NVIDIA unveils LPX rack with Groq 3 LPU at GTC
NVIDIA just announced a bold new strategy at the GTC conference, aiming to reshape how AI data centers work.
CEO Jensen Huang introduced the LPX rack, saying it was built around the Groq 3 LPU for super-fast AI performance.
This isn't just about speed: NVIDIA wants to make its hardware work seamlessly together, and is even eyeing robotics and space tech as part of its bigger vision.
LPX boosts tokens per 2nd
The LPX rack uses Groq 3 LPU chips and Rubin GPUs to process data faster and more efficiently with SRAM tech.
Ian Buck, NVIDIA's head of hyper-scale and high-performance computing, said that this setup lowers delays and costs while boosting output, meaning more tokens per second for your favorite AI tools.
The system also includes the Vera-Rubin NVL72 rack and the Spectrum-6 SPX rack, all designed to help handle heavy-duty AI tasks.
The first LPX racks are set to roll out later this year.