ESC8000-E12 4U ASUS Server 32x DDR5 DIMMs 8xDual Slot GPUs Custom

ASUS ESC8000-E12: 4U GPU Server with 8x Dual-Slot GPUs

The ASUS ESC8000-E12 is a high-density 4U GPU server powered by dual Intel Xeon 6 processors, purpose-built for AI training, AI inference, and HPC workloads. It directly supports up to eight dual-slot GPUs (up to 600W each) with flexible cooling and NVIDIA NVLink Bridge compatibility.

Key Specifications

• Processors: 2x Intel Xeon 6 (LGA 4710), up to 86 cores per CPU, 350W TDP

• Memory: 32x DDR5 DIMMs (16 per CPU, 8-channel) – up to 4TB at 6400MHz (1DPC) / 5200MHz (2DPC)

• GPU Support: 8x dual-slot GPUs – NVIDIA H200, RTX PRO 6000 Blackwell Server Edition, Intel Gaudi 3 PCIe

• GPU Power: Up to 600W per GPU

• GPU Slots: 8x PCIe Gen5 x16 (FHFL)

Storage & Expansion

• Drive Bays: 8x 2.5" hot-swap bays – up to 6x NVMe (SKU1) or 4x NVMe (SKU2)

• M.2: 2x Gen5 x4 (up to 22110) for boot/acceleration

• PCIe Expansion (SKU1): 8x GPU + 1x NIC/BlueField-3 (x16) + 1x NIC (x8) + 1x RAID/HBA (x8)

• PCIe Expansion (SKU2): 8x GPU + 1x NIC/BlueField-3 (x16) + 1x NIC (x8)

Networking & I/O

• Management: 1x dedicated RJ-45 management LAN

• Front I/O: Mini DisplayPort, 2x USB 5Gbps, Debug port

• Rear I/O: 1x USB 5Gbps, 1x management LAN

Management & Security

• BMC: ASMB12-iKVM (ASPEED AST2600) with KVM-over-IP

• Software: ASUS Control Center Enterprise, Redfish API

• Security: Hardware-level Root-of-Trust

Power & Cooling

• PSU: 3+1 redundant 3200W 80 PLUS Titanium (supports 8x 600W GPUs)

• Cooling: 5x GPU fans + 5x system fans (8080 tool-less hot-swappable)

Physical

• Dimensions: 439.5 x 800 x 175 mm (17.3" x 31.5" x 6.9") – 4U

• Net Weight: 42 kg

• Operating Temp: 10°C to 35°C

Why choose this server?

• Maximum GPU density: 8x dual-slot GPUs (600W each) in 4U – industry-leading for AI training

• Latest Intel Xeon 6 platform: Dual CPUs with up to 86 cores, 350W TDP, 8-channel DDR5-6400

• Massive memory capacity: 4TB DDR5 across 32 DIMM slots

• PCIe 5.0 architecture: 2x bandwidth over PCIe 4.0 for GPU, NIC, and storage

• Direct-CPU connectivity: GPU, NIC/DPU, and M.2 drives connect directly to CPU – minimal latency

• High GPU power support: 3+1 redundant 3200W Titanium PSUs

• Optimized thermal design: 10 hot-swappable 8080 fans for efficient cooling

• Tool-less maintenance: ASUS-exclusive design for streamlined servicing

• Broad GPU compatibility: NVIDIA H200, RTX PRO 6000, Intel Gaudi 3

• NVLink support: NVIDIA NVLink Bridge for GPU-to-GPU interconnect

Ideal for organizations seeking a powerful, scalable, and future-proof 4U GPU server platform for AI training, AI inference, HPC, and data analytics workloads in modern data center environments.