Specifications
MPN
TCS-DELL-XE9680
Model
PowerEdge XE9680
Form Factor
6U rackmount
GPU
8x NVIDIA HGX H100 80GB / H200 141GB / H20 96GB 700W SXM5
GPU (alt 1)
8x AMD Instinct MI300X 192GB 750W OAM (Infinity Fabric)
GPU (alt 2)
8x Intel Gaudi 3 128GB
GPU Interconnect
NVIDIA NVLink + NVSwitch, 900 GB/s per GPU (H200 config)
CPU
2x 5th Gen Intel Xeon Scalable, up to 64 cores each
Memory
32x DDR5 RDIMM (registered ECC only), up to 4 TB
Expansion
10x PCIe Gen5 x16 FHHL slots
Storage
8x 2.5" NVMe U.2 front bays
Management
iDRAC9 with Redfish / IPMI 2.0
Cooling
Air cooled (XE9680L adds direct liquid cooling)
System Power
Up to 8 kW across the SXM5 GPU complex
PowerEdge XE9680 accelerator configurations
| Accelerator option | Memory | TDP | Interconnect | Modules per system |
|---|---|---|---|---|
| NVIDIA HGX H100 SXM5 | 80 GB HBM3 | 700 W | NVLink 4.0 + NVSwitch | 8 |
| NVIDIA HGX H200 SXM5 | 141 GB HBM3e | 700 W | NVLink 4.0 + NVSwitch | 8 |
| NVIDIA HGX H20 SXM5 | 96 GB HBM3 | 500 W | NVLink 4.0 + NVSwitch | 8 |
| AMD Instinct MI300X OAM | 192 GB HBM3 | 750 W | AMD Infinity Fabric | 8 |
| Intel Gaudi 3 | 128 GB HBM2e | — | RoCE / 24x 200GbE | 8 |
Use Case
The XE9680 is the reference 6U eight-GPU platform for large-scale generative AI training and HPC in a standard 19-inch rack. Its value is the same chassis accepting NVIDIA HGX H100/H200 SXM5, AMD Instinct MI300X OAM or Intel Gaudi 3, so a rack built for H100 can be re-populated without changing the rack, power or cooling design. With 8x 141 GB H200 the system presents 1.1 TB of aggregate HBM3e and 38.4 TB/s of aggregate bandwidth inside one NVSwitch domain, which is the configuration most commonly quoted for frontier-model fine-tuning and long-context inference. 32 DDR5 RDIMM slots carry the host side up to 4 TB, and 10 PCIe Gen5 x16 slots absorb 400G/800G NICs, NVMe HBAs or a storage front end.
Need Dell PowerEdge XE9680?
Enterprise hardware — configured, tested, deployed. Contact us for pricing and availability.
Request Quote