“The most powerful chip in the world” – Nvidia says its new Blackwell will power the next generation of AI
The next generation of AI will be powered by Nvidia hardware, the company has said after unveiling its next generation of GPUs.
Company CEO Jensen Huang presented the new Blackwell chips at Nvidia GTC 2024 today, promising a big step forward in AI power and efficiency.
The first Blackwell “superchip”, the GB200, will arrive later this year, with the ability to scale from a single rack to an entire data center, as Nvidia looks to continue its leadership in the AI race. .
Nvidia Blackwell
Representing a significant step forward for the company’s hardware over its predecessor Hopper, Huang noted that Blackwell contains 208 billion transistors (compared to 80 billion in Hopper) across the two GPU chips, which are linked by 10 TB /second chip-to-chip coupling into a single, unified GPU.
This makes Blackwell up to 30x faster than Hopper when it comes to AI inference tasks, and offers up to 20 petaflops of FP4 power, far ahead of anything else on the market today.
During his keynote, Huang emphasized not only the enormous power jump between Blackwell and Hopper, but also the large difference in size.
“Blackwell is not a chip, it is the name of a platform,” Huang said. “Hopper is great, but we need bigger GPUs.”
Despite this, Nvidia says Blackwell can reduce costs and energy consumption by up to 25x, giving the example of training a 1.8 trillion parameter model – which previously required 8,000 Hopper GPUs and 15 megawatts of power – but now with just 2,000 can be done. Blackwell GPUs consume only four megawatts.
The new GB200 brings together two Nvidia B200 Tensor Core GPUs and a Grace CPU to create what the company simply calls “a massive superchip” capable of advancing AI development, delivering 7x the performance and four times the training speed of an H10O powered system.
The company also unveiled a next-generation NVLink network switch chip with 50 billion transistors, meaning 576 GPUs can talk to each other, creating 1.8 terabytes per second of bidirectional bandwidth.
Nvidia has already signed up a host of key partners to build Blackwell-powered systems, with AWS, Google Cloud, Microsoft Azure and Oracle Cloud Infrastructure already on board alongside a host of big industry names.