NVIDIA DGX Spark Review

The NVIDIA DGX Spark packs specialized AI power into a small desktop, but its custom OS and average CPU make it a tool for a very specific job.

CPU NVIDIA GB10
RAM 128 GB
Storage 4 TB
GPU NVIDIA
Form Factor Desktop
Psu W 240
OS NVIDIA DGX
NVIDIA DGX Spark desktop
64.1 التقييم العام

The 30-Second Version

The NVIDIA DGX Spark is a specialized AI workstation in a tiny box. It has best-in-class RAM and storage for model work, but its CPU and GPU are average, and it runs a non-standard OS. Only worth it for developers who need a dedicated local AI machine. Skip it for anything else.

Overview

The NVIDIA DGX Spark Desktop is a weird one. It's a tiny desktop that packs a Grace Blackwell Superchip and a frankly absurd 128GB of RAM, all aimed at letting you run AI models locally. NVIDIA is calling it a 'personal AI computer,' and that's exactly what it is: a specialized tool, not a general-purpose PC.

You're getting a 4TB NVMe SSD and some seriously high-speed networking ports, like 200G QSFP112, which you won't find on any normal desktop. But it runs a custom NVIDIA DGX OS, not Windows or Linux you're used to. This thing has a very specific job description.

Performance

Let's be clear: for AI workloads, this is a beast. That 128GB of RAM is best-in-class, and the 4TB SSD is top of the charts. It's built to chew through training and inference tasks. The networking ports are a standout feature for moving huge datasets. But our data shows the GPU is only about average for this category, and the CPU scores as middle of the pack. For anything not AI-related, you're leaving a lot of that raw power on the table.

Performance Percentiles

CPU 41.5
GPU 58.7
RAM 99.1
Ports 91
Storage 98.7
Reliability 13
Social Proof 5.9

Pros & Cons

Pros

  • Massive 128GB of RAM is perfect for large AI models. 99th
  • Huge 4TB NVMe SSD offers tons of fast local storage. 99th
  • Extreme networking ports (200G QSFP112) for data transfer. 91th
  • Incredibly compact form factor for the hardware inside.

Cons

  • Runs a specialized DGX OS, not a standard desktop operating system. 6th
  • GPU performance is just average for the price. 13th
  • CPU performance falls behind many competitors.
  • Reliability scores in our database are a real letdown.

Specifications

Full Specifications

Processor

CPU NVIDIA GB10
Cores 8

Graphics

GPU NVIDIA
Type discrete
VRAM 8 GB
VRAM Type GDDR6

Memory & Storage

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

Build

Form Factor Desktop
PSU 240
Weight 1.2 kg / 2.6 lbs

Connectivity

USB Ports 4
HDMI 1x HDMI 2.1a Output
DisplayPort 1x DisplayPort
Wi-Fi WiFi 7
Bluetooth Bluetooth 5.4
Ethernet 1x Ethernet

System

OS NVIDIA DGX

Value & Pricing

With prices swinging wildly from $4,300 to nearly $7,500, the value proposition is all over the map. At the low end, if you're a developer who needs a dedicated local AI sandbox, it starts to make sense for the RAM and networking alone. At the high end, you're paying a massive premium for the NVIDIA branding and compact size. Shop very carefully, because that $3,200 price spread is huge.

Price History

‏٤٬٠٠٠ US$ ‏٤٬٥٠٠ US$ ‏٥٬٠٠٠ US$ ‏٥٬٥٠٠ US$ ‏٦٬٠٠٠ US$ ٩ مارس١٥ أبريل ‏٥٬٤٢٢ US$

vs Competition

Stack it up against a traditional workstation like the HP OMEN 45L or Dell Alienware Aurora, and you lose. Those machines have stronger overall CPUs and GPUs for gaming or general compute, and they run Windows. Compared to another AI-focused system like the MSI EdgeXpert, the DGX Spark wins on port density and compactness but likely loses on raw, balanced compute power and software flexibility. The Lenovo Legion Tower 5i offers better all-around performance for less money, but without the specialized AI chops.

Spec NVIDIA DGX Spark Dell Alienware Dell Alienware Aurora Gaming Desktop HP OMEN HP OMEN 45L Gaming Desktop, Intel Core Ultra 7 MSI EdgeXpert MSI EdgeXpert-11SUS AI Supercomputer Lenovo Legion Lenovo - Legion Tower 5i Gaming Desktop - Intel Acer Nitro Acer Nitro 60 Desktop Computer
CPU NVIDIA GB10 Intel Core Ultra 9 285K Intel Core Ultra 7 265K NVIDIA GB Intel Core Ultra 7 265F AMD Ryzen 9 7900
RAM (GB) 128 32 32 128 32 32
Storage (GB) 4096 2048 2048 4096 1000 2048
GPU NVIDIA NVIDIA GeForce RTX 5080 NVIDIA GeForce RTX 5080 NVIDIA NVIDIA GeForce RTX 5060 Ti NVIDIA GeForce RTX 5070 Ti
Form Factor Desktop Desktop Desktop Mini mid-tower Desktop
Psu W 240 1000 850 240 500 850
OS NVIDIA DGX Windows 11 Home Windows 11 Pro NVIDIA DGX OS Windows 11 Home Windows 11 Home
Compare Compare Compare Compare Compare
Product CpuGpuRamPortStorageReliabilitySocial Proof
NVIDIA DGX Spark 41.558.799.19198.7135.9
Dell Alienware Aurora Gaming Compare 97.887.986.399.493.171.993.8
HP OMEN 45L Gaming Compare 96.587.979.579.993.171.999.8
MSI EdgeXpert EdgeXpert-11SUS AI Supercomputer Compare 99.19599.1919841.285.9
Lenovo Legion Tower 5i Gaming Compare 87.574.688.599.459.471.999.8
Acer Nitro 60 Compare 86.884.779.576.993.136.187.1

Common Questions

Q: Can I install Windows or Linux on the DGX Spark?

It ships with the custom NVIDIA DGX OS. While it's likely based on Linux, it's a specialized environment tuned for AI workloads, and installing a standard desktop OS isn't supported or recommended.

Q: Is this good for gaming or video editing?

No. The GPU is average for its class, and the software stack isn't designed for it. You'd get much better performance from a similarly priced gaming PC or a Mac Studio for creative work.

Q: What are the 200G QSFP112 ports for?

They're for extremely high-speed network connections, like linking to a NAS or another server to transfer massive AI datasets and model files at speeds far beyond standard Ethernet.

Who Should Skip This

If you need a general-purpose desktop for gaming, everyday work, or content creation, look elsewhere. The software limitations and average consumer-grade performance make it a poor choice. Also, if reliability is a top concern, our data shows this category scores disappointingly low, so maybe consider a more established workstation brand.

Verdict

Buy this only if you are a developer or researcher who needs a dedicated, local appliance for AI model work and you value the small size and insane networking. It's a single-purpose tool. For literally any other use—gaming, office work, content creation—there are far better and cheaper options.