Compact desktop supercomputer with NVIDIA® GB10 Grace Blackwell superchip for advanced AI workloads.

• Powered by the NVIDIA® GB10 superchip with 20-core Arm CPU and Blackwell GPU with 5th generation tensor cores.
• Up to 1,000 TOPS of performance for AI processing and complex model inference.
• 128 GB of unified memory, ideal for handling model development.
• NVLink-C2C interconnect for ultra-fast communication between CPU and GPU, with 5x the bandwidth of PCIe 5.0.
• FP4 data format support, optimised for generative AI and efficient performance.
• Extended connectivity via NVIDIA ConnectX-7 network module, allowing two GX10s to be interconnected to scale up to models such as Llama 3.1 (405B parameters).
• Scalable, local architecture – ideal for prototyping, fine tuning and inference from the desktop.
• Seamless transition to cloud environments thanks to NVIDIA’s software ecosystem, such as NVIDIA DGX Cloud.
• Compact, quiet, and efficient design – perfect for labs, technical offices, and research facilities.