Nvidia urged PC gamers to unite against the new coronavirus (COVID-19) in an official Twitter tweet by donating all of their idle GPU computing power to help scientists do virus research … how to do this?
A while ago, Nvidia tweeted on the official Twitter, calling on PC gamers to unite against the new coronavirus (COVID-19) by donating all of their idle GPU computing power to help scientists do virus research-by adopting The software named Folding @ home allows everyone who uses a computer at home to donate the computing power, and everyone can build a huge supercomputer in a global way by combining them into a network.
Mainland Chinese netizens seem to have much doubts about Nvidia’s appeal. Many people think that the so-called Folding @ home is just making wedding dresses for others, and don’t know how to donate the computing power for no reason. In fact, large-scale distributed computing makes projects such as Folding @ home very feasible. It is just that such a decentralized structure is formed through the unreliable Internet to perform calculations. What should be the problem of bandwidth? How to deal with it? And if the new crown virus research really needs massive computing power, why not directly use supercomputing (EETT editing: super computing / supercomputer)?
What is Folding @ home?
The Folding @ home project was actually launched by Pande Lab of Stanfor University in the United States as early as October 2000. In addition to Pande Laboratories, companies involved in the development of Folding @ home include Sony, Nvidia, etc .; Folding @ home is a decentralized computing project, which mainly does molecular dynamics simulation of proteins. At first, Folding @ home focused on protein folding, but now it turns to more biomedical research, including Alzheimer’s disease, cancer, Ebola virus, COVID-19, etc. So COVID-19 is actually just the latest project researched by Folding @ home (joined in March this year).
So how does Folding @ home provide computing power for these studies? The core idea is to expect to use the idle computing resources of personal computers around the world-of course, these PCs need to install the client software from Folding @ home. As for who installs it, of course, they are willing to contribute computing power to volunteers. In this way, gathering the computing power of the various CPU and GPU hardware around the world can contribute to scientific research.
The entire decentralized system is actually much more complicated than we elaborated. For example, volunteers can track their contributions on the Folding @ home official website, even compare their rankings with the amount of computing power, and a point system; they can also team up Points-Folding @ home official website updates the team’s point rankings all year round to make the computing power contribution more vivid and interesting in presentation. In fact, in the paper published by Folding @ home in 2009, it was mentioned that the client program developed for volunteers integrated OpenGL, mainly to show the situation of protein simulation to volunteers in a graphical way. Folding @ home’s team believes that although this presentation is of little value to scientific research, it is very helpful for volunteers to understand what is happening on their own PC and help Folding @ home spread among the crowd.
This is a very small example of the Folding @ home project, and its overall complexity is still relatively large.
In a sense, in these years of development, Folding @ home has become the world’s most “sturdy” supercomputing system-especially at the beginning of March, Folding @ home opened a research project for the new corona virus, which was @home overall computing power pushed to a new small climax. Strictly speaking, now its computing power has reached 768 petaFLOPs. As of March 25, the computing power of Folding @ home reached 1.5 x86 exaFLOPs, several times the fastest supercomputing in the world.
(Note: The computing power unit given by Folding @ home includes native FLOPs, x86 FLOPs, etc. The floating-point operations per second of a certain type of hardware is native FLOPs; x86 FLOPs refers to if all computing power is classified as x86 CPU, how many FLOPs-1 native GPU FLOP of GPU is usually equivalent to many native x86 FLOPs.)
Many people have no idea about this hashrate number: In September 2007, due to Folding @ home’s enhanced use of PS3 performance and a large number of PS3 game player players to join, Folding @ home hashrate was pushed to 1 petaFLOPs. At that time, Folding @ home became the world’s first computing system that reached petaFLOPs level. At that time, the fastest supercomputing in the world’s Top 500 was BlueGene / L, and its computing power was 0.280 petaFLOPs.
There are several small climaxes in performance in the follow-up. For example, its computing power reached 5 petaFLOPs in 2009. At that time, the performance of IBM Supercomputing Roadrunner was 1.105 petaFLOPs; in mid-2016, Folding @ home exceeded 100 x86 petaFLOPs. In fact, in March of this year, the equivalent computing power of Folding @ home is really fast. On March 20, Folding @ home stated on Twitter that its computing power reached 470 petaFLOPs (958 x86 petaFLOPs), and on March 25, this number flew to 768 petaFLOPs (1.5 x86 exaFLOPs).
As far as scientific research is concerned, these numbers certainly do not represent anything, but Pande Lab has produced 223 scientific research papers.
What is the use of computing power for new coronavirus research?
This is a question related to biological research. In theory, it should be a matter of greater concern to the biological media. Let’s talk briefly. The researchers expect to figure out the structure of potential drug targets for COVID-19, so that new treatments can be designed. COVID-19 can be considered as a close relative of the SARS virus and behaves in a similar manner. The first infection of the two occurs in the lungs. The protein on the surface of the virus will “bind” to the receptor protein of the lung cells. The receptor here is called ACE2.
The viral protein here is called spike protein, which is the red part in the picture above. The therapeutic antibody that scientists are chasing is actually a type of protein. Its goal is to block the binding of viral proteins to receptors and prevent viral infection of lung cells. Therapeutic antibodies for SARS-CoV have been developed. But to develop COVID-19, which is the therapeutic antibody of the new coronavirus, scientists need to better understand the structure of the viral spike protein and how it binds to the ACE2 receptor.
The protein will swing, fold and unfold, forming various shapes. Scientists need to study not only the shape of the viral spike protein, but the way it swings and folds into various shapes, so as to better understand how it interacts with the ACE2 receptor, so that the antibody can be designed from.
Folding @ home mentioned on the official website that the “low resolution” structure of the SARS-CoV virus already exists, plus the known differences between SARS-CoV and COVID-19, this project of Folding @ home is to help build COVID- 19 The structure of the spike protein and recognize the target of the antibody. And building a calculation model requires massive computing power, which is why Folding @ home intervenes.
Although we do not know how much computing power and how long it will take to invest in this new coronavirus spike protein study. However, in terms of its principle, investing in computing power is also a process of gathering less and more, or its effectiveness is probably difficult to manifest in a short period of time-so even more and more people are participating in Folding @ Home project, it can not immediately show results, and let the computational biology immediately show how much power, this is after all a long-term project.
* This article was originally published in EE Times China, please click here to read the full text