Dell PowerEdge XE9680 AI Server | 8x NVIDIA HGX H100 SXM5 80GB | Dual Xeon Platinum 8568Y+ | 8x NDR InfiniBand
  • Kategorie produktów:Serwery
  • Numer katalogowy:Dell PowerEdge XE9680 AI Server
  • Dostępność:In Stock
  • Stan:Nowy
  • Cechy produktu:Gotowy do wysyłki
  • Minimalne zamówienie:1 szt.
  • Cena katalogowa wynosiła:$387,999.00
  • Twoja cena: $360,294.00 Oszczędzasz $27,705.00
  • Czat teraz Wyślij e-mail

Oddychaj spokojnie. Przyjmujemy zwroty.

Wysyłka: Międzynarodowa wysyłka produktów może wiązać się z procedurami celnymi oraz dodatkowymi opłatami. Zobacz szczegóły

Dostawa: Proszę o dodatkowy czas, jeśli międzynarodowa dostawa podlega procedurze celnej. Zobacz szczegóły

Zwroty: 14 dni na zwroty. Sprzedawca pokrywa koszt zwrotu. Zobacz szczegóły

Darmowa wysyłka. Akceptujemy zamówienia zakupu NET 30 dni. Otrzymaj decyzję w kilka sekund, bez wpływu na Twoją zdolność kredytową.

Jeśli potrzebujesz dużej ilości produktów Dell PowerEdge XE9680 AI Server - zadzwoń do nas na bezpłatny numer WhatsApp: (+86) 151-0113-5020 lub poproś o ofertę na czacie na żywo, a nasz menedżer sprzedaży skontaktuje się z Tobą wkrótce.

Dell PowerEdge XE9680 AI Server | 8x NVIDIA HGX H100 SXM5 80GB | Dual Xeon Platinum 8568Y+ | 8x NDR InfiniBand

Keywords

Dell PowerEdge XE9680, NVIDIA HGX H100, AI Training Server, Intel Xeon Platinum 8568Y+, NDR InfiniBand, High Performance Computing, Generative AI Infrastructure, Buy Dell XE9680

Description

Propel your organization to the forefront of artificial intelligence with the Dell PowerEdge XE9680. This 6U air-cooled powerhouse is purpose-built for the AI era, featuring the relentless performance of the NVIDIA HGX H100 SXM5 platform. Equipped with eight NVIDIA H100 GPUs, each boasting 80GB of high-speed HBM3 memory, this server delivers a unified 640GB GPU memory pool interconnected via NVLink. This architecture allows for massive model parallelism, making it the ultimate solution for training trillion-parameter Large Language Models (LLMs) and deploying real-time generative AI applications.

Orchestrating this massive compute array are two 5th Generation Intel Xeon Platinum 8568Y+ processors. These "Emerald Rapids" CPUs provide 48 cores and 96 threads each, running at 2.3GHz, to efficiently handle data preprocessing and pipeline management. To ensure the GPUs are never starved of data, the system is populated with 32x 64GB DDR5 5600MHz memory modules, creating a colossal 2TB high-bandwidth memory tier. This massive capacity is critical for keeping large datasets resident in memory, minimizing I/O latency during intensive training epochs.

Networking capabilities are scaled to match the compute density, featuring eight single-port NDR InfiniBand ConnectX-7 adapters. These cards provide 400Gb/s bandwidth per port, enabling a non-blocking 1:1 rail-optimized network fabric that is essential for multi-node clusters in High Performance Computing environments. Additionally, a 100GbE Mellanox ConnectX-6 card handles high-speed storage traffic, while a standard 1GbE LOM port manages basic connectivity. The storage layer includes four 3.84TB NVMe Gen4 SSDs for rapid data access and dual 960GB M.2 SSDs in RAID 1 for a resilient operating system boot.

Reliability and management are engineered for the data center. The server features the exclusive iDRAC9 Datacenter edition with 16GB of dedicated memory, offering advanced telemetry and thermal controls required for such high-wattage components. Power delivery is robust and redundant, utilizing six 2800W Power Supply Units (PSUs) in a 3+3 configuration. This setup supports GPU Power Brake features and ensures uninterrupted operation even during power grid fluctuations, making the Generative AI Infrastructure resilient and enterprise-ready.

Key Features

  • Exascale GPU Power: 8x NVIDIA H100 SXM5 80GB GPUs (HGX) for unparalleled AI training performance.
  • Elite Processing: Dual Intel Xeon Platinum 8568Y+ (48 Cores/96 Threads, 2.3GHz, 5th Gen) for data orchestration.
  • Massive Memory: 2048GB (2TB) of DDR5-5600MHz RAM maximizes throughput for large datasets.
  • Hyperscale Fabric: 8x NVIDIA ConnectX-7 NDR InfiniBand cards (400Gb/s each) for cluster interconnects.
  • High-Speed Uplink: 1x 100GbE Mellanox ConnectX-6 for storage networking + 1GbE LOM.
  • Fast Storage Tier: 4x 3.84TB NVMe Gen4 SSDs (Read Intensive) + 2x 960GB M.2 Boot (RAID 1).
  • Datacenter Management: iDRAC9 Datacenter Edition with 16GB RAM for deep system visibility.
  • Power Redundancy: 6x 2800W PSUs (3+3 Redundant) handle peak loads with ease.

