Consumer Level Nvidia GPU Outperforms A Supercomputer
We all know that GPUs in our PCs and laptops perform large amounts of calculations to bring the games to life and their characters. Scientists have been trying to use this power for more academic purposes. Some Russian scientists have used a consumer level Nvidia GPU to solve some equations that are usually solved by supercomputers.
The scientists observed that the consumer level Nvidia GPU solved the equation in 15 minutes, which is way faster compared to supercomputer’s two to three days time.
A GPU is designed with multiple threads of processing power, which allows the GPU to perform more simultaneous calculations compared to CPUs. The researchers from Lomonosov Moscow State University were testing that whether or not the consumer level tech would make an accessible alternative to supercomputer, in situations where multiple equations has be run parallel to each other.
The GPU simulated some body scattering equations, which are used to describe how Quantum particles interact with each other. When three or more bodies are included these equations become very difficult to calculate, involving a table containing tens or even hundreds of thousands of rows and columns of data.
According to researchers the the GPU was running Nvidia software and custom programs by researchers, and the GPU performed better then expected.
“We reached a speed we couldn’t even dream of. The program computes 260 million complex double integrals on a desktop computer within three seconds. No comparison with supercomputers! My colleague from the University of Bochum in Germany carried out the calculations using one of the largest supercomputers in Germany with the famous blue gene architecture, which is actually very expensive. And what took his group two or three days we do in 15 minutes without spending a dime.
The processor used by researchers costs somewhere between $300 to $500, which way less compared to millions of dollars spent on supercomputers. The fact is that GPUs have been capable of these kind of calculations for almost past 10 years, but their value is now being appreciated.
It seems that if these tests are a complete success we can see a revolutionary change in the supercomputer, by using consumer level tech.