Price drop 10%

ASUS Ascent GX10-GG0016BN 2025, NVIDIA GB10 Grace Blackwell Superchip, 128GB, 4TB SSD, Grey

★★★★★ 5.0 (1)

The NVIDIA GB10 Grace Blackwell Superchip with 20 ARM cores and integrated Blackwell GPU delivers 128GB of unified LPDDR5x memory and a 4TB NVMe SSD in a 1.48kg mini chassis, making it a compact AI development platform. Its Ubuntu Linux environment with the NVIDIA AI stack, 10G LAN, and Wi-Fi 7 supports local model training and inference without cloud dependency. This is best for AI researchers and developers who need a portable, high-memory workstation for prototyping and running LLMs, not for gaming.

CPU NVIDIA GB10
RAM 128 GB
Storage 4 TB
GPU NVIDIA Grace Blackwell
Form Factor workstation
psu w 240
OS NVIDIA DGX
ASUS Ascent GX10-GG0016BN 2025, NVIDIA GB10 Grace Blackwell Superchip, 128GB, 4TB SSD, Grey desktop
36 Overall Score
Price CA$10,000
Also available in:

Snapshot

The 30-Second Version

The ASUS Ascent GX10 is a compact AI workstation with 128GB of unified memory and a 4TB SSD, built for developers who want to run large models locally. It's tiny, quiet, and comes with NVIDIA's AI stack preinstalled on Ubuntu. Gaming performance is terrible, and prices vary by thousands of dollars, so shop carefully.

Pros & Cons

Pros

  • 128GB unified memory is best-in-class for this form factor 99th
  • Tiny footprint and near-silent operation 98th
  • 4TB NVMe SSD gives you room for large datasets 84th
  • 10GbE, Wi-Fi 7, and ConnectX-7 SmartNIC built in 73rd
  • Runs Ubuntu with NVIDIA's AI stack preconfigured

Cons

  • Gaming performance is one of the worst we've seen
  • Reliability scores are mediocre at the 37th percentile
  • Price varies wildly between retailers
  • Only one customer review so far, so long-term data is thin
  • 240W PSU limits future expansion options

What owners think

The Word on the Street

5.0/5 (1 reviews)
👍 Buyers consistently praise the small footprint, low noise, and low power consumption while still handling large models and long context windows.
🤔 The single review available is enthusiastic, but there's not enough long-term data yet to judge reliability or software quirks.
👎 Some potential buyers express concern about the high price and the lack of upgrade options given the 240W power supply.

The proof

Performance

The GB10 Grace Blackwell Superchip is the star here. It combines a 20-core ARM v9.2-A CPU running at 4.0GHz with an integrated Blackwell GPU. The 128GB of unified memory is the headline spec, landing in the 99th percentile for RAM in our database. That's the kind of memory capacity you need for large model inference and fine-tuning. The 4TB NVMe SSD also sits in the 98th percentile, so you won't be juggling model files on external drives.

In practice, the GX10 handles large models and long context windows without breaking a sweat. The unified memory architecture means the CPU and GPU share the same pool, which eliminates the usual bottleneck of copying data between system RAM and VRAM. The GPU itself lands in the 84th percentile, which is strong but not top of the charts. The CPU is well above average at the 79th percentile. For AI development, that combination is more than enough for prototyping and running mid-size models locally. Just don't expect it to replace a multi-GPU server for training massive models from scratch.

