● Model No.: FusionServer G8600
● Maximum memory: 32 x DDR5 DIMMs, 5600MT/s; Maximum 8TB (32 x 256GB)
● Processor: 2 x 4th/5th generation Intel Xeon Scalable processors, up to 385W/CPU
● Main features: Top-tier 8-GPU computing power, fully modular architecture, extreme energy efficiency, and high reliability.
● Application scenarios: Large-scale model training, supercomputing HPC, and massive data analysis
● Warranty: 1 years
● MOQ: 1 pieces
Core selling points
1. Flagship-level 8-GPU computing power, the first choice for large model training
Supports top-tier HGX H20/H100 8-GPU modules, leading the industry in single-machine computing power, suitable for LLM, multimodal, and autonomous driving large model training.
PCIe 5.0 retimer-free design, zero data link loss, maximizing GPU performance.
2. Fully modular, cable-free, extremely easy to maintain
Six modules—CPU/GPU/Storage/IO/Power Supply/Fan—are all independent, hot-swappable, and cable-free.
Components can be replaced without removing the server, reducing maintenance time by 2.5 times.
3. Dual-zone power supply, ultimate energy efficiency
54V (GPU) + 12V (CPU) dual-plane power supply reduces conversion losses, saving 76W+ per unit.
Self-developed 96% titanium power supply, 2% more efficient than industry platinum power supplies, saving up to 200W per unit.
Innovative GPU MPC intelligent cooling reduces fan power consumption by 1.1% and improves 8080++ fan cooling capacity by 30%.
4. Flexible topology, adaptable to all scenarios. One-click software switching between NVIDIA high-performance and balanced dual topologies balances training speed and cluster scalability.
12×PCIe 5.0 + OCP 3.0, supporting IB/200G high-speed network cards, eliminating I/O bottlenecks.
Application scenarios
AI Training and Inference: It forms the crucial computational foundation for training deep learning models, enabling faster iteration of models in image, speech, and natural language processing, allowing for quicker optimization and performance improvements.
High-Performance Computing (HPC): Research institutions, meteorological departments, and oil companies often require large-scale parallel computing, such as calculations in molecular dynamics and fluid mechanics. HPC meets these needs, helping to solve complex problems.
Intelligent Cloud and Data Analytics: In public or private clouds, it's a key component providing AI computing power, supporting various AI services on the cloud and enabling faster operation of applications like big data analytics and video/image analysis.
Healthcare and Finance: In the healthcare industry, it assists in PACS image analysis and supports AI-assisted diagnosis in medicine; in the financial industry, core businesses such as high-frequency trading, risk control, and robo-advisors rely heavily on it.
Parameters
| parameter category | Specification Details |
| product form | 8U rack mounted server |
| processor | 2 x 4th/5th generation Intel Xeon Scalable processors, up to 385W/CPU |
| chipset | Emmitsburg PCH |
| memory | 32×DDR5 DIMM,5600MT/s; Maximum 8TB (32 × 256GB) |
| local storage | Front end: 8 × 2.5 "SAS/SATA HDD; Rear end: 8 × 2.5" NVMe SSD; Startup: 2 × M.2 SSD (RAID 1) |
| RAID support | 0/1/10/5/50/6/60; Cache supercapacitor protection |
| GPU module | 1 × 8-GPU module (compatible with NVIDIA HGX H2O/H100/A100 8-GPU) |
| PCIe Expansion | 12 x PCIe 5.0 x16+1 x OCP 3.0 network card; Retimer free design |
| network | 1 x OCP 3.0 (100G/200G/IB), supports hot swapping |
| heat dissipation | GPU area: 10 × 54V intelligent fan; CPU area: 5 × 12V intelligent fan; N+1 redundancy |
| power supply | 54V (N+M redundancy)+12V (1+1 redundancy) titanium power supply; Dual bus power supply |
| management | iBMC V6; Redfish/SNMP/IPMI; KVM; FusionDirector optional |
| Size | 352.8 × 447 × 925 mm; approximately 65kg (fully equipped) |
| reliability | Full module hot swappable TPM 2.0、 Safe startup, dual plane power supply |
Email: Lilicheng0510@163.com