Home Computing Nvidia Outlines Jensen ‘Huang’s Law’ of Computing

Nvidia Outlines Jensen ‘Huang’s Law’ of Computing

by Amelia Ramiro

In the midst of ongoing debates about the fate of Moore’s Law, Nvidia scientists have introduced a new concept that they believe is set to revolutionize computer performance. Known as Huang’s Law, it heralds the impressive momentum behind Nvidia’s GPU AI-processing prowess, claiming that it has grown 1000-fold over the past decade.

Unlike Moore’s Law, which focuses on the shrinking of transistors and the subsequent increase in computing power, Huang’s Law is centered around human ingenuity. Nvidia Chief Scientist Bill Dally believes that the impressive chart showcasing the increase in GPU AI inference performance is just the beginning of this new era.

Dally points out that process shrinking has had little impact on the progress of Huang’s Law. Instead, Nvidia has achieved significant gains through various technological advancements. For example, changing the underlying number handling of the GPU led to a 16x gain, while the introduction of the Nvidia Hopper architecture with its Transformer Engine delivered a 12.5x performance leap. Other advances, such as structural sparsity and networking technology, have further contributed to these impressive gains.

What makes Huang’s Law particularly intriguing is that it has seen compounded gains in AI inference performance that far outweigh the gains attributed to process improvements. Despite shifting from 28nm to 5nm processes over the past decade, semiconductor process improvements have only accounted for a 2.5x increase in performance, according to Dally.

Looking ahead, Dally and his team see “several opportunities” to further accelerate AI inference processing under Huang’s Law. These include simplifying how numbers are represented, creating more sparsity in AI models, and designing better memory and communication circuits.

While the future of Moore’s Law may be in question, Nvidia is confident in the potential of Huang’s Law to drive ongoing advancements in computer performance. With a strong focus on human ingenuity, it appears that Nvidia is well-positioned to push the boundaries of what is possible in the realm of AI processing.

You may also like