SADID GOSTARAN TECHNOLOGY SOLUTIONS WhatsApp : +98 912 427 1734
ASUS · NVIDIA GB300 · DESKSIDE AI SUPERCOMPUTER

ASUS ExpertCenter Pro ET900N G3 Price, Specifications & Availability

A procurement-focused guide to ASUS's deskside AI supercomputer, built on NVIDIA DGX Station architecture and powered by the GB300 Grace Blackwell Ultra Desktop Superchip. With 748GB of coherent memory and up to 20 PFLOPS of AI performance, ET900N G3 brings data-center-class compute to a deskside form factor.

748GBCoherent CPU–GPU memory
20 PFLOPSPeak AI performance
72-CoreNVIDIA Grace CPU
2 × 400GConnectX-8 QSFP112 ports
Pricing, exact configuration and lead time are confirmed individually against regional availability and requested options.
ASUS ExpertCenter Pro ET900N G3 NVIDIA GB300 deskside AI supercomputer
LOCAL AI DEVELOPMENTLarge-model development and inference at the deskside
COHERENT MEMORYCPU and GPU memory in one coherent address space
AI RESEARCHFine-tuning, simulation and deep-learning workflows
ENTERPRISE READYHigh-speed networking and managed procurement
01 / COMPUTE PLATFORM

NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip

ASUS ExpertCenter Pro ET900N G3 is built on NVIDIA DGX Station architecture and uses the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip. The platform combines a 72-core NVIDIA Grace CPU with a Blackwell Ultra GPU in a deskside system designed for demanding AI development and inference.

Rather than treating the CPU and GPU as isolated components, the platform uses NVIDIA NVLink-C2C to provide a high-bandwidth coherent connection between the processors. This architecture is designed to support large AI models and data-intensive workloads that benefit from a large shared memory space.

The product is an ASUS system based on NVIDIA DGX Station architecture. It should not be described as an NVIDIA-branded DGX Station product.
02 / MEMORY ARCHITECTURE

748GB of Coherent CPU–GPU Memory

The headline 748GB capacity is coherent CPU–GPU memory, not 748GB of GPU VRAM. ASUS lists the memory as 496GB of LPDDR5X system memory on the CPU side and 252GB of HBM3e GPU memory.

Total coherent memory748GB
CPU memory496GB LPDDR5X · 396GB/s
GPU memory252GB HBM3e · 7.1TB/s
CPU–GPU interconnectNVIDIA NVLink-C2C

This architecture gives developers a larger coherent memory pool for local AI workflows. The usable capacity and performance of a specific model still depend on model architecture, precision, software, context length and workload configuration.

03 / TECHNICAL DATA

ASUS ExpertCenter Pro ET900N G3 Technical Specifications

The table below summarizes the manufacturer's published platform specifications. Storage and optional components can differ by exact SKU.

ModelASUS ExpertCenter Pro ET900N G3
AI platformNVIDIA GB300 Grace Blackwell Ultra Desktop Superchip
CPUNVIDIA Grace · 72 Arm Neoverse V2 cores
GPUNVIDIA Blackwell Ultra
AI performanceUp to 20 PFLOPS (manufacturer-stated peak)
Coherent memory748GB total: 496GB LPDDR5X + 252GB HBM3e
CPU memory bandwidth396GB/s
GPU memory bandwidth7.1TB/s
CPU–GPU interconnectNVIDIA NVLink-C2C
Networking2 × NVIDIA ConnectX-8 SuperNIC QSFP112 ports (400Gb/s class), 1 × 10GbE, 1 × 1GbE dedicated BMC LAN
Operating system / softwareUbuntu with NVIDIA AI Developer Tools; confirm exact software image with the offered configuration
Warranty3-year on-site warranty is listed by ASUS; regional coverage must be confirmed
AI performance is a manufacturer-stated peak figure. Actual results depend on precision, sparsity, model, software stack and workload. Verify the final regional datasheet and offered SKU before purchase.
04 / HIGH-SPEED INTERCONNECT

NVLink-C2C and ConnectX-8 Networking

NVIDIA NVLink-C2C

Provides a high-bandwidth coherent connection between the Grace CPU and Blackwell Ultra GPU, helping the platform access its large coherent memory architecture.

NVIDIA ConnectX-8 SuperNIC

Two QSFP112 networking ports support high-speed connectivity for data-intensive workflows and integration into suitable AI infrastructure networks.

Management Networking

The system includes a dedicated 1GbE management LAN for BMC and a separate 10GbE network interface, according to ASUS specifications.

Remote Management

Dedicated BMC connectivity and management interfaces support system administration. Confirm the exact management workflow with the deployment configuration.

05 / STORAGE AND EXPANSION

NVMe Storage and PCIe Expansion

ASUS lists up to four M.2 2280 NVMe SSD positions. The platform specification identifies two PCIe 5.0 x4 M.2 positions pre-configured for OS RAID 1 and two additional PCIe 6.0 x4 M.2 positions intended for training data. The installed capacity depends on the ordered configuration.

