News on November 14th, NVIDIA recently announced the launch of a new generation of chip H200, marking a new level of its chip technology. The new generation of chips has achieved significant improvements in overall performance, demonstrating Nvidia's unremitting efforts in the chip field.
According to the editor’s understanding, the H200 chip is an upgraded product of Nvidia’s H100 chip and adopts the advanced Hopper architecture. The chip integrates 141GB of memory and uses HBM3e memory technology to provide users with a speed of 4.8TB per second. The memory capacity is almost twice that of the previous generation A100 chip, and the bandwidth is increased by 2.4 times.
Nvidia said that the performance of the H200 chip in reasoning or generating answers to questions has improved by 60% to 90% compared with the previous generation H100. Although the performance has been improved, the H200 chip still maintains compatibility with the H1100, providing users with more flexible choices
Under the influence of the new regulations on AI chip control in the United States, H200 chips are temporarily unable to be supplied to Chinese manufacturers. This is not due to price or capacity issues, but to export control restrictions. Under the current circumstances, NVIDIA Intelligent only provides reduced versions of chips with 80% performance degradation to Chinese manufacturers to comply with relevant regulations. With the changes in the international trade environment, this has brought new challenges to the industry
The above is the detailed content of Nvidia releases new chip H200, with soaring performance leading to industry change. For more information, please follow other related articles on the PHP Chinese website!