Computing Science

mukul Dave - "Powers of a Super computer"

SLAM Finalist 3Q4

How did you initially get interested in science?

I grew up watching history, science, and engineering shows on the TV due to my brother’s influence. Whereas history still remains an interest, I felt at ease with the math and logical thinking that are an integral part of science and engineering. Looking back, this shows the importance of good science communication in influencing young minds.

What is your favorite place at the Lab?

Although the break room of my office in Wang Hall (B59) provides an exquisite view of Berkeley and beyond, it becomes even more stunning at higher elevations such as the viewing platform near the old cafeteria area.

Most memorable moment at the Lab?

This would be easily the time when I first got a tour of the machine floor at Wang Hall housing NERSC’s two supercomputers at the time - Cori and Perlmutter. It was fascinating to see the complexity of the wiring, the cooling, the data archive tape handling robots, and other auxiliary systems, and understanding what goes into operating these.

What are your hobbies or interests outside the Lab?

I like reading and listening to podcasts regarding politics, philosophy, and spirituality.

Mukul's Script - "Powers of a Supercomputer"

You may have seen these large wind turbines on your road trip and wondered, what goes into designing those? Let’s say, you want to understand how a turbine blade will react to the air flowing around it. How much power will it generate? The mathematical equations for these physics are impossible to be solved using just pen and paper. So, we use a computer to crunch the numbers. This, as you may know, is called a simulation. Simulating even a single blade on your laptop would take days on end, and this is where a supercomputer comes in, which is basically hundreds to thousands of such laptops, or “processing units”, that are arranged in huge racks in a room and are able to talk to each other with a complex wiring network. Many of you use computer simulations for your research, be it for decoding the genome sequences of DNA, or resolving the mysteries of stars and black holes.

However, in all this talk about simulations, there is a catch! Each of the top 10 supercomputers in the world today consumes at least 3 megawatts of power, which is around what a large turbine would generate. Wouldn't it be great if the computers that we use were more energy efficient? So, the current supercomputer at Berkeley Lab, Perlmutter, has around 4,500 process units and along with each of them, it has graphic cards. Yes, you heard it right. These are the same graphic cards that are used for video games but on Perlmutter, they are used for crunching numbers rather than rendering graphics. These “graphical processing units” or GPUs are able to do more calculations per second and hence increase the energy efficiency of applications by 5 to 10 times. But, they are a bit more complex to program. They need to be instructed exactly what work is to be done, how it is to be distributed, and so on.

My job at the compute center is to help you, the researchers, in using Perlmutter. I am currently working with a weather modeling code that predicts the wind patterns around these turbines. The code runs 30 times faster using GPUs as compared to just the conventional processing units, and I continue pushing these boundaries. So, the next time you see these turbines on your road trip, or check the weather on your phone, do remember the supercomputer and programmers like us who made it happen.

Report abuse