GS-SYS-06Engineering datasheet

NVIDIA DGX Spark

Statusshipping
Verified2026-07-10

DGX Spark is a desk-side box built around NVIDIA's GB10 Grace Blackwell superchip, 128 GB of unified memory, and the full CUDA stack, aimed at a single developer prototyping and fine-tuning models locally before moving to real GPU cluster capacity. It is not a cluster node: its 273 GB/s memory bandwidth is well behind a datacenter GPU, and while two units can be linked over ConnectX-7 to handle models up to 405B parameters, that is a two-box desk setup, not a scalable rack architecture. Buy it to develop against CUDA and iterate on model code without renting cloud GPU time; don't buy it expecting DGX B200-class throughput.

1 PFLOP FP4 (sparse)
AI performance · GB10 Grace Blackwell superchip
20-core Arm
CPU · 10x Cortex-X925 + 10x Cortex-A725
128 GB LPDDR5x unified
Memory · 273 GB/s bandwidth, 256-bit interface
1 TB or 4 TB NVMe
Storage · M.2 2242, self-encrypting
ConnectX-7, up to 200 Gbps
Networking · dual QSFP; links two units together
240 W external PSU
Power · GB10 SoC rated 140 W TDP
Compute01/05
SuperchipGB10 Grace Blackwell · Grace CPU + Blackwell GPU on one SoC
CPU cores20-core Arm · 10x Cortex-X925 (4 GHz) + 10x Cortex-A725 (2.8 GHz)
GPUBlackwell, 5th-gen Tensor Cores, 4th-gen RT Cores · 6,144 CUDA cores
FP4 performance1 PFLOP (sparse)
FP32 performance31 TFLOPS
Memory02/05
Capacity128 GB LPDDR5x · coherent unified CPU+GPU memory
Bandwidth273 GB/s · 256-bit interface, 16 channels
Inference ceilingup to 200B parameters · single unit
Fine-tuning ceilingup to 70B parameters · single unit, full-parameter
Clustered inference ceilingup to 405B parameters · two units linked over ConnectX-7
Connectivity03/05
Scale-out NICConnectX-7, 200 Gbps · 2x QSFP; PCIe lane sharing caps aggregate at 200 Gbps despite two ports
Ethernet1x RJ-45, 10 GbE · standard LAN
USB4x USB-C · one used for power delivery
Display1x HDMI 2.1a · plus up to 3x DisplayPort via USB-C alt mode
WirelessWi-Fi 7, Bluetooth 5.4
Power & physical04/05
Power supply240 W external · included
GB10 TDP140 W · of the 240 W budget
Dimensions150 x 150 x 50.5 mm · L x W x H
Weight1.2 kg
Operating temperature5C to 30C · non-condensing, 10-90% humidity
Software05/05
OSNVIDIA DGX OS · Ubuntu-based
StackNVIDIA AI Enterprise, NIM · preloaded
EcosystemFull CUDA access · distinguishes it from Apple Silicon or other ARM desktop AI boxes
Field notes
  • 273 GB/s memory bandwidth is the real ceiling, not the 128 GB capacity. Prefill-heavy and fine-tuning workloads are compute-bound and look great; decode-heavy inference is memory-bound and noticeably slower than the raw PFLOP figure suggests.
  • It is not a cluster building block. The dual QSFP ConnectX-7 ports look like 400 Gbps of fabric, but PCIe lane sharing caps actual throughput at 200 Gbps aggregate, and NVIDIA's own clustering story tops out at two linked units.
  • The 2242 M.2 storage form factor has been reported to lock up under sustained heavy I/O in third-party testing; treat local SSD as working storage, not the system of record for large datasets.
  • For raw token throughput on models that fit in 32 GB, a workstation RTX 5090 is faster and cheaper. DGX Spark's reason to exist is the 128 GB unified memory footprint for models a 5090 cannot hold, not raw speed.
  • Buy this for CUDA-native local development and fine-tuning iteration, not as a production inference box. Budget cloud GPU time for anything latency-sensitive or high-concurrency.
Questions we get on this part

How much does DGX Spark cost?

The 4 TB Founders Edition has listed between $3,999 and $4,699, reported. NVIDIA raised its own-store price to $4,699 in early 2026 citing memory supply constraints, while retailers like Micro Center have continued to sell at $3,999.99.

What is DGX Spark used for?

Local AI development: prototyping, full-CUDA-stack coding, and fine-tuning models up to roughly 70B parameters or running inference on models up to roughly 200B parameters, thanks to its 128 GB unified memory. It targets a single developer's desk, not production inference serving.

Can I cluster two DGX Spark units together?

Yes. Two units connected over their ConnectX-7 networking can handle inference on models up to roughly 405B parameters. This is a two-box desk setup for larger model capacity, not a scalable multi-node cluster architecture.

Is DGX Spark good for inference or training?

It handles both, but memory bandwidth (273 GB/s) is the bottleneck for decode-heavy inference. It is strongest for full-parameter fine-tuning and prefill-heavy workloads where compute, not memory bandwidth, is the limiting factor.