Hacker News Reader

Hacker News

Comment by squarefoot | Hacker News Reader

Parent Full thread

squarefoot

One word: yes.

https://www.hardwaretimes.com/gpu-mining-farm-w-over-4000-ge...

https://hothardware.com/news/a-look-inside-a-massive-rtx-307...

https://www.tomshardware.com/news/gpu-shortages-worsen-crypt...

And energy prices, of course. More demand = higher prices.

Also keep in mind that cryptocurrencies become harder to mine with time, therefore the pursuit of more powerful hardware and more energy isn't going to stop, therefore there's absolutely no way prices of hardware and energy will return to normal until governments won't stop this nonsense.

Pushing for nuclear or carbon is just hiding the problem under the carpet: we don't need more energy because we already have plenty, but we must stop wasting it, especially in a way that is clearly bringing us to a path in which we will need more and more every year.

View on HN

But we do need more energy.

Energy can fix essentially any problem we have. Need to desalinate water? Energy. Need to heat your Arctic home? Energy.

Being able to use energy in crazy and inefficient ways is a feature, not a bug. Everything you ever do outside of subsistence level agriculture is proof of that.

The ideal state of being would be one in which we have limitless fusion power or similar and we basically just do whatever we wish. If we get that and alchemy we win the Universe.

The problem is in not pricing in externalities. If I have a fusion reactor in my garden then I can use as much of it as I want and it has no negative effect on you.

If we tax carbon heavily then Bitcoin and friends will actually get us to that future more quickly.

That politicians are fucking around with stuff like low emission zones and labels on meat or whatever rather than just taxing carbon is almost genocidal at this point. It's hilariously off the mark.

> we do need more energy

Sounds like ideology to me. I, for one, need many thing from basic food to nice music, but energy on its own in not in that list.

I wonder what people picture when they say "energy" but probably not that thing that is equivalent to movement and that is always a constant by definition (at least here and now), and that increase entropy when we manipulate it (entropy being that thing every life forms is designed to fight).

Anyway, carbon in the atmosphere may be a bigger problem than a fusion reactor in your garden, yet I'm a bit concerned by how much faster "we" could destroy our environment by producing more useless stuff if it became cheaper to do so.

> I, for one, need many thing from basic food to nice music, but energy on its own in not in that list.

Energy is a dependency of both of the things you listed - and an abundance of energy makes those things more accessible to more people, yourself and myself included.

It's also a dependency of everything detrimental. My point was: it's not central in itself, but is framed to appear so because of ideology.

> The ideal state of being would be one in which we have limitless fusion power or similar and we basically just do whatever we wish. If we get that and alchemy we win the Universe.

No, using energy inevitably changes the state of world we live in (that is actually the point). Use crazy amounts of energy in a distributed and unregulated manner, and the world will be nowhere like the one we evolved in, and we will die.

"therefore there's absolutely no way prices of hardware and energy will return to normal until governments won't stop this nonsense." It will stop once the crypto prices reach parity with the hardware investments plus or minus further speculation.

Can you be more specific here? I see no mechanism for this to happen. Difficulty will simply increase from that point.

I believe they mean that once is gets to a point where crypto mining is no longer profitable (revs < costs) the demand for more powerful machines / tech will decrease

At least on the ethereum side they are moving to a Proof of Stake system which doesn't require the insane hardware and energy that Proof of Work consensus requires. So there will be some relief on the horizon.

Also, I expect over time that fewer cryptos will use Proof of Work. Unless you're Bitcoin/Ethereum, proof of work chains get attacked all the time. Look at Ethereum Classic.

How many years has etherium been putting off proof of stake now? Two? Three?

I gave up on crypto because of the failed promise of proof of stake in ethereum especially. We are talking 2015 here. AFAIR it’s been minimum 6 years. Tezos has been looking interesting though I never looked in detail anymore. OCaml based, COQ checked code base and truly proof-of-stake last time I checked? Did anyone play with it in here? I know Ubisoft jumped on the lets-create-artificial-scarcity-where-we-can NFT hype train for some DLC. A terrible idea on the whole which should never find traction but interesting to me again because it’s apparently built on greater Tezos “stack”.

