 3.0 (rear) |
| COM | 1x COM (2×5-pin header) |
| Video | 1x VGA port |
| ENVIRONMENTAL | |
| Chassis | 2U 19″ rack enclosure |
| Cooling | 4x 80mm + 3x 60mm fan with smart fan control |
| Power Supply | 1600W 80+ AC/DC Platinum power adapter; 1+1 redundancy Input: 1000W: 100 ~ 127V AC / 12.9A 1600W: 220 ~ 240V AC/ 9.5A |
| Operating Temperature | 0°C ~ 35°C |
| Storage Temperature | -10° ~ 60°C |
| Humidity | 5% ~ 95%, non-condensing |
| Dimensions (W x H x D) | 435 x 88 x 620mm |
