Shulou(Shulou.com)11/24 Report--

CTOnews.com, October 1 (Xinhua)-- at a time when people are debating whether Moore's Law will slow down, still apply, or even die in the 2020s, Nvidia announced "Huang's Law's Law".

According to a new blog post on Nvidia, the processing power of Nvidia's GPU artificial intelligence has increased 1000-fold in the past decade. Huang's Law means that the acceleration seen in "single-chip reasoning performance" will not gradually disappear, but will continue to occur.

According to the article, Nvidia's chief scientist Bill Dally described a "structural shift in computer performance in the post-Moore's law era" based mainly on human ingenuity. It seems difficult to establish a law for this feature, but Dally believes that the chart below marks the beginning of Huang's Law.

CTOnews.com query found that "Huang's Law" is not Nvidia's own original statement. IEEE Spectrum was the first to propose the law named after Huang Renxun, who later became known for a column in the Wall Street Journal.

According to Dally's recent presentation at the Hot Chips 2023 conference, the figure above shows a 1000-fold increase in GPU AI reasoning performance over the past decade. Interestingly, unlike Moore's Law, process shrinking has little effect on the progress of Huang's Law, Dally said.

Dally recalls how to achieve 16x growth by changing the underlying digital processing of Nvidia GPU. The advent of the Nvidia Hopper architecture that uses the Transformer engine brings another huge boost. It is said that Hopper uses a dynamic combination of 8-bit and 16-bit floating-point and integer mathematics to achieve 12.5x performance leaps and save energy.

Dally also said that despite the declining benefits of Moore's Law, he believes Huang's Law will continue to exist.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.