Tezos has the largest grass roots adoption after ethereum for nft art. It’s very under the radar so I’m glad you asked. It has a burgeoning ecosystem and many users on ethereum are aware and have used tezos. As tezos ecosystem evolves, I believe it’s going siphon off a large amount of users from ethereum.

Proof of stake has been running for a year already. https://beaconcha.in

They are currently working on merging the v1 and v2 chains.

At this point, it feels like the corporate perpetual “13 months out“. i.e. never coming, always on the horizon.

this is a major update to a system that handles trillions in transactions, i'm happy they're taking their time on it.

tbh though very seriously likely this year, all the major infrastructure is in place

And, as ever, the exact same thing was said early last year, too. And the year before, if my memory serves me. I’ll have more faith that it’s actually coming once Star Citizen comes out.

Until it’s actually released, skepticism’s the safest bet.

That's not true. This time last year, the Beacon chain had only been running for a little more than a month. Also the code was not complete to merge Pow/PoS chains. The merge has already happened on Testnet, so literally the full merge could be done on Mainnet tomorrow if the Ethereum core team wanted to.

Speaking of betting,

What do prediction markets say about whether it will be out by what times?

There are forks of Ethereum already running PoS consensus (e.g. Binance Smart Chain, Songbird and an upcoming fork called PulseChain to name a few). I hear the difficult part is getting the sharding to work.

BSC isn't strictly a fork of Ethereum, though it is EVM compatible and uses the geth protocol. It also uses the Tendermint consensus protocol from Cosmos, which is different from the PoS implementation being worked on by Ethereum.

Sharding is definitely proving to be challenging for Ethereum, but my understanding is that it's being worked on independently of the move to PoS now.

The difficult part is figuring out why anyone would ever want to use any of this stuff for anything other than money laundering.

First POS was in 2012. Ethereum mentioned it in 2015.

So it's been 7 years since announcement.

Ever since I quit my job in finance - that was 4 years ago now.

As someone deeply interested and invested in crypto I strongly disagree. Proof of work has it's place in the ecosystem and isn't going anywhere. Plenty of modern chains are utilizing proof of work for various reasons.

We're seeing evidence for a preference for PoS already [1]. Proof of Work only makes sense if you'll have majority hash rate. Would you trust PoW in the case of Ethereum Classic?

1: https://coinmarketcap.com/view/pos/

Edit: Also I should highlight I'm not arguing to shift away from PoW where it makes sense.

> We're seeing evidence for a preference for PoS already

This isn't evidence of proof of work going away. I'm not sure why you think it is. More coins will be made with PoW and the ones that have PoW today likely will stay PoW. There's no reason to believe PoW is going to shrink, just because PoS is growing.

> Would you trust PoW in the case of Ethereum Classic

No, but I trust PoW with coins that have ASIC/FGPA resistant hashing algorithms. PoS isn't without trust issues either. What do you do if/when defi starts offering greater returns than staking? Obviously there will be a large exodus in staking making it more vulnerable to majority attacks as well.

Even if no existing coins switch to PoS (spoiler: many will), the PoS coins will suck up fee-paying users from the PoW coins, reducing the returns on mining, which will result in lower network hashpower. It may also bring a second-order effect of reducing the prices of PoW coins, which will reduce dollar-denominated block rewards on those networks, and that will bring hashpower down even farther.

ETH continues to grow its hashrate every day. Even in the face of growing competition, hashrate grows. It's an interesting theory, but there's no evidence of you being right. We already have many extremely high market-cap PoS coins, many with more features than ETH will have even after it's PoS transition, yet PoW hash continues to skyrocket.

Because ETH is still the king, and those coins haven't proven themselves yet. Next year ETH will be PoS and there will be no other major PoW coins outside of Bitcoin in the top 10. At which point, other PoS coins will lose their competitive advantage. So from a user and developer perspective, there's no reason to jump ship.

