NVIDIA chief scientist on 3D chip, China's rise

GTC 2012, the Nvidia Chief scientist Bill Dally, who served as dean of Computer science at Stanford University in an interview with EE Times, talked about 3D integration circuits, technical aspects of China's rise and the status of U.S. investment in research and development.

On the 3D chip, Dally said that the future of the GPU is a way to consolidate a number of 3D-tiered memory, a design that has the potential to play a higher bandwidth effect and lower overall power consumption. The Hyper Memory cube had actually been discussed with Nvidia, but Dally said the US-light request had "ambitions" to grab a bigger stake in the value chain because Nvidia wanted to design its own chips alone, and they refused.

In addition to the production of 3D stack memory capacity of manufacturers only Japan's delta, unfortunately, the company has entered bankruptcy protection procedures this year and received the aid of the United States to be integrated. So Dally is now pinning its hopes on the largest Samsung in the storage industry to act as the manufacturer of the product.

When it comes to the possibility that Chinese manufacturers will launch the GPU to compete in the market, the former dean of Computer science at Stanford University didn't answer the question directly, but suggested that China's pace of development was "very scary": five years ago, Godson launched only to be ridiculed, but now China's products are fully capable of many applications. If the pace continues, Dally expects China to overtake the west in 3-5 years.

Speaking of domestic research and development funding, Dally says the US ZF is gradually cutting back on computing research and development, and only innovation can bring competitive products, and sometimes innovation cannot rely solely on private companies but also on ZF investment assistance. A special example is parallel computing, although now parallel computing based on GPU or CPU is a big hit in supercomputers and other systems, but it is not taught in schools, and the talent still writes old code.