Hacker News Reader

Comment by brndnmtthws | Hacker News Reader

Let's also acknowledge that the "sky is falling" proclamations around Bitcoin mining energy use are based on numbers that aren't even close to accurate. Current network hash rate is about 12EH/s, and an Antminer S9 does about 14TH/s at 1372W. Thus,

    (12000000THs / 14THs) * 1372W = 1,176,000,000W

Which is about 1.176GW. The article cites 8.27TW, which is off by several orders of magnitude.

QED.

Edit: numbers maybe not so far off after all, as it appears to say 8.27TWh/year now.

View on HN

Those Antminer numbers don't actually take into account cooling, other than the fan on the system. Those are the "at the wall" numbers. Since these machines dump their heat into the datacenter presumably if you're running any significant number of them you'll need a cooling system.

You're also operating under the assumption that every miner is already using these systems, which I don't believe is the case. Lots of people are likely running older systems, and as the price of bitcoin gets higher older less efficient machines become more profitable to turn back on.

I think reading the original source of their power consumption numbers will give a much better insight to where those numbers came from- http://blog.zorinaq.com/bitcoin-electricity-consumption/

Even if you double the estimate, it's significantly less than the BS numbers being tossed around. By a lot.

Which of these assumptions leads you to be different than them by 4-5 orders of magnitude? Their numbers are only 3x higher total hashrate estimate and 3x less efficient GH/s over the cutting edge Antminer, which seems reasonable. I'm confused where you diverge so much.

https://digiconomist.net/bitcoin-energy-consumption#assumpti...

Okay, here's the real issue with your comment. You have your units all wrong. You're measuring "watt seconds" instead of "watt hours". Lets fix our units here -

(12000000 (TH per second s) / 14 (TH per second)) * 1372W = 1,176,000,000W/s

Now we need to convert seconds to hours (60 * 60)-

1,176,000,000W/s * 60 * 60 = 4,233,600,000,000W/h

That is roughly 4.23 terawatt hours. Take into account cooling and the fact that not everyone is using the same efficient miner from your example and the numbers look pretty damn accurate to me.

Watts is an instantaneous measurement, so I don't think that's right. If a lightbulb uses 100W and it's running for 1h, it consumes 100Wh of energy.

Exactly my point. When you take a 100w bulb and run it for an hour you get 100Wh of energy. If you run a 100W bulb for 1 second then you have 100W/s of energy used.

During your calculations above you use the "terra hashes per second" number- as you said, "14TH/s at 1372W". The device runs at 1372W, and it also calculates hashes at 14TH/s. You can use the same formula you did to calculate the 10Wh for the lightbulb- if you leave the device running for 1h it will consume 1372W/h of electricity (or 1372 Watts * 1 hour).

If you want to convert the W/h unit to W/d you can multiple the "hour" unit by "24" (since there are 24 hours in a day). To convert in the other direction you do the opposite (divide the day by the number of hours in it). Converting from Wh to Watt seconds is similar.

You're not the first person to be confused by this, and wikipedia has a whole section devoted to explaining it[1].

[1] https://en.wikipedia.org/wiki/Watt#Difference_between_watts....

I just noticed the article now says "terawatt-hours per year", which is more realistic. I can't tell if they amended it, or if I misread.

W/s doesn't make any sense. Its Ws or Wh instead

Are you confusing tw with twhr/y?

I think they amended the article.

Reply on news.ycombinator.com