Hacker News Reader

Comment by 013a | Hacker News Reader

Its bad. Its never been this bad before. Just go on pcpartspicker and realize that there are GTX 1080 Tis selling on third party websites for over $1500, and that's the only place you can get them beyond getting lucky.

Nvidia isn't going to ramp up production significantly because crypto could crash any day, simultaneously destroying their new sales and flooding the used market with old mining hardware. Combine it with the industry-wide RAM shortage... What a disaster.

The funny thing is, I remember back when ASICs first started coming out for Bitcoin, and a whole bunch of new cryptos came out saying "we're specifically designed to thwart ASICs so everyone can mine on the cards they have!" Well, here's what you've created. No one can afford cards anymore, and you made making more efficient hardware harder. Thanks for destroying the planet.

View on HN

ASICs aren't saving the planet either - people are incentivized to burn up to $BLOCK_REWARD of electricity per block, the number of hashes it takes to do so is irrelevant. If you come up with a better ASIC, people will buy more of them, the network difficulty increases, and you're back where you started. It's just another stage in the Red Queen's Race.

The only thing ASICs do is centralize control (because economies of scale favor companies who can produce+operate their own chips). Full stop.

Proof of stake doesn't actually work either, or at least not without a centralized master key to prevent stake grinding/rewriting history (at which point why have staking in the first place). And in many cases, it may devolve to proof of work anyway.

There is no efficient cryptocurrency, and the cat's out of the bag now, there is nothing that can (practically) be done to stop it. Apart from charging people exponentially-increasing rates on their utility bills, that is, but China isn't going to do that, so the planet is just going to have to take another one for the team.

Is Cryptocurrency a form of entropy exchange from Energy (increasing entropy via heat) to information entropy (decreasing information entropy by hashing)?

I am thinking more broadly in terms of thermodynamics. I understand classical Thermodynanamics but I don't understand information entropy well.

From the standpoint of physics, Energy is the ultimate currency and it seems like we are trading it to decrease information entropy. Typically, entropy is the tax when performing work (for e.g in a heat engine) if that process is irreversible. I wonder if this kind of theory applies to Cryptocurrencies?

Proof of work is about consensus among mutually distrusting parties. It has approximately nothing to do with entropy.

Of course there is some connection, but the simple "exchange" you're looking for does not exist.

>> Proof of work is about consensus among mutually distrusting parties. It has approximately nothing to do with entropy.

Think more abstractly - at the end of the day, Energy is being consumed to compute some problem. That solution has intrinsic value precisely because it took energy to solve it. Extrinsic value is the market economy of CC, and that's where the 'proof of work' and 'consensus amongst distrusting parties' comes into play - that concept of a currency is only valuable if sufficient number of parties agree to its value. I am weeding into the underlying fabric that holds Crytocurrency together.

Imagine if Energy was free, despite of 'consensus', the supply of such currency would be virtually unlimited and it would be worthless.

None of that makes entropy any more relevant.

Energy is consumed and that is why there is a cost. I'm not sure there is any such thing as "intrinsic value", but there is a cost.

Because of that cost, the larger system works, because that is how the system keeps itself honest, by making it expensive to lie.

But there are other ways to have an honest system that don't involve wasting energy, so there's no direct connection between the energy that's wasted and any value the system has.

If energy was free, silicon would be more expensive, mining difficulty would go up, and bitcoin would still "work" just like it does today.

> Imagine if Energy was free, despite of 'consensus', the supply of such currency would be virtually unlimited and it would be worthless.

They would just ramp up the network difficulty. Soon mining would be counted in exhausted stars per second.

Cryptocurrencies are the perfect example of the maxim that "just because you can, doesn't mean you should". They're funny in theory, proof of work being an interesting solution to a particular mathematical problem. In the real world, with problems determined by laws of physics and (in consequence) the structure of human minds, this is about the worst imaginable way to run the economy. The constraints of the problem solved by cryptocurrencies are not very meaningful in reality.

