NVIDIA is pushing boundaries in the world of AI and high-performance computing (HPC) with the launch of its Blackwell platform. Named after renowned mathematician, David Harold Blackwell, the platform introduces two innovative Graphics Processing Units (GPUs) – the B100 and the B200 – which promise to shake up AI and HPC with groundbreaking advancements.
The B100 GPU, projected to debut in 2024, and its successor, the B200, planned for a 2025 launch, are set to be NVIDIA’s next-generation AI accelerators. Global technology company, Dell, is both a key partner and potential customer for NVIDIA’s new offering and has outlined the timeline and expectations for these GPUs, emphasising their crucial role in AI data centres.
The Blackwell architecture features the GB200––NVIDIA’s innovative “Grace Blackwell” Superchip combo––and the interconnected GB200NVL platform for supercomputing. Together, these components demonstrate NVIDIA’s ambition to become a leader in the AI and compute-ready hardware sector.
NVIDIA’s Blackwell GPUs utilise TSMC’s advanced 3nm manufacturing process, reflecting a significant stride towards heightened efficiency and superior performance. The B100 GPU is forecasted to double the AI performance of its predecessor, the Hopper H200, impressively demonstrating its AI-related capabilities in benchmarks like the GPT-3 175B inference performance. This advance in capacity shows NVIDIA’s determination to continually redefine what is possible in AI and HPC.
However, the immense computational power of the Blackwell GPUs carries substantial power requirements. The B200 is expected to have a power budget of 1000W per GPU, underlining the need for progress in power efficiency and cooling technologies to effectively harness this power. Despite the significant power demands, Dell is confident that it can manage these without direct liquid cooling, reflecting a broader industry trend towards high-performance, energy-efficient computing solutions.
The launch of the NVIDIA Blackwell platform marks a significant turning point in the evolution of the computing landscape – a transition to more capable, efficient, and AI-optimised computing resources. The Blackwell B100 and B200 GPUs propose exceptional performance for AI and HPC applications and pave the way for future technological breakthroughs in the industry. Their arrival on the market is anticipated to profoundly influence data centres, AI research, and the broader technological ecosystem, propelling us into a new epoch of computing excellence.