Proof of Stake implementation has been as never-ending as COVID restrictions.

Besides, it's not clear whether PoS won't create governance problems. PoW is a pretty smart solution of proving you spent some money and pays the miners so that people who have computing power won't attack the system.

PoS is trusting rich people, hoping they won't bring down everything for who knows what profit - a government printing money to delay a market crisis, is not that far.

Crypto currency is an instantiation of Bostrom's paperclip maximizer running on a hybrid human/machine topology.

Yes, very likely, although there's no AI involved and are the humans who are digging their own grave. I can easily picture a future in which 50% of the space available on the planet is allocated for energy production plants, the other 50% for mining farms, and there will be essentially only jobs related to these two activities available, just like using every resource to make paperclips. We started something that could make a nice Twilight Zone plot, and that's what worries me the most.

>therefore there's absolutely no way prices of hardware and energy will return to normal until governments won't stop this nonsense.

For all its faults, the market is great at solving these problems. Though I suppose it depends on what you mean by "normal".

The problem is that miners are buying GPUs. Without denying the problem, how can the market fix this?

GPUs are manufacturable. The market is meant to signal to manufacturers that this is profitable to make more of, and thus, more would become available in the future. Unfortunately, the current world events conspired to not allow this balance to happen - chip shortages and logistics problems etc.

Give it a few more years, and it will balance back out.

Crypto is carefully designed to model Parkinson's Law. (Proof of) Work always grows to fill the compute available.

Network difficulty adjusts to pull the demand curve to the right.

That makes the GPU market supply-constrained. Any increase in supply will be met with a lower price equilibrium which will result in more GPUs bought and used for mining. Which results in a higher THR which will be met with additional increases in network difficulty which will pull the demand curve to the right ad infinitum.

I'm not sure how much more economics 101 I can make this. As long as GPUs are part of the system, the GPU market will be supply-constrained. The supply and demand curves will continue to drift to the right.

There is not enough competition, people got conditioned to high prices for 10 years - once prices are raised they are not coming back.

Good luck starting your GPU company

> Good luck starting your GPU company

There is a chance you may mean that sarcastically, so...

Yes, if you think sale prices are going to stay much higher than manufacturing costs it is a great time to start a new GPU factory. Mass produce 5-10 year old tech in bulk, soak up enough of the market to survive and move as quickly as possible into modern manufacturing techniques.

We've seen in the last number of years that investors/lenders are perfectly happy to take multi-billion dollar losses of multiple years while chasing a market. It'd work fine.

> Mass produce 5-10 year old tech in bulk

I really wonder if the tech/equipment to do this is actually also unavailable - preventing any competition from starting up. I think it's the same logistical bottleneck that is hitting everything, not some monopoly conspiracy, and if that's the case, by the time bottlenecks clear up, the demand for such items would also drop (as the higher end stuff becomes available again).

> Yes, if you think sale prices are going to stay much higher than manufacturing costs it is a great time to start a new GPU factory.

So then as PoW difficulty increases, demand for GPS increases. Necessitating more factories until the primary output of world industry is GPU factories.

Does nobody see how absurd this is?

The market is stuck in a feedback loop. People are naievely looking at this and saying "Aha, just follow the rules of capitalism and the market and it will work out well. That's the religion." Not seeing this is a spiral. There is no market solution here.

It's like we're looking at a while loop, with condition "true == true", the body of the loop doubles the length of a string each time. Somewhere here is saying, this will not end well, we're gonna use up all of our heap trying to double the string size. And somebody else's solution is, "close some running programs so that ram can be allocated to the string doubling program".

Ok yah, we can run a bit longer, but the end result is still 70%+ of RAM being used to allocated these strings and eventually program crash.

Building more GPU factories isn't a solution is crypto's exponential consumption of resources.

