Neu

ASUS Ascent GX10

★★★★★ 5.0 (1)

The NVIDIA GB10 Grace Blackwell Superchip delivers petaflop-scale AI compute with 128GB of unified LPDDR5x memory, enabling large model inference in a compact, stackable metal chassis. Its 10GbE networking and Wi-Fi 7 connectivity, combined with a 2TB NVMe SSD, provide high-speed data throughput for secure, on-device agentic workflows. This workstation is best for AI developers running private, long-duration agentic tasks with frameworks like OpenClaw and NemoClaw.

CPU 20-Core: Up to 3.25 GHz Cortex A725 (10 Cores)Up to 3.8 GHz Cortex X925 (10 Cores)
RAM 128 GB
Storage 2 TB
GPU NVIDIA GB10 Grace Blackwell Superchip
form factor workstation
ASUS Ascent GX10 desktop
74 Gesamtbewertung
Auch erhältlich in:

Überblick

The 30-Second Version

The ASUS Ascent GX10 is a specialized AI beast with best-in-class CPU and memory for running massive models locally. It's dead silent, surprisingly compact, and absolutely useless for gaming. Worth it for AI devs who can find it near the $4,700 mark.

Pros & Cons

Vorteile

  • 128GB of unified memory lets you run enormous AI models locally. 100th
  • The GB10 Superchip delivers best-in-class AI compute in a compact chassis. 99th
  • Near-silent operation and low power draw for this level of performance. 92nd
  • Stackable design and 10GbE networking make it easy to cluster. 75th

Nachteile

  • Gaming performance is essentially nonexistent, don't even try.
  • Port selection is mediocre with only four USB-C and one HDMI 2.1.
  • Reliability scores are below average based on early data.
  • No pricing listed, but expect a hefty premium over a standard desktop.

Was Besitzer sagen

The Word on the Street

5.0/5 (1 reviews)
👍 Owners love the tiny footprint and quiet operation, saying it handles large models and contexts without breaking a sweat.
👍 Several buyers mention the low power consumption is a pleasant surprise for this level of AI performance.

Die Fakten

Performance

The CPU sits at the absolute top of our charts, and the 128GB of LPDDR5x RAM is best-in-class for this category. That unified memory is the real star here, letting you load massive 200B-parameter models that would choke most desktops. The 2TB NVMe SSD is fast and well above average. But the integrated GPU, while purpose-built for AI compute, lands in the bottom 10% for traditional graphics tasks. Don't expect to game on this thing at all, it scored a 13.4 out of 100 in our gaming tests. For AI inference and training, though, it's a petaflop-class monster in a shockingly small footprint.

Performance Percentiles

CPU 99.6
GPU 9.6
RAM 98.8
Anschlüsse 46.9
Speicher 91.6
Zuverlässigkeit 36.4
Nutzerresonanz 75.2

Specifications

Full Specifications

Processor

CPU 20-Core: Up to 3.25 GHz Cortex A725 (10 Cores)Up to 3.8 GHz Cortex X925 (10 Cores)
Cores 20
Frequency 3.3 GHz

Graphics

GPU NVIDIA GB10 Grace Blackwell Superchip
Type Integrated

Memory & Storage

RAM 128 GB
RAM Generation DDR5
Storage 2 TB
Storage Type NVMe SSD

Build

Form Factor workstation
Weight 3.2 kg / 7.0 lbs

Connectivity

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

vs Competition

Stacked against the Apple Mac Studio M4 Max, the GX10 trades blows depending on your workflow. The Mac has better GPU versatility, more ports, and a mature ecosystem, but it can't touch the GX10's 128GB of unified memory for massive model inference. The Lenovo Legion 34IAS10 and HP Omen 45L are gaming-first machines that completely outclass the GX10 in graphics but can't even load the models this thing handles natively. The MSI EdgeXpert and Dell Tower Plus are more traditional workstations with better reliability scores and port selection, but they lack the specialized AI silicon. Pick your poison: general purpose or AI-focused.

