# Line below added 29 Nov 2024
AHGPI |ESC8000-E12P H200/RTX PRO 6000/4500 Blackwell

AHGPI - Intel® Xeon® 6 dual-processor - NVIDIA H200 / NVIDIA RTX PRO™ 6000/4500 Blackwell Server Edition / Intel® Gaud®3 PCIe AI Accelerator - 4U Form Factor Eight dual-slot GPUs in a high-dens

AHGPI

- Intel® Xeon® 6 dual-processor

- NVIDIA H200 / NVIDIA RTX PRO™ 6000/4500 Blackwell Server Edition / Intel® Gaud®3 PCIe AI Accelerator

- 4U Form Factor

Eight dual-slot GPUs in a high-density 4U compute server

Powered by Xeon 6 dual-processors, AHGP delivers exceptional performance, featuring top-tier GPUs, faster GPU interconnects, and higher-bandwidth fabric. This powerhouse server supports up to eight dual-slot GPUs, either active or passive, and offers flexible scalability with options for NVIDIA® NVLink® Bridge. This configuration allows seamless scaling and performance enhancement, making AHGPI an ideal choice for tackling the most Enterprise AI and HPC tasks.


Key Features

  • Intel Xeon 6 processors featuring 86 cores, eight-channel DDR5 memory up to 6400MHz and a maximum TDP of 350 watts per socket
  • Fully compatible with NVIDIA® MGX architecture for fast, scalable deployment
  • High-density 4U server for eight dual-slot high-end NVIDIA H200 GPUs, NVIDIA RTX PRO™ 6000 Blackwell Server Edition, NVIDIA RTX PRO™ 4500 Blackwell Server Edition, or Intel® Gaudi® 3 PCIe AI Accelerator, supporting up to 600 watts each
  • Optimized design with 5x PCIe® 5.0 slots for high-bandwidth NICs and DPUs
  • Exclusive tool-less design for easy maintenance and efficiency

Target market

• Virtual Machines
• Data Center
• HPC Application
• AI Training
• LLM Inference

Scale with ease and flexibility using MGX architecture

The AHGP advances AI infrastructure with NVIDIA® MGX™, an open, modular reference architecture designed for long-term scalability and future-ready compatibility. Supporting up to 160 configurable design options, it enables tailored deployments for AI and high-performance computing workloads. MGX redefines modular server design by improving ROI, accelerating deployment, and helping businesses adapt quickly to evolving AI demands.

System Layout


Front Panel


Rear Panel


AHG Block Diagram


Supercharge performance with PCIe 5.0 switch connectivity

This server integrates the latest Broadcom PEX89000 series switches, delivering up to 1,024Gbps (128GBps) of raw bandwidth per x16 port for high-availability hyperscale environments. Built on PCI Express® 5.0, it enables fast, efficient communication between in-rack hosts and endpoints while simplifying connectivity across servers, storage, and networking. Designed for demanding AI, ML, DL, and hyper-converged workloads, it provides industry-leading PCIe switching performance for modern data centres.

Processor / System Bus

2 x LGA 4710 sockets

Intel® Xeon® 6500/6700 processors

Up to 350 TDP


Memory

Total Slots:

32 (8 channel per CPU, 16 DIMM per CPU)

Capacity:

Maximum up to 4TB

Memory Type:

DDR5 6400/5600 RDIMM

Memory Size:

128GB, 96GB, 64GB, 32GB RDIMM


Expansion Slots


Rear:

8 x PCIe x16 for dual-slot GPU cards (Gen5 x16 link, FH/FL)

5 x PCIe x16 for NIC/BlueField-3 cards (Gen5 x16 link, FH/HL)

Front:

1 x PCIe x16 (Gen5 x8 link, FH/HL)


Storage

8 x 2.5" Front Hot-swap Storage Bays

(Backplane supports up to 8 x NVMe)

2 x M.2 socket (Gen 5 x4 link, up to 22110)


Networking

2 x 10GbE LAN ports (RJ45, X710-AT2)

1 x Management Port (RJ45)

Multi-GPU Support


Front I/O Ports

1 x Mini DisplayPort

2 x USB 5Gbps ports

1 x Debug port


Rear I/O Ports

1 x USB 5Gbps port

2 x RJ-45 LAN ports

1 x RJ-45 management LAN port


Switch/LED

Front:

1 x Power switch/LED

1 x Location switch/LED

1 x Message LED

1 x Clear CMOS switch

1 x Reset Switch

2 x M.2 LEDs

2 x LAN LEDs

1 x Q-code/Port 80 LED

Rear:

1 x Power switch/LED

1 x Location switch/LED

1 x Message LED


Dimensions

800mm x 439.5mm x 175mm

31.5” x 17.3” x 6.9”


Form Factor

4U


Weight

Net Weight: 42 Kg

Gross Weight: 44.23 Kg


Power Supply

3+1 redundant 3200W 80 PLUS Titanium power supply

220-240 Vac, 16A (x4) 50/60Hz


Environment

Operating temperature: 10°C ~ 35°C

Non-operating temperature: -40°C ~ 60°C

Non-operating humidity: 20% ~ 95% (Non-condensing)


*For operating temperatures above 30°C, please contact us

Datasheet 1
Title Version Date Size
AHGPI Brochure
CATALOGUE 1
Title Version Date Size
HS Product Catalog

AHGPI |ESC8000-E12P H200/RTX PRO 6000/4500 Blackwell

  • 4U dual-processor server powered by Intel® Xeon® 6 CPUs
  • Built on the NVIDIA® MGX server architecture
  • Supports up to 8 dual-slot GPUs
  • Compatible with NVIDIA® H200 GPUs
  • Compatible with NVIDIA RTX PRO™ 6000/4500 Blackwell Server Edition GPUs / Intel® Gaud®3 PCIe AI Accelerator
  • Up to 32 DIMM memory slots
  • 5x PCIe® 5.0 expansion slots
  • 8x 2.5" NVMe drive bays

  • $AUD $32,738.10

    *RRP Pricing*

    To View Channel Discounts Please Login


System Configurator


Tags: AHGPI, ESC8000-E12P, H200, RTX PRO 6000 Blackwell, Intel Xeon 6 CPU, PCIe 5.0, GPU, HPC Server, AI, AUS, Australia