Lenovo PGX Workstation, NVIDIA GPU, 128GB RAM, 2TB SSD, DGX OS

The Lenovo PGX Workstation is a compact, high-performance computing solution designed specifically for AI development, model testing, and data-intensive research. By integrating the powerful NVIDIA GB10 Grace Blackwell superchip with 128GB of unified memory, this workstation allows developers to prototype and fine-tune large language models locally. It eliminates the need for constant cloud reliance, providing a secure, energy-efficient, and agile environment for professional AI workflows right on your desktop.

Advantages

This workstation offers a significant leap in local computing power, enabling the execution of language models with up to 200 billion parameters on a single unit, or up to 405 billion parameters when two units are daisy-chained. The inclusion of the NVIDIA DGX OS and a comprehensive AI software stack ensures that users can begin development immediately without complex configuration. Its compact form factor, combined with high-speed connectivity like 10GbE and WiFi 7, makes it an ideal choice for modern office or laboratory environments where space and performance are equally critical.

Suitable for

• Prototyping and fine-tuning large language models
• Local AI model development and testing
• Data science and complex computational research
• Enterprise-level AI software development

When it may not be suitable

This system is specialized for AI development and may not be the optimal choice for general-purpose office tasks, standard gaming, or creative media production. Users requiring extensive internal storage expansion or traditional multi-GPU desktop configurations may find the compact, unified-memory architecture limiting for their specific hardware needs.

Key specifications

• NVIDIA GB10 Grace Blackwell AI superchip
• 20-core Arm processor architecture
• 128GB LPDDR5x unified memory
• 2TB PCIe 4.0 NVMe Opal 2.0 SSD
• Up to 1000 TOPS AI compute performance
• NVIDIA DGX OS (Ubuntu Linux Pro-based)
• Connectivity: 4x USB4-C, 10GbE RJ45, WiFi 7, Bluetooth 5.3, HDMI 2.1a
• Dual QSFP ConnectX-7 Smart NIC ports for device chaining
• Compact dimensions: 150x150x50.5mm
• Security: TPM 2.0, Secure Boot, and self-encrypting NVMe storage

How to use

To begin, connect the workstation to a power source using the provided 240W USB-C adapter and attach your peripherals via the USB4 or HDMI ports. Upon powering on, the preinstalled NVIDIA DGX OS will guide you through the initial setup. Utilize the pre-loaded NVIDIA AI Workbench and CUDA 13 environment to access your preferred frameworks like PyTorch. For advanced scaling, use the QSFP ConnectX-7 ports to link a second PGX unit to increase your model parameter capacity.

What to consider before purchase

• Verify that your specific AI frameworks are compatible with the NVIDIA GB10 architecture.
• Ensure your workspace has adequate ventilation for high-performance computing.
• Confirm if your workflow requires the daisy-chaining capability of two units for larger model support.
• Check the availability of 10GbE infrastructure if you intend to utilize the high-speed network port.

Usage recommendations

For optimal performance and security, ensure that the NVIDIA firmware is kept up to date using the provided recovery and management tools. Always utilize the NVMe self-encryption features to protect sensitive research data, and leverage the AI Workbench to streamline your development lifecycle.