> So then as PoW difficulty increases, demand for GPS increases. Necessitating more factories until the primary output of world industry is GPU factories.

it's not absurd. If you replace GPU with food, and PoW with population, then you will see that it is not absurd. It is only absurd because of the a priori assumption that having more PoW is worthless - an assumption you wouldn't make with population.

And i would think that if in the future, demand for PoW crypto grew, it means it must be satisfying a demand (for example, currency collapse). As long as a demand is being satisfied, and paid for by individual's earned money, it should be fine (as opposed to being paid for by somebody else's money - aka, externalization of cost).

> it's not absurd. If you replace GPU with food, and PoW with population, then you will see that it is not absurd. It is only absurd because of the a priori assumption that having more PoW is worthless - an assumption you wouldn't make with population.

> And i would think that if in the future, demand for PoW crypto grew, it means it must be satisfying a demand (for example, currency collapse). As long as a demand is being satisfied, and paid for by individual's earned money, it should be fine (as opposed to being paid for by somebody else's money - aka, externalization of cost).

I'm not even certain how to respond to this. I called out that it's becoming evident that the growing problem with crypto is a collection of people who fundamentally don't understand markets and society. That the understanding is as simplistic as "if someone is willing to pay money for it, then it must be good for society". Which is beyond naieve. And how people need a deeper understanding of what's happening.

And the response "countering" my argument is essentially saying I'm wrong because paraphrasing: "If someone is willing to pay for it, it is good for society". And equating GPU production with food production because in this hypothetical they are both things people pay for.

I'm not certain we can have a productive discussion. The goal of society isn't to create markets. Markets are a tool society uses to improve life for people. Markets aren't the goal.

Making legal murder for hire is not a net positive for society, even though you could create a market place for it and people would pay for it. Arguing something is good because people will pay for it is so beyond flawed that I don't have a way to describe it.

Historically, for many products, high demand leads to lower unit prices as manufacturers become more efficient. Since we're a decade into GPU mining, there's been a lot of time for manufacturers to adjust. While so far they have not been able to keep up with demand, I don't think it's clear that they never will.

Sorry, not following. Can you explain how efficiencies in GPU production interact with network difficulty to produce lower GPU prices? I'm not finding much in the literature.

What makes hash-based proof of work function is that to produce some number of hashes you need to use some amount of real-world resources. In practice, the two main resource inputs are electricity and computational hardware. If hardware is scarce, as it is now, the hardware will be a large proportion of mining cost, while the more efficiently hardware can be produced, the more of the cost is electricity.

How does that make miners not buy GPUs?

LHR was one attempt at it with varying success.

And it won’t take a few years to balance out. If ETH really does go PoS in the next six months then you’ll see a lot of RTX 3080 and 3090 come on the market for very cheap.

A lot of ETH miners say they’ll go to RVN or others but there is no way other coins can take the hash rate currently devoted to ETH.

> The problem is that miners are buying GPUs.

That's not a real problem, I can stop buying GPUs and keep playing most 3-5 years old games with a decent integrated one in a relatively modern CPU at a fraction of the price, and that's it. I'm not a gamer however, so I would understand hardcore gamers to be pissed off, but that's not a real problem as nothing vital is taken away from anyone.

The real problem is the steadily raising demand in energy which doesn't seem to stop anywhere in the future. If we accept the fact that more energy translates into more money, and that more money allows miners to add more and more GPUs (or ASIC dedicated hardware, energy-wise, a cluster of GPUs and a box that draws 3000 watts are exactly the same thing) then we're going straight to a future in which every resource will be used to produce energy.

How to stop something that screams "put more energy in = get more money out"?

No way, it's probably too late, and I expect a huge misinformation campaign from the parties involved, which already amassed a lot of money to pay for it. Therefore we'll see a stronger push for nuclear and other non clean sources, but also soaring prices of every other source, including clean ones, because the demand dictates the price, and there is where most demand is coming from.

And there's also the warming problem, because you know, consuming so much energy doesn't come without a price to pay.

Reply on news.ycombinator.com