At Nvidia's GPU Technology Conference, Dell unveiled the Pro Max desktop featuring the GB300 superchip, capable of 20 petaFLOPS processing power. This workstation is equipped with 784 GB of unified memory, including HBM3e and LPDDR5X, to facilitate extensive AI training and inference tasks. Additionally, it features NVLink-C2C for high-speed CPU-GPU access and the Nvidia ConnectX-8 SuperNIC for optimized AI networking. Dell aims to support developers in scaling AI models akin to server environments, with complementary servers utilizing Nvidia's Blackwell Ultra silicon also on the horizon.
The Pro Max with GB300 is a revolutionary machine that can achieve 20 petaFLOPS, making it a game-changer for AI application development.
With its high-speed NVLink-C2C connectivity and 784 GB of unified system memory, the GB300 is designed to efficiently handle large-scale AI training workloads.
Kevin Terwilliger emphasized that the Pro Max will significantly simplify the workflow for developers, enabling better prototyping, testing, and scalability of AI models.
In addition to desktops, Dell is preparing servers that incorporate Nvidia's Blackwell Ultra silicon, broadening the scope of AI capabilities across platforms.
Collection
[
|
...
]