Jensen Huang’s Speech at Unveiling of Energy Department’s Summit Supercomputer

by Jensen Huang

Secretary Perry, Governor Haslam, Senator Alexander, Director Zacharia, Ginnie, John, and the IBM team, and all of the amazing researchers of Oak Ridge National Laboratory, it’s a great pleasure to be here today to inaugurate the Summit supercomputer.

Summit is a remarkable machine. The fastest computer ever made. Every statistic of Summit is amazing. It has 200 petaflops of high-precision computation — many times faster than the world’s fastest.

And because of a new invention we call the Volta Tensor Core GPU, it also does multi-precision mathematics and pumps out 3 quintillion operations per second — 3 exaops. Exa is a big number — the universe is only half an exa-seconds old.

With 3 quintillion operations per second, Summit can simulate a human brain where each of the 100 billion neurons is modeled with 30 million instructions each second. Summit is fast.

Summit is a magnificent scientific instrument that will attract the world’s great scientists. Supercomputers are essential instruments of science. Like other magnificent scientific instruments — the Large Hadron Collider, the LIGO gravitational wave observatory, the 39-meter ELT Extremely Large Telescope or the Square Kilometer Array radio telescope — the Summit supercomputer will attract the world’s great scientists — the Einsteins of our time — to do their lives’ work, in their lifetime. To them, we have built a time machine.

Summit is a milestone in a global race. Not a race to space, but a race to all human knowledge — a race to understand everything. The prize for winning is the science, which is then shared with all humanity.

So why the fuss when everyone gets the prize?

The key is in the derivatives. The community of world-class researchers attracted to the instrument. The methods and engineering breakthroughs derived to create Summit. If bold, conventional approaches won’t cut it — everything breaks — too slow, too costly, too large, impossible to cool, too hard to program. All of that was true at the start of Summit.

Summit has a remarkable backstory that has shaped the future of computing. Summit is possible today because of four giant bets:

  • In 2005, we bet our company on a new way to do computing because CPU scaling was going to end and the world needs a path forward. In 2007, we announced the first CUDA GPU.
  • In 2009, the leaders of Oak Ridge bet their careers and the lab on a then-nascent NVIDIA technology. Then in 2012, we announced Titan, the fastest supercomputer at the time and five times more energy efficient than anything else.
  • In 2013, IBM, the world’s most important computer systems company, made a big bet on NVIDIA, a then-tiny company, to build its future supercomputers.
  • In 2014, NVIDIA bet the farm again — this time on a new approach that fuses high-precision computing and deep learning into one architecture — the Volta Tensor Core GPU was born.

These were four “bet-the-farm” decisions that led to Summit.

Summit is a new breed of computer. Summit is the world’s largest AI supercomputer, a machine that learns — its software will write software — amazing software no human can write.

Summit will simulate the world from existing knowledge, from first principles, while learning to recognize complex patterns, “connecting the dots” in a sea of incomprehensible data to discover new knowledge.

Today is a proud moment for our nation — not just for being first or building the biggest — but for having the courage of conviction to make big bets on our future.

Today is the start of an exciting new path for computing.

Today is a great day for science.

To my friends and colleagues at Oak Ridge National Lab, a huge congratulation for reaching this Summit.

From here, we can all jump to the next peak.

Read the blog on Summit’s launch.