Bitcoin Forum

What could possibly explain this highly periodic crash in overall network hashing speed?  I think this is not a random thing, but rather some effect due to actions taken by minors.  What is going on?  That is a pretty strong down trend late in the cycle. 

https://i.imgur.com/u8sqCwR.jpg

Livenet testing/shipping cycles could help explain some of it.

Test Lot 1 causes the first increase.  Shutdown Lot 1 and ship causing the first decrease.  During the second cycle, Test Lot 2 while at the same time Lot 1 is arriving at the purchasers and being powered on.  Test Lot 2 gets shut down and shipped to end users.  Test Lot 3 comes online for testing while Test Lot 2 arrives at the end-user.

Rinse. Repeat.

However, I think this scenario, while possible, is unlikely as a manufacturer wouldn't necessarily start/test/shutdown/ship in lots like this.  They'd probably cycle HW in and out of deployment testing and shipping.

some entity switches on their quantum computer prototype in the middle of the cycle, then powers it off as not to cause too much disturbance.
It helps with the expenses,  ;).

It's just statistical noise. No-one would rationally turn off mining capacity and then switch it back on and the power requirements to make a 10% dent in the hash rate are huge (> 10 MW and probably closer to 20 MW).

Here's the last year in log form (which shows the relative size of swings evenly across the last year):

https://blockchain.info/charts/hash-rate?timespan=1year&showDataPoints=false&daysAverageString=1&show_header=true&scale=1&address=

I can't see anything that's dramatically different about the last 4 difficulty periods than the preceding 30-ish and all are well within predictable statistical norms.

Two years ago were guys that were switching their GPU rigs between BTC and LTC (coin-hoppers), now there are ASIC manufacturers test-burning their equipment on the bitcoin network .. However, there`s nothing I know out-there which can impact the hashrate with about 20% at _every_ difficulty jump over the whole history. The chart shape must be influenced this way by the mathematical model used to adapt the difficulty. The LTC chart is showing the same profiles.

It helps to see the Y axis values, just to see how big the jumps are .. https://bitcoinwisdom.com/bitcoin/difficulty

The spikes are not from manufacturers "testing" before shipping.

A facility of that size would take months to build-out and fill with hardware. It is currently being done by Bitfury and others. So it's certainly possible to do. But no manufacturer is going to go through that expense just for testing.

And no massive DC is going to shutdown 20mw of hashing power just to play with the difficulty. It would be financial suicide not to keep everything running 24/7.

The Chinese are putting up farms in 25 days start to finish: http://www.thecoinsman.com/2014/08/bitcoin/inside-one-worlds-largest-bitcoin-mines/
Cement isn't even dry yet and they're hashing away...

As for the pattern, it's just statistical noise.  Like you state, nobody is turning stuff on and off.

I believe it's the Bitmine.ch facility in Iceland.  ;D
One day a catastrophe it is occurred (avalanche, polar bears rush, volcano eruption) and this stop their activity.
After the resume, the hashing rate spike on chart again.

Yeah, I read that article too. It's amazingly frightening how quick they can be. But they don't mention the time it took to plan the construction and acquire the hardware.

Yeah I also had this idea. But shouldn't the renewal of equipement be progressive ?

While I wish that were true it's not.  For every 300W 60GH/s BFL single that shuts down a 300W 450GH/s Antminer S3 replaces it.  Even if 1/2 the people who started mining last year never buy new equipment the hashrate will still go up since we have seen greater than a doubling in efficiency in the last year.

Also now people have options to set up in places with cheaper electricity rather than mining at home.