📦
Extreme Rack Density
2U form factor with 4 SXM GPUs and direct liquid cooling delivers up to twice the GPU core density per rack height compared to 6U air-cooled alternatives.
UK’s trusted IT infrastructure partner since 2003
4× H100/H200 GPUs
SXM5 NVLink full-mesh
564GB GPU Memory
4× H200 141 GB HBM3e
2U Form Factor
Liquid-cooled — dense rack
2× Rack Density
vs 6U air-cooled alternatives
4TB System Memory
32 DIMMs DDR5 5600 MT/s
4× PCIe Gen5 Slots
InfiniBand NIC + DPU support
✅ Authorised Reseller
📦 Free UK Delivery
🔒 Full Warranty
📞 24/7 Support
Download Official Datasheet
PDF · Official manufacturer document
DellSKU: PE-XE9640
Dell PowerEdge XE9640 4-GPU Dense AI Server
Maximum GPU Density. Minimum Rack Space.
Dell's densest 2U GPU server — 4× NVIDIA H100/H200 SXM5 GPUs with direct liquid cooling. Delivers up to 2× GPU core density per rack versus standard air-cooled configurations.
- 4× NVIDIA H100 SXM5 (80 GB) or H200 SXM5 (141 GB) with full NVLink interconnect
- 2U liquid-cooled chassis — up to 2× more GPU core density per rack than air-cooled
- Dual Intel Xeon Scalable 5th Gen (up to 64 cores) — up to 4 TB DDR5 @ 5600 MT/s
- Direct liquid cooling for CPUs and GPUs — 45% energy saving over air cooling
- 4× PCIe Gen5 slots for InfiniBand networking and DPU connectivity
- Purpose-built for inference and distributed training requiring maximum rack GPU density
Get Pricing — Speak to Our Team
Request a Quote
Competitive pricing · Response within 4 hours
Key Features
Everything you need in one device
❄️
Liquid-Cooled GPUs
Direct liquid cooling removes heat from CPUs and all four H100/H200 SXM5 GPUs at the source — enabling sustained 700W SXM5 TDP in a 2U chassis without thermal throttling.
🔗
NVLink Full-Mesh
All four GPUs are fully NVLink connected — enabling 4-GPU all-reduce operations for smaller model training and fine-tuning with near-linear scaling within the node.
⚡
Inference Scale-Out
2U form factor makes XE9640 ideal for scale-out inference clusters — pack more GPU nodes per rack for higher token throughput per rack unit than any air-cooled GPU server.
🤝
Pairs with XE9680
XE9640 and XE9680 can be mixed in the same InfiniBand cluster — use XE9640 for dense inference and XE9680 for large training runs within one AI infrastructure.
📊
Dell Validated Design
Ships with Dell AI Reference Architectures — pre-validated designs for PyTorch distributed training, NVIDIA TensorRT-LLM inference, and popular frameworks reduce time-to-production.
About the PowerEdge XE9640
The Dell PowerEdge XE9640 is a 2U, 4-way GPU server engineered for maximum GPU density in space and power-constrained data centres. Unlike the 6U XE9680, the XE9640 uses direct liquid cooling to pack four NVIDIA H100 or H200 SXM5 GPUs into a 2U chassis — delivering up to twice the GPU core density per rack unit.
Each XE9640 houses four NVIDIA H100 (80 GB) or H200 (141 GB) SXM5 GPUs with full NVLink interconnect at 900 GB/s. The liquid cooling system removes heat directly from GPUs and CPUs, enabling sustained 700W SXM5 GPU TDP without throttling — something air-cooled 2U chassis cannot achieve. Dual Intel Xeon Scalable processors provide up to 4 TB of DDR5 system memory for hosting large inference KV-caches alongside GPU workloads.
The XE9640 is particularly well-suited for inference-at-scale deployments, where the goal is maximising total GPU count per rack for token throughput. Four PCIe Gen5 x16 slots provide connectivity for 400GbE InfiniBand adapters and BlueField-3 DPUs, connecting XE9640 nodes into high-bandwidth all-reduce clusters for distributed inference and fine-tuning. XE9640 and XE9680 nodes can be mixed in the same cluster, offering a flexible approach to combining dense inference nodes with larger 8-GPU training nodes.
Technical Specifications
📄 Download Datasheet (PDF)Use Cases
Ideal for a wide range of deployments
⚡
Scale-Out Inference
Pack more GPU nodes per rack for maximum tokens/second throughput — XE9640's 2U density means double the inference capacity of 6U alternatives in the same rack space.
🧬
Fine-Tuning at Scale
4× NVLink H100/H200 GPUs enable fine-tuning of 7B–70B parameter models on a single 2U node — deploy many XE9640s across a rack for parallel fine-tuning pipelines.
🏗️
Dense AI Clusters
Pairs with XE9680 in shared InfiniBand clusters — XE9640 for dense inference racks, XE9680 for large training runs, sharing storage and networking infrastructure.
🔬
HPC Modelling
SXM5 NVLink GPUs and 4 TB DDR5 support molecular dynamics, genome sequencing, and materials science computations in a highly space-efficient chassis.
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.
