search
Cart shopping-cart

ASUS Ascent GX10 | Ultra‑Small Desktop AI Supercomputer

ASUS Ascent GX10

In Stock

* Unlock desktop-class AI with the ASUS Ascent GX10—a palm-size powerhouse built on the NVIDIA GB10 Grace Blackwell Superchip. Run LLMs, RAG, and vision AI faster than you thought possible with 128 GB unified memory, up to 4 TB NVMe, 10 GbE, and DGX Base OS readiness. Perfect for AI labs, edge pilots, and demo rigs, the GX10 delivers serious acceleration without a server room, noise, or hassle.


Why wait on cloud queues? Spin up experiments on your desk today and scale across multiple GX10s tomorrow.


Order in Australia & New Zealand with local warranty, expert integration, and rapid delivery.

 

Ready to move? Talk to a DiGiCOR Architect—get pricing, lead times, and a tailored config now.

  • Box PC
  • Single
  • ARM
  • DDR5
  • 1
  • 10Gb/s
  • 240
headset

Local Support in Australia and New Zealand

heroicons/outline/truck

Quick Delivery Wide stock availability to respond to most requirements.

Need help from a specialist?

Our specialist support is available!

heroicons/outline/phone-outgoing 1300 192 308 heroicons/outline/chat Talk to an expert

ASUS Ascent GX10 – Ultra-Small Desktop AI Supercomputer

Powered by the NVIDIA GB10 Grace Blackwell Superchip, the Ascent GX10 delivers petaflop-class AI acceleration in a 150 × 150 × 51 mm chassis—ideal for rapid model development and edge AI labs across Australia and New Zealand.

NVIDIA GB10 Grace Blackwell Superchip 20-core Arm v9.2-A CPU + Blackwell GPU Up to ~1 PFLOP (FP4) 128 GB LPDDR5X unified memory 3× USB-C (DP 2.1) + 1× USB-C PD-in HDMI 2.1, 10 GbE, ConnectX-7 NVIDIA Base OS (DGX Base OS) ready 1.48 kg | 150×150×51 mm

Compute architecture

  • NVIDIA GB10 Grace Blackwell Superchip (CPU+GPU SiP).
  • 20-core Arm v9.2-A CPU paired with next-gen Blackwell GPU.
  • AI performance up to ~1 petaFLOP (FP4) for accelerated inference.

Unified memory & storage

  • 128 GB LPDDR5X unified memory (shared CPU↔GPU pool).
  • Factory SSD options: 1 TB PCIe 4.0, 2 TB PCIe 4.0, or 4 TB PCIe 5.0.

I/O & connectivity

  • 3 × USB-C (USB 3.2 Gen 2x2, 20 Gbps) with DisplayPort 2.1 alt-mode.
  • 1 × USB-C PD-in (180 W EPR PD 3.1 spec) for power.
  • 1 × HDMI 2.1, 1 × 10 GbE LAN, Kensington lock.
  • NVIDIA ConnectX-7 SmartNIC for high-speed clustering/scale-out.

Software stack

  • Developer-ready NVIDIA Base OS (DGX Base OS) support.
  • CUDA-accelerated frameworks (e.g., PyTorch, TensorFlow).
  • Optimised for GB10 unified memory & Blackwell FP4/FP8 inference.

Form factor & thermals

  • Ultra-small: 150 × 150 × 51 mm; ~1.48 kg.
  • Quiet, efficient operation tailored for 24/7 desktop/lab use.
  • Fits desks, labs, and field enclosures without compromise.

Benefits for Australia & New Zealand deployments

  • Power-dense footprint: Rack-free AI acceleration ideal for labs, PoCs, and edge sites with space or power limits.
  • Unified memory agility: 128 GB LPDDR5X simplifies training/inference workflows and reduces CPU↔GPU copies.
  • Faster time-to-value: NVIDIA Base OS with mainstream frameworks shortens setup for data scientists and MLOps teams.
  • High-speed networking: 10 GbE and ConnectX-7 enable rapid dataset movement and multi-node scaling.
  • Quiet, reliable operation: Desktop-friendly acoustics suit offices, labs, and kiosks without additional cooling.
  • Portable cluster building: Link multiple GX10 units to scale experiments as requirements grow.

Applications aligned to the platform

LLM/RAG prototyping

Fine-tune and serve compact-to-mid models on unified memory with NVIDIA-validated tooling.

Vision AI & analytics

Accelerate detection, tracking, and QA using Blackwell-class inference (FP8/FP4).

Edge MLOps & demo rigs

Self-contained, quiet systems for client demos, pilots, and on-site experimentation.

Higher-ed & research

Lab-friendly AI compute for coursework, optimization research, and rapid iteration.

Ordering information

Model / SKU Short description Memory Storage Networking & I/O OS / Stack MPN
GX10-GG0032BN ASUS Ascent GX10 AI Supercomputer (GB10) 128 GB LPDDR5X unified 4 TB NVMe (PCIe 5.0) 3× USB-C (DP 2.1), 1× USB-C PD-in, HDMI 2.1, 10 GbE, NVIDIA ConnectX-7 NVIDIA Base OS (DGX Base OS) 90MS0371-M00100
GX10-GG0013BN ASUS Ascent GX10 AI Supercomputer (GB10) 128 GB LPDDR5X unified 1 TB NVMe (PCIe 4.0) 3× USB-C (DP 2.1), 1× USB-C PD-in, HDMI 2.1, 10 GbE, NVIDIA ConnectX-7 NVIDIA Base OS (DGX Base OS) 90MS0371-M000D0

Note: Storage options (1 TB / 2 TB / 4 TB) and regional bundles may vary. Confirm exact MPN and regional inclusions at order time.

Note: Specifications reflect ASUS’ published tech sheet for Ascent GX10. Features and software are subject to change with firmware and OS updates.

AI Model Sizes & Recommended Hardware

Hardware Setup GPU Memory Max Model Size (params)
Intel NUC 15 (RTX 5070) 32 GB Up to ~51 B
Workstation (RTX PRO 6000) 96 GB Up to ~153 B
ASUS Ascent GX10 100 GB Up to ~200 B
2× GX10 (clustered) 200 GB Up to ~405 B
4× Workstation (RTX PRO 6000) 384 GB Up to ~600 B

Recommended NVIDIA Stack by Model Size

Model Size (params) GPU Memory Recommended Hardware Scale When...
Up to 50 B 32 GB Intel NUC 15 (RTX 5070) Move to GX10 for >50 B
50–153 B 96 GB Workstation (RTX PRO 6000) Switch to GX10 at ~153 B
153–200 B 100 GB 1× ASUS Ascent GX10 Cluster 2× GX10 for >200 B
200–405 B 200 GB 2× GX10 (clustered) Consider cloud H100 for >400 B
405–600 B 384 GB 4× Workstation (RTX PRO 6000) Cloud H100/A100 for >600 B
Weight 7.000000
Brand Asus
Number of Nodes per System 1
Application AI/Deep learning Workstation, AI Training, Artificial Intelligence / Machine Learning Development, Scientific Research
Form Factor Box PC
Total Power (W) 240
Platform ARM
CPU Sockets Single
Co-processor/GPU NVIDIA Blackwell GPU
GPU 1
Drive Form Factor M.2
Drive Interface Technology M.2
Max Number of M.2 Drives 1
Memory Generation DDR5
Maximum RAM Capacity (GB) 128
Max Network Throughput (Gb/s) 10Gb/s
Number of Onboard Networking Ports 1
Write Your Own Review
Only registered users can write reviews. Please Sign in or create an account