Jensen Huang, CEO of Nvidia, gave a keynote speech at COMPUTEX 2024 on June 2, 2024. Here are the main points:
* GPUs are used for computer graphics, simulation, and artificial intelligence.
* Nvidia believes that AI is entering a new phase where it can understand the laws of physics and is based on real-world data.
* Nvidia has developed a new chip called Blackwell designed specifically for AI.
* Blackwell is the most complex and high-performance computer ever built. The Blackwell B200 GPU has a total of 208 billion transistors and is manufactured using NVIDIA's self-designed 4NP TSMC process, allowing the connection of two GPU chips on the same single chip. Furthermore, Blackwell uses the second generation of Transformer Engine, doubling the computational load with 4-bit AI inference. Compared to the Hopper GPU, the Blackwell B200 GPU has 128 billion more transistors. leads, 5x the AI performance at 20 petaFlops per chip, and 4x the onboard memory. Additionally, this GPU is paired with 8 HBM3e stacks, delivering the world's fastest memory with 8 TB/s data transfer speeds and up to 192 GB of HBM3e memory. NVIDIA Blackwell B200 GPU is one of two chips Blackwell is used in many different designs. This is NVIDIA's first GPU to use a new type of design called a chiplet (MCM), which consists of two computational chips. Chiplet design helps NVIDIA improve performance and reduce production costs. This is an important step in NVIDIA's technology development.* Architecture: Grace Hopper
* Size: 5nm
* Number of transistors: 208 billion
* AI performance: 20 petaflops
* Memory: HBM3
* Memory bandwidth: 2TB/s
* Interface: PCIe Gen 5
Jensen Huang's speech shows that Nvidia is betting big on the future of AI. The company believes that AI has the potential to revolutionize many industries, and it is investing heavily in developing new technologies that will help realize this potential.
IS COIN INTO THE NEW WAVE?