What is data mining and what does it do to hardware?


What does it mean when a top of the line GPU or something like a bunch of the latest Playstation consoles are being used for data mining? What is data mining and what does it do to hardware? If a GPU was used in data mining,why shouldn’t I buy it?

In: 1

You’re likely conflating data mining (pulling focused data out of giant data sets) with crypto coin (or crypto) mining, which is solving hard math problems to generate unique tokens others are willing to buy.

I think you may be conflating data mining and crypto-currency mining.

Data Mining is a very CPU and storage activity intensive activity. Loosely speaking, data mining is about finding patterns in huge piles of data you already have. Example: Facebook, Google, Apple etc all collect a HUGE amount of information about their users. Where you are, where you go, what ads previously got your attention and so on. And most people figure “who cares if Google knows I go to the CVS at the corner of 1st and Main every week?” And, to most people, that’s all Google can determine. However, when you collect enough data, you can start connecting the dots, making inferences about your users that can reveal some surprising details. You can connect the data you collected from the users phone to data you bought from the DMV, credit rating agencies, lots of things to build up a full profile of an individual. This process is called Data Mining and it is hugely important to companies from a customer relations, customer engagement and marketing stand points. The process requires powerful servers hosting very large databases and sophisticated data analysis algorithms . As far as I know, no one is seriously doing data mining using GPUs and video cards.

Crypto Mining is a very GPU intensive activity. Very loosely speaking, to get a Bitcoin, you have to have a computer do some complex math. There are only a fixed number of Bitcoin math answers to be found, gradually taking longer and longer to solve/find all the time. Once you find/solve a Bitcoin, your computer announces that fact to every one else also trying to mine Bitcoin. Whoever submits the correct math first gets the Bitcoin. (while the math to find a Bitcoin is hard, the math needed to prove you have the correct answer is fairly easy.) The way the math works is easy to speed up by splitting the job among portions of the processor. (experts call this parallelism). It just so happens that rendering complex graphics for things like computer games is also very easy to run in a parallel fashion. Thus, GPUs are the faster and more energy efficient way to mine Bitcoin. They split the job up among the hundreds or thousands of tiny cores built to handle simple math.

The gold rush-like boom in crypto has put a huge demand on GPU makers to keep up. This has really jacked up the prices of video cards for gamers. We’re now starting to see that stabilize and we’re starting to see older cards that were used for crypto being sold. The problem is, crypto miners do not replace the GPU card until the increased speed and energy efficiency of a newer card makes it worthwhile. Usually what they do is start up an entirely separate operation before disposing of the older hardware. (They are in a race against other crypto miners after all.) Until that card is decommissioned and sold, it is running at 100% 24/7 for up to two or three years. Even the most avid gamer can’t max their card at 100% for a week straight. So graphics hardware aimed at the gamer crowd is designed around the expectation of a “peaky” duty cycle. Peg at 100% for a gaming session, idle at maybe 10-20% the rest of the time. That means that a gamer graphic card used for crypto mining is being exposed to high heat loads its entire life. This extra heat can degrade the performance of the capacitors on that board. This leads to possible early and unpredictable failure. Moreover, it is a failure that can potentially damage the PC it is installed in.