我們接受加密貨幣付款 — 讓您持有的幣種也能買到真正的硬體設備。以加密貨幣付款，直接購入真正的硬體設備。 全球訂單滿 US$399，即享免費保險運送。簡約外箱，不印任何品牌標誌，從德國出貨。訂單滿 US$399，即享免費保險運送。 結帳一開始，您的匯率即鎖定 30 分鐘。匯率鎖定 30 分鐘。 每件商品皆原廠密封、序號經過核對，並享有 24 個月保固。原廠密封，保固 24 個月。

追蹤清單 登入
Two compact silver cubes side by side with a connecting cable

PNY Technologies, Inc.

PNY Technologies, Inc. NVDGXSPARK2PK-KIT barebone mini PC 20-core ARM 128GB LPDDR5x 4TB NVME WiFi 7

SKU CH-6HWY-A1K54 MPN NVDGXSPARK2PK-KIT

Two NVIDIA DGX Spark units linked by a ConnectX-7 cable deliver 20-core ARM compute, 128GB LPDDR5x unified memory, 4TB NVME storage and WiFi 7 connectivity in a compact barebone kit

  • CPU Type20 core Arm, 10 Cortex-X925 + 10 Cortex-A725 Arm
  • CPU Cores20-Core
  • Memory installed128GB LPDDR5x, unified system memory
  • Storage Installed4TB NVME M.2 SSD with self-encryption
  • Discrete Graphic CardNVIDIA Blackwell Architecture
  • Max LAN Speed10Gbps

Option 2-Pack

  • 20-core ARM CPU with 10 Cortex-X925 and 10 Cortex-A725 cores
  • 128GB unified LPDDR5x memory delivering 273 GB/s bandwidth on 256-bit interface
  • 4TB NVME SSD storage with self-encryption for data protection
  • WiFi 7 and Bluetooth 5.3 with 10Gbps Ethernet for high-speed connectivity
  • Two DGX Spark units with ConnectX-7 Smart NIC and included connect cable

NVIDIA DGX Spark twin pack

This bundle contains two NVIDIA DGX Spark mini PCs built around the GB10 Grace Blackwell Superchip. Each unit integrates a 20-core Arm processor with ten Cortex-X925 and ten Cortex-A725 cores, 128 GB of unified LPDDR5x memory and a 4 TB self-encrypting NVME M.2 SSD. The pair is intended for developers and researchers who need desktop-scale AI compute without a full rack.

DGX Spark Connect Cable included

A single DGX Spark Connect Cable ships in the box to link the two systems directly, enabling multi-node workloads and cluster-style software stacks. The unified 256-bit memory interface delivers 273 GB/s bandwidth per unit, so the interconnect can move model shards and training data between nodes without saturating the local fabric. DGX OS is pre-loaded on both machines for immediate containerised deployment.

Verify power and network provisioning

Each Spark requires its own power supply, so confirm two suitable outlets or a PDU with adequate headroom before ordering. The rear panel provides one HDMI 2.1a port, four USB Type-C ports, a 10 Gbps RJ45 jack and a ConnectX-7 Smart NIC per unit; ensure your switch, cabling and any required VLANs support 10 GbE and the Smart NIC feature set. Wireless clients will need Wi-Fi 7 and Bluetooth 5.3 compatible infrastructure.

商品特色

  • 20-core ARM CPU with 10 Cortex-X925 and 10 Cortex-A725 cores
  • 128GB unified LPDDR5x memory delivering 273 GB/s bandwidth on 256-bit interface
  • 4TB NVME SSD storage with self-encryption for data protection
  • WiFi 7 and Bluetooth 5.3 with 10Gbps Ethernet for high-speed connectivity
  • Two DGX Spark units with ConnectX-7 Smart NIC and included connect cable

商品規格

BrandPNY
ModelNVDGXSPARK2PK-KIT
CPU Type20 core Arm, 10 Cortex-X925 + 10 Cortex-A725 Arm
CPU Cores20-Core
Memory installed128GB LPDDR5x, unified system memory
Storage Installed4TB NVME M.2 SSD with self-encryption
Discrete Graphic CardNVIDIA Blackwell Architecture
Max LAN Speed10Gbps
Wireless LAN802.11be
WiFi GenerationWi-Fi 7
BluetoothBluetooth 5.3
HDMI1 x HDMI 2.1a
Rear USB4 x USB Type C
RJ451
Other Rear Ports1 x ConnectX-7 Smart NIC
Dimensions1.99" x 5.91" x 5.91"
Weight2.65 lbs.
Packaging2 x NVIDIA DGX Spark
1 x Quick Start Card
1 x Support Card
1 x Power Supply
1 x DGX Spark Connect Cable
出貨重量7.42 kg
包裝尺寸541 × 292 × 178 mm

其他版本

同系列還有 1 款

同系列產品，記憶體、散熱器或時脈不同 — 下單前請先比較。

商品問答

Will the 128GB LPDDR5x unified memory ever need upgrading or can it be replaced?

The 128GB LPDDR5x is soldered as unified system memory on a 256-bit interface with 273 GB/s bandwidth, so it cannot be upgraded or replaced after purchase.

What limits sustained workloads first — the 20-core ARM CPU or the Blackwell GPU?

The 20-core ARM CPU with 10 Cortex-X925 and 10 Cortex-A725 cores shares the 128GB LPDDR5x pool with the Blackwell GPU, so memory bandwidth and capacity become the first bottleneck under continuous heavy loads.

How loud and hot does the unit get when running AI workloads all day?

The compact 1.99 x 5.91 x 5.91 inch chassis uses active cooling for the GB10 Superchip; expect noticeable fan noise and warm surface temperatures during sustained FP4 petaFLOP workloads.

同類商品

顧客也比較過

同一分類，同一套結帳流程 — 支援八種加密貨幣，匯率鎖定 30 分鐘。