ZC4850
Inference and training in one
ZC4850 Overview
All-new architecture for extreme performance: supports 1 or 2 4th/5th Gen Intel® Xeon® Scalable processors; backed by the latest DDR5 and PCIe 5.0 technologies to accelerate data transfer rates; direct GPU-CPU interconnect delivers extreme multi-card GPU computing power.
Flexible Topology for Key Applications
CPU-GPU supports direct connection, accommodating up to 8 x 600W dual-width GPU cards for high-speed, stable operation; supports common and performance CPU-GPU interconnect topologies to suit more application scenarios.
Rich Ecosystem, Stable & Reliable
Compatible with both foreign and domestic AI accelerator cards; critical system components use redundant design and functional parts use modular design, improving product reliability and serviceability; supports smart remote management and smart remote fault diagnosis, ensuring efficient and stable 24/7 operation.
| System Model | ZC4850 |
|---|---|
| Processor | Supports 1 or 2 4th/5th Gen Intel® Xeon® Scalable processors, up to 350W TDP |
| Chipset | Intel® C741 series server-grade chipset |
| GPU Card | Supports up to 8 x 600W dual-width GPUs |
| Memory | 32 DDR5 memory slots, supports DDR5 4400/4800/5600MHz memory |
| Storage Controller | 2 x Slimline 4x ports (SATA), 2 x 2280/22110 M.2 (SATA/PCIe 3.0 x1) |
| Storage | 12 hot-swap 3.5"/2.5" SAS/SATA/NVMe drives |
| I/O Ports | Front: 2 x USB3.0, 1 x VGA; Rear: 1 x serial port, 2 x USB3.0, 1 x VGA, 1 x RJ45 management port (optional 2 x 10G RJ45 service ports) |
| PCIe Expansion | Supports up to 13 PCIe expansion slots (OCP 3.0 card optional PCIe 5.0 x8 or x16, mutually exclusive with 1 PCIe 5.0 standard single-width slot) |
| Power Supply | 3+1 or 2+2 redundancy (2000W/2200W/2600W/3200W) |
| Fans | Standard 12 x 6056 hot-swap N+1 redundant fans |
| Management | 1 x dedicated RJ45 management port |
| Security | Supports TPM 2.0 module, open-cover alarm, BMC/BIOS redundancy |
| Operating System | Microsoft Windows Server, Red Hat Enterprise Linux, SUSE Linux Enterprise Server, CentOS, Ubuntu, VMware ESXi, etc. |
| Server Dimensions | 176.5mm (H) x 444mm (W) x 850mm (D) |
| Operating Temperature | Operating temperature: 5°C - 35°C (no direct sunlight); Storage temperature: -40°C - 70°C |
Related Products
ZC2230
General-purpose server. A dual-socket general-purpose computing server based on the 3rd Generation Intel® Xeon® Scalable processors, delivering outstanding computing performance with versatile I/O expansion. It also provides ample storage capacity and is ideal for virtualization, data analytics, public and private clouds, as well as scenarios such as VDI, AI vision, big data processing and enterprise applications.
ZC7850
ZC7850 is an outstanding 7U dual-socket 8-GPU AI server with distinct advantages. Built on the 4th/5th Gen Intel® Xeon® Scalable processors and paired with high-performance GPUs, it delivers comprehensive upgrades across CPU, GPU and I/O specs, featuring high computing power, strong scalability, rich configuration and high reliability. It is ideal for AI inference, large...
ZC8850
Inference and training in one. With direct CPU-GPU interconnect, it supports 8 x 600W dual-width, triple-width, 3.5-slot and 4.5-slot GPUs. It features high performance and parallel computing, high scalability, high reliability, high energy efficiency, smart maintenance, modular design and an open ecosystem, meeting the needs of large-model training, AI inference, graphics processing and generative AI. Ideal for large models...
ZC6850
Inference and training in one. With direct CPU-GPU interconnect, it supports 600W dual-width, triple-width and 3.5-slot GPUs. It features high performance and parallel computing, high scalability, high reliability, high energy efficiency, smart maintenance, modular design and an open ecosystem, meeting the needs of various AI workloads. Ideal for cloud gaming, AI...