Performance Percentiles

CPU 72.2
GPU 84.3
RAM 98.9
Ports 73.3
Storage 98.1
Reliability 37.3
Social Proof 70.4

Specifications

Full Specifications

Processor

CPU NVIDIA GB10
Cores 20
Frequency 3.3 GHz

Graphics

GPU NVIDIA Grace Blackwell
Type Discrete
VRAM 128 GB
VRAM Type LPDDR5X

Memory & Storage

RAM 128 GB
RAM Generation DDR5
Storage 4 TB
Storage Type NVMe SSD

Build

Form Factor workstation
PSU 240
Weight 1.5 kg / 3.3 lbs

Connectivity

USB-C Ports 4
USB Ports 4
HDMI 1x HDMI 2.1
Wi-Fi Wi-Fi 7
Bluetooth Bluetooth 5.4
Ethernet 10 GbE

System

OS NVIDIA DGX

vs Competition

The Apple Mac Studio M5 Max is the most direct competitor. It's similarly compact, runs quieter, and has a more mature software ecosystem for creative work. But it can't touch the GX10's 128GB of unified memory at this price point, and NVIDIA's CUDA support is still the default for most AI frameworks. If you're deep into PyTorch or TensorRT, the GX10 is the safer bet. The Lenovo Legion Tower 5a Gen 10 and HP OMEN TPC-W062-MT are gaming towers that happen to have decent GPUs. They'll crush the GX10 in gaming but fall apart when you try to load a 70B parameter model. The Dell ECT1250 and CLX SET TGMSETRTU5204BM are traditional desktops that don't really compete in the AI workstation space. They're cheaper, but they're solving a different problem.

Spec ASUS Ascent GX10-GG0016BN Apple Mac Studio M3 Ultra Lenovo T Series Towers Legion Tower 5a Gen 10 Dell Desktop ECT1250 HP OMEN TPC-W062-MT CLX SET TGMSETRTU5204BM
CPU NVIDIA GB10 Apple M3 Ultra AMD Ryzen 7 7800X3D Intel Core Ultra 7 265 Intel Core Ultra 9 285K Intel Core i9 14900KF
RAM (GB) 128 96 32 32 64 64
Storage (GB) 4096 1024 4096 2048 2048 8192
GPU NVIDIA Grace Blackwell Apple M3 Ultra 60-core NVIDIA GeForce RTX 5070 Ti Intel UHD Graphics NVIDIA GeForce RTX 5090 NVIDIA GeForce RTX 5070
Form Factor workstation sff mid-tower sff tower mid-tower
Psu W 240 - 850 180 - 850
OS NVIDIA DGX macOS Windows 11 Home Windows 11 Home Windows 11 Home Windows 11 Home
Compare Compare Compare Compare Compare
Product CPUGPURAMPortsStorageReliabilitySocial Proof
ASUS Ascent GX10-GG0016BN 72.284.398.973.398.137.370.4
Apple Mac Studio M3 Ultra Compare 99.776.195.793.653.699.386.2
Lenovo T Series Towers Legion Tower 5a Gen 10 Compare 73.593.27888.398.170100
Dell Desktop ECT1250 Compare 88.458.781.774.582.87099.1
HP OMEN TPC-W062-MT Compare 98.197.893.899.290.17067
CLX SET TGMSETRTU5204BM Compare 94.989.396.584.199.111.694.7

Price

Value & Pricing

Value is tricky here because the price spread is enormous. In Canada, you'll find this machine listed anywhere from CAD 6,532 to CAD 10,000. That's a CAD 3,468 difference for the same hardware, which is frankly ridiculous. Staples.ca has the best deal we've seen, so start there if you're buying in Canada. US pricing runs USD 7,999 to 8,468, and UK buyers are looking at GBP 10,000. At the low end of that range, the GX10 makes sense for developers who want a local AI rig without building a custom system. At the high end, you're paying a premium for the compact form factor and turnkey setup. The Apple Mac Studio M5 Max is the obvious alternative if you're not locked into NVIDIA's CUDA ecosystem, and it often undercuts the GX10 on price while offering better general-purpose performance.

