Name: 比特幣電源問題 - Computerphile (Bitcoin Power Problem - Computerphile)
Uploaded: 2021-01-14T10:16:13.000Z
Duration: 8 min 22 s
Description: 【看影片學英語】數萬部 YouTube 影片，搭配英漢字典即點即查，輕鬆掌握單字發音與用法，長久累積看電影不必再看字幕。

程度副詞

It's remarkable just how much energy is involved in doing these calculations associated with Bitcoin mining.

The way that it's approached right across the Internet is for a number of years has been well, let's look at how many hash attempts are being taken.

You know, what's the What's the Realtor of actually attempting to crack this code if we're gonna put it in that context versus what's the sort of energy cost in terms of efficiency off the hardware that's being used to solve this?

A better way of doing it, which is described in this rather good paper, which I'm sure Sean will link to below is to look at the economics.

But from a from a physics of perspective, it's really quite neat in many ways, from many perspectives is what's called a Fermi problem.

The physicist for me was really adept out working out reasonable estimates for quantities are really good.

How many on this was the first sort of the archetypal, firmly problem?

How many piano tuners other in Chicago are we to do that?

How many piano tuners are there in Nottingham.

So you gotta think about well, first, you get hit by that.

I don't know how much you ever gonna answer that.

You have per house, you know, let's say, probably isn't in Notting.

Many do you have per house, say, two or three somewhere like that, maybe two and 1/2 kids.

Then we'd have to think about how often does a piano need to be tuned Once a year?

Maybe, Let's say once a year you take up all these and you stuck them up and its estimate piled upon estimate.

But what's remarkable is that you can get are pretty good estimate.

In many cases, obviously, each one of those estimates has got its own, never valued its uncertainty.

But sometimes what we want to do is get your calculator the 10th decimal place, all the times.

We just want those estimates on when you're coming back to the Bitcoin issue and the energy consumption issue.

When you look across the Web, you'll find many, many different estimates, and it's effectively affirming because nobody really knows.

No, unless you can take each piece of equipment or each piece of hardware and see, you know, somebody could be running the absolute state of the art kit and then somebody kind of a farmer.

PlayStations we're not gonna get a very, you know, on estimate that's accurate to the end decimal place.

But we can at least think about well, could we put a lower limit on this?

Well, let's think about, you know, absolute state of the art on Let's think about the energy efficiency of the absolute state of the art.

Um, this paper is quite useful because we turned to terrible one we find right at the top, and I'm sure many in the audience will be much more familiar with this than I am.

Arrest nine house rate of 14 terror hashes per second, so terror tend to the 12.

We find that the power efficiency so the jewels per gig a hash.

So the amount of energy that's chewed up a Giger harsh, which means a total of 10 to the nine hush calculations per second is off the order of 0.98 Jules per gig a hush.

It's a very recent paper, but what you can see is the power efficiency.

So we need to think about, well, how many actual operations many attempts hush attempts are happening per second on it is an absolute astronomical number, literally astronomical.

You know that that's the number of hash attempts that are happening across the entire sort of Bitcoin equal system, if you want to put it, which is a remarkable about.

It is truly astronomical because the number of Galaxies in the observable universe is only off order.

10 to the nine, So this is 10 to the nine by 10 to the nine dot number.

That's just a mind bogglingly astronomical number, so it's tend to the 18.

Let's put it that way, So that means what we have to do is is we need to take this number this 0.9 it multiply it by the 2.6 by 10 to the ninth we worked out.

It turns out that you were talking about 2.6 gigawatts of power, so it just translated there from power to energy.

Remember, power is the rate of doing work, which is the rate of energy expenditure effectively, which is different from energy, will come back and think of energy in a second.

So it's the difference between, for example, kilowatts power and kilowatt hours energy.

But you work it out and it turns out to be two points off the order 2.6 gigawatts.

Adding on lots of other desperate place is pointless because of the estimates that we've used in there.

In fact, you could make an argument that even having one desperate place under the circumstances, probably pushing it because we have sold this.

This is just really we're thinking about the lower limit.

Actually, obviously all machines involved in Bitcoin mining are not of this specifications.

We also haven't bundled in, for example, the costs associated with cooling.

We haven't balled, and this is this is purely the efficiency we haven't bundled in cost assorted actually with making the devices as well.

So there are capital costume or usage costs, etcetera.

It's a physics problem, but it's also a couple of physics and economic problem.

One of the first papers looked at a lot of attention that looked into this was from some researchers in Ireland back in 2014 2.6 ago.

Watch is not that far removed from the 3.1 gigawatts off that Ireland users, which is given my accent, is fairly close to home close to my heart.

On its now there's ah website called DJ Economist, where you can look up the current energy consumption and all the values associated with with Bitcoin mining on, we're now we've gone beyond Ireland and we're now up.

I think it exceeds the Czech Republic's consumption.

So and and it's it's going up, it's steadily going on.

So if you got a DJ economist website which I thoroughly recommend, there many items associated.

But one of those items is that the the cost per transaction basically in terms of the energy cost in terms of kilowatts per hour and it turns out that it's off the order off per transaction of 1000 kilowatts per hour.

It's difficult to put that in sort of contact sometimes, but that's an awful lot of energy when you total it up in terms off cost per kilowatt hour in the U.

That's something like Fluctuates obviously, but it's something like 14 pence per kilowatt are in the U.

This paper producers much smaller estimates in terms of its average is talking about maybe five cents per kilowatt are.

Anyway, we're talking about orders of magnitude is firmly problem again on DDE.

You compare 14 pence per kilowatt hour times 1000 kilowatt hours.

There's a question that a coolie you've got massive farms.

You've obviously got out in the financial cost of keeping them cool.

You've got our guests that you know the bar goes open, open, open, open, open terms off.

The computational power you need to process is and it's built into the system to have this sort of arms, wrists, terms of building up, that the complexity.

But you add that all in still at the moment, the pure economics of this, or is that we haven't hit the ceiling.

That 2.6 Guillot gigawatts is definitely not the ceiling.

There's an awful lot Maur in terms of where the energy is going upscale.

These are all estimates that would be coming back to the Fermi problem idea.

You might say we're plucking values out of thin air.

We're not, You know, they're obviously informed values, but it makes me smile when I see you know, people quoting these values 23 and four decimal places when the ever bars like that on their Colton with a precision like that, the interesting courses this is considering the whole system on aggregate for the individual prospectors and where for the individual minor is there's a hell of a lot of risk or there can be a hell of a lot of risk involved on.

字幕列表影片播放

比特幣電源問題 - Computerphile (Bitcoin Power Problem - Computerphile)

sort

individual

approach

quote