🤖
Multi-GPU Ecosystem
Supports NVIDIA HGX H200 (141 GB), H100 (80 GB), H20 (96 GB), AMD Instinct MI300X (192 GB), and Intel Gaudi3 (128 GB) — choose the GPU best matched to your AI workload.
UK’s trusted IT infrastructure partner since 2003
8× GPUs per Node
H100 / H200 / MI300X / Gaudi3
1.12TB GPU Memory
8× H200 NVLink pool
900GB/s GPU Bandwidth
NVLink full-mesh
4TB System Memory
32 DIMMs DDR5 5600 MT/s
10× PCIe Gen5 Slots
Full-height half-length x16
6U Form Factor
Air-cooled — no liquid required
✅ Authorised Reseller
📦 Free UK Delivery
🔒 Full Warranty
📞 24/7 Support
Download Official Datasheet
PDF · Official manufacturer document
DellSKU: PE-XE9680
Dell PowerEdge XE9680 8-Way GPU AI Server
Eight GPUs. One Platform. Limitless AI.
Dell's flagship 6U, 8-way GPU server supporting NVIDIA H100/H200 SXM5, AMD MI300X, or Intel Gaudi3 — purpose-built for LLM training, inference, and HPC at enterprise scale.
- 8× NVIDIA H100 SXM5 (80 GB) or H200 SXM5 (141 GB) with NVLink full-mesh interconnect
- Dual Intel Xeon Scalable 5th Gen (up to 64 cores/CPU) — up to 4 TB DDR5 @ 5600 MT/s
- Up to 1.12 TB combined GPU memory (8× H200 141 GB)
- 10× PCIe Gen5 x16 slots for 400/800GbE InfiniBand/Ethernet NICs and DPUs
- Six 3200W Titanium PSUs in 3+3 fault-tolerant redundant configuration
- Air-cooled (10–35°C ambient); XE9680L variant offers direct liquid cooling
- iDRAC9 with GPU telemetry, CloudIQ AI predictive health, full Redfish API
Get Pricing — Speak to Our Team
Request a Quote
Competitive pricing · Response within 4 hours
Key Features
Everything you need in one device
🔗
NVLink Full-Mesh
All 8 NVIDIA GPUs interconnect via NVLink Switch with 900 GB/s bidirectional bandwidth — enables training of models too large for a single GPU with near-linear scaling.
🛡️
Fault-Tolerant Power
Six PSUs in 3+3 FTR arrangement: if two or more fail, GPU power-brake engages instead of shutting down — long training runs survive PSU failures without checkpointing loss.
📡
GPUDirect Storage
Direct path from NVMe drives and InfiniBand NICs to GPUs bypasses CPU — reduces data loading latency for large training datasets and high-throughput inference KV-caches.
📊
AI-Optimised Management
iDRAC9 provides real-time GPU utilisation, temperature, power consumption telemetry. CloudIQ integration uses ML to predict hardware failures before they disrupt training.
🌡️
Robust Air Cooling
16 gold-grade fans moving 1,200 CFM support 10–35°C ambient — handles NVIDIA H100/H200, AMD MI300X, and Intel Gaudi3 at full rated TDP without liquid cooling infrastructure.
About the PowerEdge XE9680
The Dell PowerEdge XE9680 is Dell's flagship enterprise AI server — a 6U, 8-way GPU platform designed for the most demanding generative AI training, large-scale inference, and HPC simulation workloads. Combining a PowerEdge R760-style compute node with a massive 4U GPU drawer, the XE9680 delivers exceptional GPU density and flexibility within a well-engineered, air-cooled chassis.
At the heart of the XE9680 are eight high-performance GPU accelerators connected via NVLink full-mesh at 900 GB/s. GPU options span NVIDIA HGX H200 (141 GB), H100 (80 GB), and H20 (96 GB); AMD Instinct MI300X (192 GB); and Intel Gaudi3 (128 GB). This multi-vendor GPU support makes the XE9680 genuinely flexible for organisations evaluating different AI accelerator ecosystems. All eight NVIDIA GPUs can train models collectively with 1.12 TB of shared NVLink-addressable GPU memory, enabling LLMs with 70B–400B+ parameters to be trained on a single node.
System compute is provided by dual Intel Xeon Scalable processors — 5th Generation with up to 64 cores per CPU at 5600 MT/s DDR5, or 4th Generation with up to 56 cores at 4800 MT/s. With 32 DIMM slots supporting up to 4 TB of DDR5 memory, the XE9680 can hold the large datasets and system-side KV-caches demanded by modern inference workloads. Ten PCIe Gen5 x16 slots provide connectivity for 400GbE/800GbE InfiniBand adapters, GPUDirect-capable NICs, and DPUs essential for multi-node AI cluster networking.
Power resilience is a key differentiator. The XE9680's six 3200W/2800W Titanium PSUs operate in a 3+3 fault-tolerant redundant (FTR) configuration. Unlike systems that simply shut down on PSU failure, the XE9680 throttles GPU clocks to approximately 20% of typical performance — allowing long-running training jobs to survive PSU incidents without checkpoint loss, then return to full speed at the next maintenance window.
Technical Specifications
📄 Download Datasheet (PDF)Use Cases
Ideal for a wide range of deployments
🧬
LLM Training
1.12 TB NVLink-addressable GPU memory and 900 GB/s interconnect enable single-node training of 70B–400B+ parameter models — ideal for fine-tuning Llama 3, Mistral, and custom foundation models.
⚡
High-Throughput Inference
8× H200 GPUs serve large language models at scale — XE9680 clusters achieve 117,248 TOPS across 8 nodes (64 GPUs) with 93.2% linear scaling efficiency.
🔬
HPC & Simulation
Dual Xeon with up to 4 TB DDR5 and AMD MI300X support supports molecular dynamics, genome sequencing, climate modelling, and CFD simulations alongside AI workloads.
🏗️
AI Infrastructure
10× PCIe Gen5 slots accommodate 400/800GbE InfiniBand NICs and DPUs — XE9680 nodes interconnect into SuperCluster AI factories with multi-node all-reduce at wire speed.
FAQ
Frequently Asked Questions
Why Servnet
Why buy from Servnet?
Trusted UK IT reseller since 2003 — supplying businesses of all sizes with genuine, competitively priced technology.
🏆
Authorised UK Reseller
Servnet is an authorised IT reseller with direct access to leading product lines at competitive pricing.
💬
Expert Pre-Sales Advice
Our engineers will help you select the right product and advise on compatibility with your existing infrastructure.
🚚
Fast UK Delivery
Most products ship from UK stock. Express delivery available for urgent deployments.
🔒
Genuine, Warranted Kit
All products are genuine, brand-new, with full manufacturer warranty and RMA support.
📋
Configuration Services
We can pre-configure your hardware before shipping — reducing on-site deployment time and cost.
📞
24/7 Support
Post-sale support from our certified engineers. We're here if anything goes wrong.
