Google Computes Pi to Well Beyond 3.14 in 157 Days

Googl Cloud

While most of us know the value of pi to be 3.14, a team at Google Cloud appears to have set a new record by calculating the value to at least 100 trillion digits.

Pi, the standard ratio of a circle’s circumference to its diameter, has fascinated mathematicians for millennia. As an irrational number, it can’t be presented as just a two-integer ratio as its true decimal representation is infinite. Before the Google initiative, the record number for the decimal representation of pi was pegged at 62.8 trillion digits.

Before this achievement, Google Cloud hit a record in 2019 when it computed the value of pi to 31.4 trillion digits. Over time, the company’s cloud infrastructure has been bolstered greatly, increasing by around 600%. 

For Emma Haruka Iwao, one of the Google Developer Advocates involved, this massive increase in bandwidth played a major role in the initiative as it enabled programmers to shift nearly four times the amount of data from the 19.1 petabytes achieved in the 2019 attempt.

As of press time, this new record still needs to be verified by judges for Guinness World Records. However, the final numbers were verified through the use of a secondary algorithm called the Bailey Borwein Plouffe formula.

How Did They Do It?

To give you an idea of the immensity of the initiative, Google programmers began running the core program last October 14, 2021. They just achieved 100 trillion digits on March 21st of this year. 

The total time it took to run the calculations was pegged at 157 days, 31 minutes, and 7.651 seconds. The process also added 37.2 digits to the previous total of 62.8 trillion. The process resulted in the machine reading and writing a total of 82 petabytes of data.

To make calculations of this magnitude, programmers require a great deal of both hardware and software. For this, Google Cloud’s pi calculation was made possible using an n2 high-memory 128 machine while running on a Debian Linux 11 operating system. In total, programmers used an Intel Xeon with 128 virtual CPUs (vCPUs0, 864GB of RAM, as well as 663 terabytes of storage, of which 515 were used.

While this record may seem frivolous to some, this initiative demonstrates the full capability of Google Cloud’s cadre of programmers. Moreover, it may lead to breakthroughs down the line for improved processing speed and accuracy.