Boot / OS storage2 × M.2 2280 PCIe 5.0 x4, pre-configured for OS RAID 1
Additional data storage2 × M.2 2280 PCIe 6.0 x4 positions
Maximum listed M.2 positions4
PCIe expansion3 physical PCIe x16 slots: one Gen 5.0 x16 and two Gen 5.0 x8 electrical configurations

Confirm SSD count, capacity and included accessories against the exact part number. ASUS lists multiple ET900N G3 configurations, including 2SSD and 4SSD variants.

06 / AI APPLICATIONS

What Can the ET900N G3 Be Used For?

Large Language Models

Local development, inference and fine-tuning workflows for large language models, subject to model size, precision, memory use and software support.

Generative AI

Development and deployment of generative AI applications, multimodal pipelines and advanced model workflows.

Agentic AI

Development and evaluation of autonomous AI agents and agent-based workflows using supported NVIDIA software.

Physical AI and Simulation

Research workflows involving robotics, simulation, scientific computing and accelerated deep learning.

Enterprise AI Development

A deskside environment for AI teams that need local compute resources while developing and validating models.

AI Inference

Run supported inference workloads locally and evaluate model behavior before deployment to larger infrastructure.

07 / ORDERABLE CONFIGURATIONS

ET900N G3 Models and Optional Components

ASUS publishes multiple model configurations. The exact suffix matters when comparing storage and included networking accessories.

ET900N G3-2SSDGB300 · 748GB coherent memory · 2 × 2TB NVMe SSD (default configuration listed by ASUS)
ET900N G3-4SSDGB300 · 748GB coherent memory · 4 × 2TB NVMe SSD
ET900N G3-2SSD-T2 × 2TB NVMe SSD · includes 2 GB300 QSFP transceivers
ET900N G3-4SSD-T4 × 2TB NVMe SSD · includes 2 GB300 QSFP transceivers
ET900N G3-4SSDT-Pro2K4 × 2TB NVMe SSD · 2 QSFP transceivers · NVIDIA RTX PRO 2000 listed as an additional GPU
This is a summary of ASUS-published configuration names, not a guarantee of regional availability. The included SSDs, transceivers, optional GPU and warranty coverage must be confirmed on the quotation.
08 / COMMERCIAL INFORMATION

ASUS ExpertCenter Pro ET900N G3 Price

The ASUS ExpertCenter Pro ET900N G3 is an enterprise-class AI supercomputer. Pricing can vary according to the selected model suffix, SSD configuration, included QSFP transceivers, optional GPU, regional availability, warranty coverage, delivery destination and order quantity.

For an accurate quotation, share the required configuration, quantity and destination country. Sadid Gostaran can coordinate a procurement enquiry and confirm current pricing and supply conditions.

09 / SUPPLY AND DELIVERY

Availability and Lead Time

ASUS states that ET900N G3 is available to order, while regional availability, exact configurations and service coverage vary. Current stock status and delivery lead time should therefore be confirmed at the time of quotation.

To receive a relevant estimate, provide the preferred model suffix, SSD configuration, optional accessories, quantity and destination. Any indicative schedule remains subject to written confirmation and current supply conditions.

10 / BUYER QUESTIONS

Frequently Asked Questions

What is ASUS ExpertCenter Pro ET900N G3?

It is a deskside AI supercomputer from ASUS built on NVIDIA DGX Station architecture and powered by the NVIDIA GB300 Grace Blackwell Ultra Desktop Superchip.

Does the ET900N G3 have 748GB of GPU VRAM?

No. The 748GB figure refers to coherent CPU–GPU memory: 496GB LPDDR5X CPU memory and 252GB HBM3e GPU memory.

What is the maximum AI performance?

ASUS advertises up to 20 PFLOPS of AI performance. Actual performance depends on precision, sparsity, software, model and workload.

Which CPU is used in the ET900N G3?

The platform uses an NVIDIA Grace CPU with 72 Arm Neoverse V2 cores.

Does every configuration include an additional RTX PRO GPU?

No. ASUS lists an optional/configuration-specific RTX PRO 2000 in the ET900N G3-4SSDT-Pro2K configuration. Do not assume it is included in every model.

How much storage can be installed?

ASUS lists up to four M.2 2280 NVMe SSD positions. The exact installed drives and capacity depend on the selected configuration.

Does it support high-speed networking?

ASUS lists two NVIDIA ConnectX-8 SuperNIC QSFP112 ports, along with 10GbE and dedicated 1GbE BMC networking.

How can I request the current price?

Contact Sadid Gostaran with your preferred configuration, quantity and destination to request a current quotation and lead-time estimate.

SADID GOSTARAN · AI INFRASTRUCTURE PROCUREMENT

Request a Quotation for ASUS ExpertCenter Pro ET900N G3

Contact our team for current pricing, configuration options, supply status, delivery lead time and procurement enquiries.

Chat with Sales on WhatsApp ↗

WhatsApp: +98 912 427 1734

Available in Iran

For local product information and purchasing options, visit the Persian ASUS ExpertCenter Pro ET900N G3 product page.

View Persian Product Page

Explore more: NVIDIA product directory · AI infrastructure

WhatsApp ↗ ↑