Asus Ascent GX10 AI Supercomputer

Brand: ASUS

ServersComplete Servers

  • Delivered nationwideDoor-to-door, anywhere in South Africa
  • 1 YearLocal support
  • 7-day returnsChange your mind within 7 days of delivery. Returns policy
  • Secure checkoutPaid on an encrypted payment page

Description

Petaflop AI performance on your desk | 128GB coherent unified memory | Ultra-compact 150 x 150 x 51 mm form factor | NVIDIA ConnectX-7 scalable networking | Preloaded NVIDIA AI software stack

Quick Specs

  • Achieve petaflop-scale AI performance with the NVIDIA GB10 Grace Blackwell Superchip delivering 1 PFLOP of theoretical FP4 tensor performance
  • Experience seamless large-model work backed by 128 GB of LPDDR5x coherent unified system memory — enough to fine-tune models up to 200B parameters
  • Scale your AI infrastructure by linking up to multiple Ascent GX10 systems via the integrated NVIDIA ConnectX-7 NIC at 200 Gbps
  • Develop and deploy immediately with the preloaded NVIDIA AI Software Stack and NVIDIA DGX OS, ready from first boot
  • Fit petaflop-class compute anywhere — the Ascent GX10 measures just 150 x 150 x 51 mm and weighs 1.48 kg

Supercomputer-class AI, sized for your desk

The ASUS Ascent GX10 is a compact desktop AI supercomputer built for AI researchers, developers, data scientists, and machine learning engineers who demand serious compute without a dedicated server room. Powered by the NVIDIA GB10 Grace Blackwell Superchip — the same architecture found in NVIDIA DGX Spark — it delivers up to 1 PFLOP of theoretical FP4 AI performance and 128 GB of LPDDR5x coherent unified system memory in a chassis that measures just 150 x 150 x 51 mm and weighs 1.48 kg. This is petaflop-scale AI development that fits on a standard office desk.

The horsepower to handle your most demanding AI workloads

Tired of offloading every serious workload to the cloud because your local hardware simply cannot cope? The Ascent GX10 changes that equation. Its NVIDIA Blackwell GPU integrates fifth-generation Tensor Cores and fourth-generation RT Cores, while the 20-core Arm CPU — combining 10 Cortex-X925 and 10 Cortex-A725 cores — handles data preprocessing and model orchestration without breaking stride. With 128 GB of unified memory and up to 273 GB/s of memory bandwidth, the system can prototype, fine-tune, and run inference on large language models with hundreds of billions of parameters directly on-premises.

Scale beyond a single unit with NVIDIA ConnectX-7 networking:

When a single unit is not enough, the integrated NVIDIA ConnectX-7 NIC at 200 Gbps enables you to link multiple Ascent GX10 systems together, scaling your AI workloads to handle even larger models. NVLink-C2C technology delivers a cohesive CPU and GPU memory model with substantially greater bandwidth than PCIe Gen 5, ensuring low-latency, high-throughput communication across distributed workloads. A dedicated 10 GbE RJ-45 port, Wi-Fi 7 2×2, and Bluetooth 5.4 round out the connectivity picture for everyday lab and office use.

Ready to develop from the moment it powers on

The Ascent GX10 ships with NVIDIA DGX OS and the full NVIDIA AI Software Stack preloaded, so you spend your time building models rather than configuring environments. Security-conscious deployments benefit from OpenShell secured containers and NemoClaw sandboxes, which provide private, on-device inference and governed access control for long-running autonomous agents. The engineered thermal design sustains peak performance across labs, offices, and edge environments alike — high-density compute that runs reliably wherever you need it.

Features

  • NVIDIA GB10 Grace Blackwell Superchip: 20-core Arm CPU (10 Cortex-X925 + 10 Cortex-A725) paired with NVIDIA Blackwell GPU
  • 1 PFLOP of theoretical FP4 AI tensor performance via fifth-generation Tensor Cores
  • 128GB LPDDR5x coherent unified system memory with 256-bit memory interface and up to 273 GB/s bandwidth
  • 1TB M.2 2242 NVMe PCIe® 4.0 x 4 SSD
  • Integrated NVIDIA ConnectX-7 NIC at 200 Gbps for multi-system scaling and distributed AI workloads
  • Wi-Fi 7 2×2, Bluetooth 5.4 with LE, and 10 GbE RJ-45 for versatile connectivity
  • Four USB 3.2 Gen 2×2 Type-C ports (3 with DisplayPort 2.1 alternate mode; 1 with 180 W EPR PD3.1 power delivery input) plus HDMI 2.1a
  • Preloaded NVIDIA DGX OS and NVIDIA AI Software Stack for immediate development and deployment
  • Engineered thermal design sustaining performance in office, lab, and edge environments
  • Kensington Lock slot for physical security; certified to BSMI/CB/CE/FCC/UL/CCC/C-Tick/Wi-Fi/RF/VCCI standards

Product information

Brand
ASUS
SKU
90MS0371-M00030
Warranty
1 Year
Architecture
NVIDIA Grace Blackwell
GPU
NVIDIA Blackwell Architecture
CPU
20 core Arm, 10 Cortex-X925 + 10 Cortex-A725 Arm
NVIDIA® CUDA® Cores
NVIDIA Blackwell Generation
Tensor Cores
5th Generation
RT Cores
4th Generation
Tensor Performance
1 PFLOP
System Memory
128 GB LPDDR5x, coherent unified system memory
Memory Interface
256-bit
Memory Bandwidth
Up to 273 GB/s
Storage
1TB M.2 2242 NVMe PCIe® 4.0 x 4 SSD
USB
3 x USB 3.2 Gen 2×2 Type-C, 20Gbps, alternate mode (DisplayPort)
LAN
1 x RJ-45 connector 10 GbE
NIC
1 x NVIDIA® ConnectX-7 NIC @ 200 Gbps
Wi-Fi
Wi-Fi 7 2×2
Bluetooth
Bluetooth® 5.4 w/LE
Power Supply
240W
Display Connectors
1 x HDMI 2.1a
NVENC | NVDEC
1x | 1x
OS
NVIDIA DGXOS
Software
NVIDIA AI Software Stack
Kensington Lock
Yes
Colour
Stellar Grey
Dimension
15 x 15 x 5.1 cm
Weight
1480g