Nvidia dgx

Lenovo ThinkStation PGX is a compact AI workstation that brings development and testing of large models directly to your desktop. It is designed as a companion device for software developers, data scientists, and researchers who want to prototype, test, and fine-tune models locally without reliance on the cloud or a data center. With 128 GB of unified system memory and the NVIDIA GB10 Grace Blackwell superchip, you can run language models with up to 200 billion parameters at your own workstation—or up to 405 billion parameters with two linked PGX units, powered by PetaFlop (1000 TOPS) AI compute. The solution is significantly more cost-effective, agile, and energy-efficient than running large models in the cloud or in a data center, because all development and testing can be done securely on-premises. ThinkStation PGX ships with a preinstalled Linux-based NVIDIA DGX OS operating system and an NVIDIA AI software package. Popular tools and frameworks such as PyTorch and Jupyter are included as well, so you can get started quickly and easily. Key features: NVIDIA GB10 Grace Blackwell AI superchip 20-core Arm processor, 10x Cortex-X925 + 10x Cortex-A725 cores 128GB LPDDR5x 256-bit unified memory Up to 1000 TOPS AI compute capability Support for a 200-billion-parameter language model Support for a 405-billion-parameter language model when daisy-chained 2TB SSD M.2 2242 PCIe 4.0x4 NVMe Opal 2.0 Ubuntu Linux Pro-based NVIDIA DGX OS operating system Preinstalled NVIDIA software: NVIDIA AI Software Stack CUDA 13 GB10 Dashboard AI Workbench Connectivity: 4x USB4-C 20Gbps (DP 2.1), one of which supports power delivery (PD 3.1) 10GbE RJ45 WiFi 7 + BT 5.3 HDMI 2.1a (4K@60Hz) 2x QSFP ConnectX-7 Smart NIC (for chaining two devices) Dimensions 150x150x50.5mm Weight 1.2kg Power adapter 240W USB-C 2x USB-C to A adapter Security features: Self-encrypting NVMe (AES SED) TPM 2.0 NVLink-C2C enclave NVIDIA FW Recovery AMI Setup Password UEFI Secure Boot One-year on-site warranty, extendable to three years