Skip to the beginning of the images gallery

NVIDIA DGX Blackwell Spark

SKU: CE10915 Brand: Nvidia Guides (1)

Run complex AI workloads and private LLMs locally around the clock. Featuring a Grace Blackwell superchip and 128GB of unified memory, this NUC-sized compute node delivers high-performance at just 140W.

$8,995.00 AUD, inc GST

$8,177.27 AUD, exc GST

Available with a lead time
Expect dispatch between Jul 20 and Jul 21

A compact, workstation-class AI appliance designed to bring the power of enterprise processing directly to your desktop workspace. Featuring the Grace Blackwell GB10 Superchip and 128GB of coherent unified memory, this system acts as a complete, out-of-the-box solution for developing, testing, and deploying complex artificial intelligence workloads locally.

Makers, Professionals, and businesses relying on cloud infrastructure for Large Language Model (LLM) inference often face unacceptable latency, recurring subscription costs, and severe data privacy concerns. The DGX Spark solves this by offering local computational power capable of running monolithic models (up to 200 billion parameters) 24/7 without exposing sensitive data, all while drawing merely a fraction of the power required by traditional servers/compute hardware.

Join the rising tide of AI-integrated workflows; now is the perfect time to begin learning, and it's so easy to get started. If your homelab or business (from sole-trader, small-local business, to corporation) has any kind of data and a repeatable workflow, compute hardware like this makes sense in this day and age.

Features

Minimise operational costs and maximise power efficiency with a highly optimised thermal footprint, drawing an average of just 140W during typical computing loads, dropping to approximately 20W at idle. The watt-per-token is exceptional, and still at a good rate.
Deploy sophisticated, private LLMs continuously using the massive 128 GB of LPDDR5x coherent unified system memory, which can comfortably house multiple large-scale models locally for 24/7 access.
Eliminate complex setup times and software headaches thanks to the pre-installed NVIDIA DGX OS. Using official OEM hardware provides the peace of mind that your ecosystem is fully supported, allowing you to go from unboxing to running models in under an hour.
Expand your compute capabilities effortlessly by leveraging the integrated 200 Gbps ConnectX-7 interface to cluster multiple systems together, scaling up your hardware to handle even more demanding AI workflows.
Scale Effortlessly: Enterprise-level software and configuration toolchains that scale to any NVIDIA infrastructure demand you might have in the future.
Ensure ultra-fast data retrieval and secure local storage with the built-in 4 TB NVMe PCIe 4.0 SSD, providing ample, high-speed space for vast datasets, featuring self-encryption for absolute data security.

Technical Specifications

Property	Specification
Compute Architecture	NVIDIA GB10 Grace Blackwell Superchip
Processor (CPU)	20-core ARM architecture (10× Cortex-X925, 10× Cortex-A725)
Graphics (GPU)	Blackwell Architecture (5th Gen Tensor Cores, 4th Gen RT Cores)
System Memory	128 GB LPDDR5x coherent unified memory (Up to 273 GB/s bandwidth)
Local Storage	4 TB NVMe M.2 PCIe 4.0 SSD (Self-encrypting)
High-Speed Networking	1× NVIDIA ConnectX-7 NIC @ 200 Gbps
Connectivity	1× RJ-45 (10 GbE), Wi-Fi 7, Bluetooth 5.4, 4× USB Type-C, 1× HDMI 2.1a
Power Consumption	140 W (Typical usage), ~20 W (Idle)
Dimensions (L × W × H)	150 mm × 150 mm × 50.5 mm
Operating System	NVIDIA DGX OS (Pre-installed)

Resources

Power Efficient - Averaging around 140W during typical use, Approx 20W idle

Lots of memory - 128 GB LPDDR5x unified system memory

Ready to go - setup in less than an hour, no assembly or complex installation required

Business-oriented - $/W is optimised and designed to scale

Locally run LLMs 247 with ease - 128GB of memory allows multiple models to be loaded into memory around the clock

Full name: NVIDIA DGX Spark Founders Edition, ANZ, 4TB

Hardware Overview - https://docs.nvidia.com/dgx/dgx-spark/hardware.html

Datasheet: https://www.nvidia.com/en-au/products/workstations/dgx-spark/

This product is listed in:

Home Lab
Other Favourites>NVIDIA
Other Favourites>NVIDIA>Workstations

Product Comments

Start the conversation

Exact shipping can be calculated on the view cart page (no login required).

Products that weigh more than 0.5 KG may cost more than what's shown (for example, test equipment, machines, >500mL liquids, etc).

We deliver Australia-wide with these options (depends on the final destination - you can get a quote on the view cart page):

$3+ for Stamped Mail (typically 10+ business days, not tracked, only available on selected small items)
$7+ for Standard Post (typically 6+ business days, tracked)
$11+ for Express Post (typically 2+ business days, tracked)
Pickup - Free! Only available to customers who live in the Newcastle region (must order online and only pickup after we email to notify you the order is ready). Orders placed after 2PM may not be ready until the following business day.

Non-metro addresses in WA, NT, SA & TAS can take 2+ days in addition to the above information.

Some batteries (such as LiPo) can't be shipped by Air. During checkout, Express Post and International Methods will not be an option if you have that type of battery in your shopping cart.

International Orders - the following rates are for New Zealand and will vary for other countries:

$12+ for Pack and Track (3+ days, tracked)
$16+ for Express International (2-5 days, tracked)

If you order lots of gear, the postage amount will increase based on the weight of your order.

Our physical address (here's a PDF which includes other key business details):

40 Aruma Place
Cardiff
NSW, 2285
Australia

Take a look at our customer service page if you have other questions such as "do we do purchase orders" (yes!) or "are prices GST inclusive" (yes they are!). We're here to help - get in touch with us to talk shop.

Have a product question? We're here to help!