NVIDIA is a technology company known for designing and manufacturing graphics processing units (GPUs) that runs on large computers that process data and power generative AI.
NVIDIA has unveiled the next-generation GH200 Grace Hopper super chip platform with the world’s first HBM3e processor for accelerated computing and generative AI.
The new platform is created to handle the world’s most complex generative AI workloads, spanning large language models, recommender systems, and vector databases.
The HBM3e processor offers groundbreaking memory, bandwidth, the ability to connect multiple GPUs for exceptional performance, and an easily scalable server design. HBM3e memory is 50% faster than HBM3. It delivers 10TB/second of combined bandwidth, allowing the new platform to run models 3.5 times larger than the previous version while improving performance with three times faster memory bandwidth.
A single server with 144 ARM Neoverse cores, eight petaflops of AI performance, and 282GB of the latest HBM3e memory technology in a dual configuration delivering 3.5 times more memory capacity and three times more bandwidth than current generation offerings.
“To meet surging demand for generative AI, data centers require accelerated computing platforms with specialized needs,” said Jensen Huang, founder and CEO of NVIDIA. “The new GH200 Grace Hopper Superchip platform delivers this with exceptional memory technology and bandwidth to improve throughput, the ability to connect GPUs to aggregate performance without compromise, and a server design that can be easily deployed across the entire data center.”
The new platform uses the Grace Hopper superchip, which can be connected to additional Superchips via NVIDIA NVLink this will allow them to work together to deploy the massive models used for generative AI. This high-speed, compatible technology gives the GPU full access to CPU memory, providing a combined 1.2TB of fast memory in a dual configuration.
NVIDIA’s next-generation GH200 Grace Hopper superchip will be available in a wide range of configurations, and the system manufacturers are expected to deliver systems based on this platform in Q2 of 2024.
