Nvidia Unveils AI Supercomputer with GB200 Superchip
By
Rafaela Lopez Fernandez
1 min read
⚠️ Heads up: this article is from our "experimental era" — a beautiful mess of enthusiasm ✨, caffeine ☕, and user-submitted chaos 🤹. We kept it because it’s part of our journey 🛤️ (and hey, everyone has awkward teenage years 😅).
Nvidia revealed its new AI supercomputer, the DGX SuperPOD, empowered by the GB200 Grace Blackwell Superchip. This system boasts 11.5 exaflops of AI supercomputing and 240 terabytes of fast memory. The GB200 Superchip, with 36 Nvidia Arm-architecture Grace CPUs and 72 Nvidia Blackwell GPUs, delivers a remarkable 30 times better performance than Nvidia's current leading H100 Tensor Core GPU. Furthermore, major cloud providers like Oracle and Google plan to integrate the GB200 NVL72 systems into their infrastructure. This new development showcases Nvidia's innovation in AI supercomputing and its ability to deliver high-performance solutions for complex AI workloads, setting new standards in the industry.