Lenovo ThinkStation PGX 4TB NVIDIA Grace Blackwell

Lenovo ThinkStation PGX 4TB NVIDIA Grace Blackwell (30KLS00K00) - 3 Year Local Warranty 

Overview

The Lenovo ThinkStation PGX 4TB NVIDIA Grace Blackwell (30KLS00K00) is an enterprise-grade AI workstation built for organisations running intensive compute workloads such as LLM training, reinforcement learning, simulation, and large-scale data science. Engineered on Lenovo’s proven workstation platform, it delivers stable, long-duration performance required for demanding AI development and production pipelines. This system is designed for businesses and research teams scaling advanced AI operations with reliability and efficiency.

AI Performance & Processing Power

Powered by NVIDIA’s Grace Blackwell architecture, the ThinkStation PGX offers exceptional AI acceleration and parallel processing capabilities. The unified Grace CPU and Blackwell GPU architecture significantly improves model training speed, inference optimisation, multi-agent compute, and high-resolution simulation workloads. It is well-suited for teams running internal LLMs, multimodal AI, enterprise inference clusters, and accelerated HPC environments that require consistent throughput and low-latency data handling.

Storage & Workflow Efficiency

Equipped with a high-speed 4TB NVMe storage, the workstation supports enterprise-scale datasets, multi-version AI checkpoints, vector stores, and multimodal training assets without performance bottlenecks. The large capacity enables faster iteration cycles by keeping models and datasets local, while high I/O throughput ensures smooth loading, preprocessing, and continuous processing even under sustained heavy workloads.

Design & Reliability

Built with an advanced thermal and chassis design, the ThinkStation PGX maintains optimum cooling and stable performance during full utilisation, ensuring reliability for mission-critical AI workloads. Its workstation architecture supports extended operation hours, quiet performance, and service-friendly components. Ideal for AI labs, R&D teams, private AI clusters, and enterprise production deployments, it delivers professional durability and long-term operational resilience.

ThinkStation PGX – NVIDIA Grace Blackwell (170W) Technical Specification

Processor & Architecture

• NVIDIA Grace Blackwell Platform (170W)

• Blackwell Architecture delivering up to 1000 AI TOPS

• 20-core ARM CPU

  • 10 × Cortex-X925 high-performance cores

  • 10 × Cortex-A725 efficiency cores

• Designed for next-generation AI workloads, LLM inference, RAG pipelines, and enterprise AI development

Memory

• 128GB LPDDR5X Unified System Memory

• Shared CPU–GPU memory architecture for maximum AI throughput

• High-bandwidth, low-latency design optimized for transformer models and multimodal workloads

Storage

• 4TB NVMe M.2 SSD with self-encryption

• High-speed sequential performance for datasets, checkpoints, and production AI workloads

Connectivity

• Wi-Fi 7 wireless connectivity

• Bluetooth 5.3

• 4 × USB4 Type-C (up to 40Gbps)

• 1 × RJ-45 (10GbE) for high-speed wired networking

Display & Audio

• HDMI 2.1a video output

• HDMI multichannel audio support

• Compatible with high-resolution displays and AV workflows

Software Platform

• Runs NVIDIA DGX™ OS (Ubuntu LTS-based, AI-optimized)

• Includes NVIDIA AI Enterprise, CUDA, TensorRT, and NGC container ecosystem