Research finds bitcoin mining can be improved by 30% with approximate computing
A new technique, developed by a student team led by CSL Associate Professor Rakesh Kumar, could maximize bitcoin mining profits by up to 30%, representing an application of a larger body of work in approximate computing.
Bitcoins are the most popular form of cryptocurrency today. A bedrock of the Bitcoin framework is mining, a computationally intensive process that monetarily rewards so-called miners when they verify the trustworthiness of online transactions between two parties. It has become increasingly difficult to profit from Bitcoin mining because of the required computational power, but this new approximation technique will work to increase profits and security for miners.
The technique relies on a field called approximate computing, an area of expertise for Kumar. Bitcoin mining is tolerant to errors, meaning that even when approximations are made in the transactions—less reliability but faster processing speeds—the mining is still accurate.
“Approximate computing allows for reliable computing on unreliable devices, so bitcoin mining is a great application for this work,” said Kumar, associate professor of electrical and computer engineering. “We can use it to improve the security of and create trust in these online transactions, while also increasing profits for miners.”
This work will appear at the Design Automation Conference this June and, in addition to receiving significant attention from the media, there are also ongoing conversations with a mining company to possibly commercialize this technology.
“I’m extremely grateful for the time, opportunity, and guidance Dr. Kumar provided me to work on a substantial project such as this while still an undergraduate. It has motivated me to continue my education in graduate school with a focus in computer architecture,” said Vilim, who will begin graduate work in the fall of 2016 at Stanford. “The project provided an opportunity to explore many research areas with which I was only loosely familiar and a chance to study topics that can’t be learned in a class alone.”
Vilim worked with CSL graduate student Henry Duwe, who is generally focused on designing low-power processors, including approximate computing techniques.
“This work is particularly interesting since it highlights an application whose subcomputations are not obviously amenable to approximate computing,” said Duwe. “Looking forward, this work suggests that future miners are very likely to use approximation in order to keep competitors from getting a significant profit advantage.”
“The successful execution of this work is a testament to the quality of students at the University of Illinois, and a good example of the high level of work undergraduate students can conduct,” said Kumar. “Work like this can help students shape and clarify what they’d like to do in the future, both in their research and career.”