From CA$10,000 2 offers across 2 retailers
Amazon 1 offer From CA$10,000
Staples 1 offer From CA$8,299

Price History

CA$6,000 CA$8,000 CA$10,000 CA$12,000 Jun 6Aug 15Sep 28Nov 16 CA$8,299

Read more

Overview

The ASUS Ascent GX10 is a mini desktop that's really an AI workstation in disguise. It's built around NVIDIA's GB10 Grace Blackwell Superchip, packs 128GB of unified LPDDR5x memory, and comes with a 4TB NVMe SSD. This isn't a machine for spreadsheets and email. It's for developers and researchers who want to run large language models locally without renting cloud GPUs by the hour. Pricing is all over the map right now, from CAD 6,532 to CAD 10,000 in Canada, so shopping around matters more than usual.

People searching for a compact AI development box will find the GX10 checks a lot of boxes. It runs Ubuntu out of the box, includes NVIDIA's AI software stack, and has connectivity that would make a network engineer smile: 10GbE, Wi-Fi 7, and an NVIDIA ConnectX-7 SmartNIC. The footprint is tiny at 5.91 inches square and 2.01 inches tall, and it weighs just 1.48kg. You could easily mistake it for a fancy router on your desk.

But this is a specialized tool. Our database puts it at 38.1 out of 100 for developer use and 39.9 for home office. Those scores reflect the narrow focus. If you're wondering whether the ASUS Ascent GX10 is good for gaming, the answer is a hard no. It scored 9.8 out of 100 there. This is a workstation for AI workloads, not a gaming PC.

Common Questions

Q: Is the ASUS Ascent GX10 good for gaming?

No. The integrated Blackwell GPU scored 9.8 out of 100 for gaming in our database, which is one of the worst results we've seen. This machine is built for AI workloads, not gaming.

Q: Can the ASUS Ascent GX10 run large language models?

Yes. The 128GB of unified LPDDR5x memory lets it handle large models and long context windows that would choke most desktops. It's designed specifically for local AI inference and fine-tuning.

Q: How does the ASUS Ascent GX10 compare to the Mac Studio M5 Max?

The GX10 has more unified memory at a similar price and native CUDA support, which matters for most AI frameworks. The Mac Studio is quieter, has better general-purpose performance, and a more polished software experience.

Q: Does the ASUS Ascent GX10 come with an operating system?

Yes, it ships with Ubuntu Linux and NVIDIA's AI software stack preinstalled, so you can start working with frameworks like PyTorch right away.

Who Should Skip This

Skip the GX10 if you're a gamer, a video editor, or anyone who needs a general-purpose desktop. The gaming performance is genuinely bad, and the ARM-based CPU means some x86 software won't run without emulation. If you're on a budget, the CAD 10,000 listings are hard to justify when the same hardware sells for CAD 6,532 at Staples.ca. And if you need to train massive models from scratch, a multi-GPU server or cloud cluster will serve you better. This is a prototyping and inference machine, not a data center replacement.

Verdict

Should you buy the ASUS Ascent GX10? If you're a developer or researcher who wants to run large language models locally, and you value a small, quiet machine over raw expandability, then yes. The 128GB of unified memory is the killer feature, and the 4TB SSD means you won't be juggling storage. The NVIDIA software stack is preconfigured, so you can be up and running with PyTorch in minutes.

But if you're looking for a general-purpose desktop, a gaming rig, or a budget AI setup, this isn't it. The gaming score of 9.8 out of 100 should tell you everything you need to know. And with only one customer review so far, we're taking the 5-star rating with a grain of salt. The reliability percentile of 37 is also a yellow flag. This is a specialized tool for a specific audience. If that's you, it's a compelling option. If not, keep looking.

Usage Scores

Overall (36.4)AI/LLM (42.3)Gaming (35.5)Portability (28.3)Creator (35.8)Business (33.5)Developer (37.4)Home Office (39.3)Workstation (37.5)
  • Overall36.4
  • AI/LLM42.3
  • Gaming35.5
  • Portability28.3
  • Creator35.8
  • Business33.5
  • Developer37.4
  • Home Office39.3
  • Workstation37.5

Other Configurations5

Similar Products