The image does not belong to buddystech.com
on June 5, 2024
During his speech at Computex 2024, NVIDIA CEO Jensen Huang surprised the tech world by showing off the company’s new AI-accelerating GPU platform, called the Rubin architecture. Here are the most important points:
- Zen 5 design and IPC Boost: The modern Zen 5 design will be used in the Rubin GPUs.
- The number of instructions per cycle (IPC) will be 16% higher than in the previous model.
- Rubin GPUs are set to come out in 2026.
Number 2: Rubin Ultra: The Next Version – NVIDIA aims to release a tick-tock version called “Rubin Ultra” after the start of Rubin.
- There aren’t many specifics about these chipsets, but Huang says they will save money and energy.
The third book is called “Vera Rubin: A Tribute to an Astronomical Pioneer.” The Rubin design is named for an American scientist named Vera Florence Cooper Rubin, who lived from 1928 to 2016.
- Vera Rubin’s important research on the rates at which galaxies spin gave strong proof that dark matter exists. On the new accelerator board called “Vera Rubin,” there will be Rubin GPUs and an ARM-based CPU named “Vera.”
- HBM4 and NVLink 6 Switch: Rubin GPUs will use HBM4, a new type of high-bandwidth memory that can work at an amazing 3,600 GBps.
- The NVLink 6 Switch will improve speed and connection.
This article from Ars Technica and this one from Tom’s Hardware have more information.
Prepare for NVIDIA’s big step into the future of AI and fast computers! 😊🚀
Learn more