Product Launches

Dell's 20 PFLOPS AI workstation targets trillion-parameter models on-prem

Share
Dell's 20 PFLOPS AI workstation targets trillion-parameter models on-prem

Dell Technologies has unveiled the 'Dell Pro Max with GB300', a desktop workstation packing 20 PFLOPS of AI performance and 748GB of total memory, designed to handle trillion-parameter models on-device. The system is aimed at enterprise AI teams and research institutions seeking on-premises AI development solutions.

TL;DR

What happened

On August 20, Dell Technologies introduced the 'Dell Pro Max with GB300', a high-performance desktop workstation powered by the NVIDIA GB300 Grace Blackwell Ultra Superchip. The system is designed for large-scale model training, multi-user collaboration, and memory-intensive workloads, with a focus on on-premises AI development.

The Pro Max with GB300 features 748GB of coherent memory, combining 496GB of system memory and 252GB of graphics memory. It delivers 20 PFLOPS of AI performance in FP4 compute and can process models with up to 1 trillion parameters on-device. The system also includes an NVIDIA RTX PRO 2000 Blackwell GPU and supports up to 16TB of storage.

Why it matters

The Pro Max with GB300 targets enterprise AI teams and research institutions looking to reduce dependence on cloud GPUs. Its ability to handle trillion-parameter models on-device makes it a compelling option for organizations seeking to bring AI development in-house.

The system's integration with Dell AI Factory with NVIDIA provides a consistent AI development environment from local to cloud, simplifying the transition from proof-of-concept to full-device deployment. This seamless integration is a key factor in enterprise procurement decisions.

Key facts

  • The Dell Pro Max with GB300 is powered by the NVIDIA GB300 Grace Blackwell Ultra Superchip, featuring 72 CPU cores and GPU cores.
  • The system offers 748GB of coherent memory, combining 496GB of system memory and 252GB of graphics memory.
  • It delivers 20 PFLOPS of AI performance in FP4 compute and can process models with up to 1 trillion parameters on-device.
  • The system includes an NVIDIA RTX PRO 2000 Blackwell GPU and supports up to 16TB of storage.
  • Dell also offers the compact 'Dell Pro Max with GB10', featuring the NVIDIA GB10 Grace Blackwell Superchip and 128GB of unified memory.
  • The GB10-powered machine can handle models with up to 200 billion parameters standalone or 400 billion parameters with two units connected.
  • Both models come with pre-installed AI software stacks, including system-specific configurations and core CUDA-X libraries.
  • The Pro Max with GB300 is currently sold out on Dell's e-commerce site.

Context

The launch of the Dell Pro Max with GB300 underscores the growing demand for on-premises AI development solutions. As organizations seek to reduce their reliance on cloud GPUs, high-performance desktop workstations like the Pro Max with GB300 are becoming increasingly important.

The integration of NVIDIA's GB300 Grace Blackwell Ultra Superchip and the RTX PRO 2000 Blackwell GPU highlights the collaboration between Dell and NVIDIA in advancing AI capabilities. This partnership is crucial in providing a consistent AI development environment from local to cloud, catering to the needs of enterprise AI teams and research institutions.

Topics

Join the discussion

Have a take on this story? Weigh in with our community on Facebook.

💬 Discuss on Facebook →