Specifications
MPN
TCS-ASUS-ESCN8-E11
Model
ESC N8-E11 / ESC N8-E11V
Form Factor
7U rackmount
GPU
8x NVIDIA HGX H100 or H200 Tensor Core GPU
GPU Interconnect
NVLink + NVSwitch (HGX baseboard)
GPU-to-NIC Topology
Dedicated 1:1 GPU-to-NIC, up to 8 NICs
CPU
2x 5th Gen Intel Xeon Scalable (4th Gen on early SKUs)
Memory
32 DIMM slots
PCIe Slots
Up to 12
Storage
8x 2.5" NVMe + 2x 2.5" NVMe/SATA/SAS hot-swap
LAN
Dual 10Gb LAN
OCP
OCP 3.0 mezzanine (SKU2)
Power
3+3 3000 W redundant PSU
ASUS HGX 8-GPU servers
| Model | Height | GPU | CPU | GPU-to-NIC |
|---|---|---|---|---|
| ESC N8-E11 | 7U | 8x HGX H100 / H200 | Dual 5th Gen Xeon Scalable | 1:1, up to 8 NICs |
| ESC N8-E11V | 7U | 8x HGX H100 / H200 (V variant) | Dual 5th Gen Xeon Scalable | 1:1, up to 8 NICs |
| ESC A8A-E12U | 4U | 8x HGX H200 / B200 | Dual AMD EPYC 9005/9004 | 1:1 |
Use Case
The ESC N8-E11 is ASUS's first NVIDIA HGX architecture server and remains a common starting point for large-scale training clusters. Its distinguishing feature is the one-GPU-to-one-NIC topology: each of the eight H100 or H200 GPUs gets a dedicated network interface rather than sharing host PCIe lanes, so collective operations such as all-reduce run at the full line rate of the fabric. Combined with the HGX NVSwitch baseboard, a single 7U chassis is a self-contained eight-way training node; rack several and connect them through the 1:1 NICs and you have a linearly scaling cluster without adding an external PCIe switch fabric.
Need ASUS ESC N8-E11?
Enterprise hardware — configured, tested, deployed. Contact us for pricing and availability.
Request Quote