Nvidia, a leader in AI chip technology, has introduced its latest breakthrough: the Blackwell B200 GPU and GB200 “superchip.” These innovations are poised to extend Nvidia’s lead in the industry and revolutionize AI computing.
The new B200 GPU boasts an impressive 20 petaflops of FP4 horsepower, powered by a staggering 208 billion transistors. Additionally, Nvidia claims that the GB200 superchip, which combines two B200 GPUs with a Grace CPU, can deliver 30 times the performance for LLM inference workloads while significantly improving efficiency, reducing cost, and consuming less energy compared to previous models.
Training massive AI models has traditionally been resource-intensive, requiring thousands of GPUs and substantial power consumption. However, Nvidia’s latest advancements have dramatically reduced the resources needed. With just 2,000 Blackwell GPUs, Nvidia claims to achieve the same performance while consuming only a fraction of the power.
Key improvements include a second-gen transformer engine that doubles compute, bandwidth, and model size by using four bits for each neuron. Additionally, a next-gen NVLink switch enables seamless communication between 576 GPUs, boasting 1.8 terabytes per second of bidirectional bandwidth.
Nvidia’s innovative solutions are designed to meet the demands of large-scale AI deployments. The GB200 NVL72, for example, combines 36 CPUs and 72 GPUs into a single liquid-cooled rack, delivering unparalleled AI training performance. Furthermore, Nvidia offers comprehensive solutions like the DGX Superpod, which combines multiple systems to deliver exceptional computing power.
Major cloud service providers, including Amazon, Google, Microsoft, and Oracle, have already expressed interest in Nvidia’s latest offerings. These companies are planning to integrate NVL72 racks into their cloud services, signaling a significant shift in AI infrastructure.
While gaming enthusiasts may have to wait for news on new gaming GPUs, Nvidia’s Blackwell GPU architecture is expected to power future desktop graphics cards. With its relentless focus on innovation, Nvidia continues to push the boundaries of AI computing, driving progress and shaping the future of technology.
Nvidia CEO estimates that the new GPUs might cost up to $40K each.