> It has approximately nothing to do with entropy.

Err.. it's removing surprisal from bit sequences. That's exactly what Shannon entropy is. There's even a boundary trade-off (which we're obviously trading well, well above) between Boltzmann and Shannon on this.

Yes, that's the "work". The value of that work is as part of a mechanism for enforcing consensus. To the question "does entropy tell us anything about the larger system" I am arguing that it does not (in any non-trivial way).

Obviously entropy is related to any computation we perform. It's just not going to give us any special insight into cryptocurrencies to point that out.

Though there is some benefit to cryptocurrencies that rely on ASICs for mining: it means that someone with a significant amount of CPUs/GPUs under their control (such as botnet operators, or government groups like the NSA) can't waltz in and trivially 51% attack the blockchain with their existing general-purpose hardware, and then go back to using it for other things. An attacker would have to commit to buying/building specialized mining hardware for the attack, and then that hardware can't be repurposed for anything else after the attack other than mining.

ASICs are supposed to be orders of magnitude more efficient at their job than GPUs. So NSA can make a 0.01x the investment and "waltz in and trivially 51% attack the blockchain", all while not touching their existing equipment.

Only if they've developed an ASIC that is orders of magnitude more efficient than what anyone else has - which seems unlikely given the enormous economic incentive to find more efficient hardware.

The only reason to invoke NSA in this context is for their ability to make a 10000x investment, with no concern for this being (directly) economically viable. The question then is, what will they do with control of the blockchain?

Idea: prevent North Korea from spending their coins?

I'm not sure if you understood my post right. Bitcoin mining relies on ASICs. If the NSA wanted to do a 51% attack, then they'd have to spend a bunch of money building ASICs that are then worthless to them once they're done with their attack.

If Bitcoin mining relied on GPU or CPU computation, then the NSA could just redirect some of their existing GPU/CPU hardware to do a 51% attack, and then go back to using it for something else when they were done. It would be cheaper and easier for them to do.

Unrelated - did you used to frequent a Halo tricking forum called HIH?

Yeah, that's me. I stuck around its game-modding subforums a lot back then.

I did, too - not sure if you’d remember me, but I used to go by Zanzang. I went from modding games to programming them, then from the game industry to startups.

Funny how 10 (plus?) years later we’d run into each other on a different site.

A huge amount of the electricity being used was unused/not-competed-for in the first place. Miner farms are coalescing in Northern America around excess production in hydroelectric areas. This is an inconvenient fact that is ignored in the "electricity oh nooooo" debate.

Except excess hydroelectric production is extremely valuable to other utilities on the interconnection who run primarily thermal plants.

Hydro is one of the few forms of baseload power that can be ramped in a hurry: you can increase or decrease production just by increasing or reducing flow through the turbine. If you use less water, it builds up in the reservoir as potential energy and you can use it later.

Compare this to thermal plants, which require many hours to have a significant swing in energy output. I have heard of ramp figures on the order of days for very large nuclear reactors.

Because thermal stations run with pretty flat power output, hydroelectric stations have a critical role smoothing out load conditions, ramping down at night and up during the day. Since electricity is traded on a market system, utilities with significant hydro operations really benefit from this: they can reduce production and import excess capacity from thermal plants at night, and increase production during the day to export at a much higher rate.

This is part of why regions with predominantly hydro power have such low rates: in addition to hydroelectric being pretty cheap once a dam is built, hydro operators make significant amounts of money by appropriate import/export to thermal operators. Since a lot of hydro utilities are state owned (eg: BC Hydro, Hydro Quebec, Bonneville Power Administration), the profits are used to subsidize lower rates for ratepayers.

I had no idea I was so lucky on my timing. It's hard to believe they're going for 1200-1500 right now! I bought my son a 1080Ti for christmas, and paid about 1/2 of what they're going for now.

Destroying the planet = upset some gamers?

Reply on news.ycombinator.com