Configuration

Component Specification
Model Platform Dell PowerEdge XE9680 (Determined by iDRAC & PSU Layout)
GPU Accelerator 8x NVIDIA HGX H100 SXM5 (80GB HBM3, 700W TDP)
Processor (CPU) 2x Intel Xeon Platinum 8568Y+ (48C/96T, 2.3GHz, 5th Gen)
Memory (RAM) 2048GB (2TB) Total - 32x 64GB DDR5 5600MHz
Boot Storage 2x 960GB M.2 SSD (RAID 1 Mirror)
Data Storage 4x 3.84TB NVMe Gen4 SSD (Read Intensive)
IB Network 8x Single-Port NDR InfiniBand ConnectX-7 (400Gb/s)
Ethernet Network 1x 100GbE Mellanox ConnectX-6 + 1x 1GbE LOM
Management iDRAC9 Datacenter (16GB Dedicated RAM)
Power Supply 6x 2800W PSU (3+3 Redundancy)

Compatibility

The Dell PowerEdge XE9680 is fully validated for the NVIDIA AI Enterprise software stack. It supports industry-standard operating systems such as Ubuntu Server 22.04 LTS and Red Hat Enterprise Linux 9, providing a stable foundation for containerized workloads via Kubernetes or Slurm. The system is optimized for NVIDIA's NVLink switch capabilities and requires compatible NVIDIA Quantum-2 InfiniBand switches to realize the full 400Gb/s potential of the installed ConnectX-7 cards.

Usage Scenarios

This configuration is the gold standard for Foundation Model Training. Companies developing proprietary Large Language Models (LLMs) rely on the 8-way NVLink interconnect to train models with billions of parameters across the GPU cluster efficiently. The massive VRAM ensures that larger batch sizes can be used, accelerating convergence times.

It is also essential for Digital Biology and Genomics. Researchers use the double-precision capabilities of the H100 to run simulations like AlphaFold2 for protein structure prediction, significantly speeding up the drug discovery pipeline compared to traditional CPU-based clusters.

For financial institutions, this server powers High-Frequency Algorithmic Trading backtesting. The combination of high-frequency Xeon processors and massive parallel GPU compute allows for the simulation of complex market scenarios and risk analysis in real-time.

Frequently Asked Questions

Q: Why does the iDRAC9 have 16GB of memory?
A: Standard servers use iDRACs with much less memory. The "Datacenter" edition on the XE9680 requires 16GB to process the massive stream of telemetry data from 8 GPUs, 6 PSUs, and high-speed fabrics. It enables granular thermal control and predictive maintenance analysis without burdening the main host CPUs.

Q: What is the benefit of the 3+3 Power Supply redundancy?
A: The server consumes significant power (up to 10kW+ under load). A 3+3 configuration splits the load across two distinct power grids (feeds). If one entire power grid fails (cutting power to 3 PSUs), the remaining 3 PSUs can instantly take over the full load, keeping the AI training job running without interruption.

Q: Do I need special cooling for this server?
A: The Dell PowerEdge XE9680 is an air-cooled server, meaning it fits in standard data center racks with proper hot/cold aisle containment. However, due to the high thermal output (approx. 35,000 BTU/hr), you must ensure your data center rack is rated for high-density airflow and power delivery.

PRODUKTY ZWIĄZANE Z TYM ARTYKUŁEM
Serwer Inspur NF5280M6 obsługujący technologię AI Dual Xeon z procesorem graficznym Tesla L2 do obciążeń korporacyjnych Polecane
Inspur NF5466M6 Dual Intel Xeon 4314 Enterprise Storage and Compute Server Polecane
Dell PowerEdge R760xs — konfiguracja korporacyjna z dwoma procesorami Xeon Silver 4410Y Polecane
Dell PowerEdge R760xs — wydajny serwer Xeon Gold 6507P Polecane
Serwer Dell PowerEdge R660 o wysokości 1U do montażu w szafie — dwa procesory Xeon Gold 6430, 1 TB pamięci RAM, karta HBA 25 GbE i Fibre Channel Polecane
Serwer HPE ProLiant DL380 Gen11 2U do montażu w szafie | Podwójny 48-rdzeniowy procesor Intel Xeon Gold 6542Y | 1 TB pamięci RAM DDR5 | 2x dysk twardy SAS 300 GB Polecane
Serwer pamięci masowej Inspur NF8480M5 4U dla przedsiębiorstw — 24× LFF SAS, platforma o dużej gęstości Quad Xeon Gold 6248R Polecane
Lenovo ThinkSystem SR850 V3 Wysokowydajny 4-procesorowy serwer z procesorem Xeon Gold 6448H i obsługą sieci korporacyjnej Polecane