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
Compact desktop AI system with matte black finish and subtle green accent lighting

NVIDIA

NVIDIA 940-54242-0000-000 DGX Spark AI edge system 20-core Arm 128 GB LPDDR5x 4 TB NVME

SKU CH-HZJF-QX8JW MPN 940-54242-0000-000

A compact AI workstation with a 20-core Arm processor, 128 GB LPDDR5x unified memory, 4 TB NVME storage, and Wi-Fi 7 connectivity

  • TypeGaming & Entertainment
  • UsageConsumer
  • Processor20 core Arm, 10 Cortex-X925 + 10 Cortex-A725 Arm
  • Processor Main Features64 bit 20-Core Processor
  • Memory128 GB LPDDR5x, unified system memory
  • StorageUp to 4 TB NVME.M2 with self-encryption

Type NVIDIA DGX Spark

  • GMKtec Ryzen Al Max+ 395
  • NVIDIA DGX Spark

被劃上刪除線的選項確實存在，但本站並無供貨。

  • Up to 1 petaFLOP of AI compute at FP4 precision with Blackwell architecture
  • 128 GB LPDDR5x unified memory runs models up to 200 billion parameters locally
  • 20-core Arm CPU (10 Cortex-X925 + 10 Cortex-A725) delivers 64-bit high-performance processing
  • 4 TB NVME self-encrypting storage with 10 Gbps LAN and Wi-Fi 7 connectivity
  • Pre-installed NVIDIA DGX OS and AI stack enables immediate development, fine-tuning and inference

DGX Spark personal AI computer

The NVIDIA DGX Spark is a desktop AI workstation built on the GB10 Grace Blackwell Superchip. It combines a 20-core Arm processor with 10 Cortex-X925 and 10 Cortex-A725 cores alongside a Blackwell GPU. The system ships with 128 GB of unified LPDDR5x memory and a 4 TB NVME SSD. It runs NVIDIA DGX OS for local AI development, fine-tuning, and inference.

240 W power supply and thermal profile

The unit operates from a 240 W power supply in a compact form factor. Unified memory reduces data movement between CPU and GPU, lowering overall power draw during model workloads. Wi-Fi 7 and Bluetooth 5.4 connectivity add minimal thermal load. The 10 Gbps LAN port supports high-throughput transfers without active cooling noise in typical office environments.

Preinstalled NVIDIA AI software stack

DGX OS includes the full NVIDIA AI software stack pre-optimised for the GB10 Superchip. This enables immediate prototyping with models up to 200 billion parameters using FP4 precision. ConnectX networking allows linking two systems to scale to 405 billion parameter models. The self-encrypting NVME drive protects model weights and datasets at rest.

商品特色

  • Up to 1 petaFLOP of AI compute at FP4 precision with Blackwell architecture
  • 128 GB LPDDR5x unified memory runs models up to 200 billion parameters locally
  • 20-core Arm CPU (10 Cortex-X925 + 10 Cortex-A725) delivers 64-bit high-performance processing
  • 4 TB NVME self-encrypting storage with 10 Gbps LAN and Wi-Fi 7 connectivity
  • Pre-installed NVIDIA DGX OS and AI stack enables immediate development, fine-tuning and inference

商品規格

BrandNVIDIA
ModelDGX Spark Founders Edition
Part Number940-54242-0000-000
TypeGaming & Entertainment
UsageConsumer
Processor20 core Arm, 10 Cortex-X925 + 10 Cortex-A725 Arm
Processor Main Features64 bit 20-Core Processor
Memory128 GB LPDDR5x, unified system memory
StorageUp to 4 TB NVME.M2 with self-encryption
GraphicsNVIDIA Blackwell Architecture
Power Supply240W
Operating SystemNVIDIA DGX OS
CPU Type20-core Arm,10 Cortex-X925 + 10 Cortex A725
Memory Capacity128GB LPDDR5x
SSD4TB NVMe
Optical Drive TypeNo
LAN Speed10Gbps
WLAN802.11be Wireless LAN
WiFi GenerationWi-Fi 7
BluetoothBluetooth 5.4
Video Ports1 x HDMI 2.1a
Rear USB4 x USB Type-C
RJ451 port
Dimensions (H x W x D)1.99" x 5.91" x 5.91"
Weight2.65 lbs.
出貨重量3.18 kg
包裝尺寸245 × 245 × 152 mm

商品問答

What operating system ships on the system and is it ready for AI development?

NVIDIA DGX OS comes pre-installed and is pre-optimised for AI development, fine-tuning, and inference workloads.

How many AI model parameters can this system handle locally?

The 128 GB unified memory supports models up to 200 billion parameters for local prototyping and inference.

Can two units be linked together for larger models?

Yes, NVIDIA ConnectX networking allows two systems to work together on models up to 405 billion parameters.

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。