Spec ASUS Ascent GX10 Lenovo Legion 34IAS10 HP Omen 45L Apple Mac Studio M4 Max MSI EdgeXpert EdgeXpert-11SUS CLX SET TGMSETRTU5204BM
CPU 20-Core: Up to 3.25 GHz Cortex A725 (10 Cores)Up to 3.8 GHz Cortex X925 (10 Cores) Intel Core Ultra 9 Intel Core Ultra 9 285K Apple M4 Max NVIDIA GB Intel Core i9 14900KF
RAM (GB) 128 64 64 36 128 64
Storage (GB) 2048 3072 8096 512 4000 8000
GPU NVIDIA GB10 Grace Blackwell Superchip NVIDIA GeForce RTX 5080 NVIDIA GeForce RTX 5080 Apple M4 Max 32-core NVIDIA Blackwell Architecture NVIDIA GeForce RTX 5070
Form Factor workstation mid-tower mid-tower sff mini mid-tower
Psu W - 1200 - - 240 850
OS - Windows 11 Pro Windows 11 Home macOS NVIDIA DGX OS Windows 11 Home
Compare Compare Compare Compare Compare
Produkt CPUGPURAMAnschlüsseSpeicherZuverlässigkeitNutzerresonanz
ASUS Ascent GX10 99.69.698.846.991.636.475.2
Lenovo Legion 34IAS10 Compare 97.68896.791.796.569.784.8
HP Omen 45L Compare 97.68895.59899.469.787.2
Apple Mac Studio M4 Max Compare 85.565.569.694.530.299.499.9
MSI EdgeXpert EdgeXpert-11SUS Compare 99.795.298.887.197.936.484.5
CLX SET TGMSETRTU5204BM Compare 94.381.496.786.399.211.195.7

Preis

Value & Pricing

ASUS hasn't published an official price yet, but third-party listings show a wild spread from $4,699 to $6,579. At the low end, it's competitive with a loaded Mac Studio M4 Max for AI workloads, though the Mac is far more versatile. At the high end, you're paying a serious early-adopter tax. If you can snag it closer to $4,700, the petaflop-level AI performance in a quiet, power-sipping box is genuinely impressive. Just know you're buying a specialized tool, not a do-everything computer.

Ab 4.699 $ 2 Angebote bei 2 Händlern
B&H Photo 1 Angebot Ab 4.699 $
Amazon 1 Angebot Ab 4.699 $

Price History

4.640 $ 4.660 $ 4.680 $ 4.700 $ 4.720 $ 4.740 $ 4.760 $ 18. Juni19. Juni 4.699 $

Mehr erfahren

Overview

The ASUS Ascent GX10 is basically a baby NVIDIA DGX Spark in a tidy metal box. It packs the GB10 Grace Blackwell Superchip, which combines a 20-core ARM CPU and a Blackwell GPU on one package, with 128GB of unified memory. This isn't a gaming rig or a general purpose desktop. It's a focused AI workstation for developers who want to run big models locally without renting cloud GPUs.

Common Questions

Q: Can I use the ASUS Ascent GX10 for gaming or video editing?

No, the integrated GPU is optimized for AI compute and scores a 13.4 out of 100 in our gaming tests. It's not built for traditional graphics workloads.

Q: What size AI models can this run locally?

With 128GB of unified memory, it can handle models up to around 200 billion parameters, depending on quantization. That's far beyond what most consumer GPUs can manage.

Q: Does it support clustering multiple units together?

Yes, the stackable chassis and 10GbE networking via the ConnectX-7 SmartNIC are designed for linking multiple GX10s to scale up workloads.

Who Should Skip This

If you need a machine that can game, render 3D graphics, or run traditional creative apps, skip this entirely. The GPU is bottom-of-the-barrel for anything outside AI compute. You'd be better off with a Mac Studio or a high-end gaming desktop that can moonlight as an AI rig.

Verdict

This is for AI developers and researchers who need to run large language models, agentic workflows, or fine-tuning jobs locally and value a quiet, compact setup. If that's you, the GX10 is a standout. If you need a machine that can also game, edit video, or drive multiple high-res displays, look elsewhere. The single-minded focus is both its superpower and its limitation.

Usage Scores

Gesamt (73.7)AI/LLM (31.8)Gaming (13.4)Mobilität (52.6)Kreative (29)Business (72.4)Entwicklung (76.5)Home Office (82.1)Workstation (75.6)

Ähnliche Produkte