Hacker News new | past | comments | ask | show | jobs | submit | Negitivefrags's comments login

None of that matters if you already bought the H100 and have no use for it. You might as well recoup as much money as you can on it.

> You might as well recoup as much money as you can on it.

Depending on how fast their value depreciates, selling them might recoup more money then renting them away. And being exposed to 3y of various risks.

Selling now at a 40% loss gets you back the equivalent of 60c/h over three years, and without having other costs (DC, power, network, security) and risks.


If you already have the H100s, renting access to them at a loss isn't better. Throwing them in the trash will lose you less money.

That's not how this works.

Imagine I own a factory, and I've just spent $50k on a widget-making machine. The machine has a useful life of 25,000 widgets.

In addition to the cost of the machine, each widget needs $0.20 of raw materials and operator time. So $5k over the life of the machine - if I choose to run the machine.

But it turns out the widget-making machine was a bad investment. The market price of widgets is now only $2.

If I throw the machine in the trash on day 1 without having produced a single widget, I've spent $50k and earned $0 so I've lost $50k.

If I buy $5k of raw materials and produce 25k widgets which sell for $50k, I've spent $55k and earned $50k so I've lost $5k. It's still a loss, sure, but a much smaller one.


The concept you're looking for is "marginal cost". The initial $50,000 for the machine has already been spent - the only calculation left is that each new widget costs 20 cents to make (that's the marginal cost) and generates $2.00 in revenue. At this point, making widgets is highly profitable.

and for GPUs, the math is even more stark because rather than having a 25k item lifespan, the lifespan is the time until GPUs improve enough to make the current one irrelevant.

GGP already showed the marginal power cost is well below $2.

There is so much more to lifecycle sustainment cost than that.

Rackspace. Networking. Physical safety. Physical security. Sales staff. Support staff. Legal. Finance. HR. Support staff for those folks.

That’s just off the top of my head. Sitting down for a couple days at the very least, like a business should, would likely reveal significant depths that $2 won’t cover.


These are all costs of any server hosting business. Other commenters have already shown that $2/hr for a racked 1U server at 400W is perfectly sustainable.

Just because you have all of those costs already doesn't make them go away. If you're cross-subsidising the H100 access with the rest of a profitable business, that's a choice you can make, but it doesn't mean it's suddenly profitable at $2: you still need the profitable rest of the business in order to lose money here.

So you terminate all of the above right now, or continue selling at a loss (which still extends the runway) and wait for better times? Also, do you know that similar situations occasionally occur in pretty much any market out there?

The market doesn't care how much you're losing, it will set a price and it's up to you to take it, or leave it.


No, if its only a “loss” due to counting amortization of the sunk cost of initial acquisition, throwing them in the trash will lose you more money. The only way you can avoid the key cost is to travel back in time and not buy them, and, yeah, if you can do that instead, maybe you should (but, the time travel technology will make you more money than the H100s would ever cost, so maybe don't bother.)

The main concern with this is how do you actually get the records of what the cost basis was from someone who is dead?

If the issue is that people are dying leaving behind significant wealth but not documenting this, just make the estate tax 100% on any assets missing documentation like this. I'm sure the lawyers would figure out the rest.

That's effectively already the rule if you sell something and can't figure out the cost basis - it counts as zero.

That isn't really the main concern. It's really a question of alienability.

If your great grandfather invested in something a hundred years ago and now 99% of its value is appreciation (or inflation), you may or may not want to continue investing in it. If you do, the step up in basis doesn't really matter because you're not going to sell it anyway.

But if you now think it's a mediocre investment, you may be inclined to sell it and invest in something else. Except that you won't if you'd lose a significant proportion of its value to taxes. This is a problem with capital gains taxes in general, but it's especially a problem for anything held intergenerationally (i.e. for a very long time) because not only will the appreciation be large, the inflation by itself would represent most of the value of the "gain". So the step-up in basis is a stupid hack to avoid this and let children make different choices than their parents and grandparents without being punished by the tax code.

There are probably better ways to handle this, but "delete it and replace it with nothing" is not one of them.


> There are probably better ways to handle this, but "delete it and replace it with nothing" is not one of them.

Why not? Why do I care about someone being deprived of a portion of some investment his great-grandfather made?

If I get money from some relative who invested in stuff and then you get money from working really hard in a way that someone thought valuable so they gave you money for your work, why should you pay taxes on that money while I don't pay taxes on the money I got from my dead relative?

Are we trying to incentivize people to be born to families that already have money or something? Like are we afraid that if we don't do this, we'll be creating incentives for people to get born into poor families instead?


> Why do I care about someone being deprived of a portion of some investment his great-grandfather made?

Because they only get deprived of it if they sell it, so that gives them more incentive not to sell it, but selling it may be more economically productive, and then you lose the positive externalities of the more productive investment and the tax revenue it would have generated, which could by itself plausibly be more than the loss from the step up in basis.

In general the problem is that capital gains taxes when implemented simplistically create a lot of perverse incentives (tax on productive investment is economically undesirable in general and some of the edge cases are especially ugly), and then the tax code gets full of warts that try to reduce the bad incentives/consequences instead of rethinking the structure of the tax.

> If I get money from some relative who invested in stuff and then you get money from working really hard in a way that someone thought valuable so they gave you money for your work, why should you pay taxes on that money while I don't pay taxes on the money I got from my dead relative?

Your dead relative already paid the taxes on any money earned in the equivalent way. Capital gains are on asset appreciation, which is an industrial-sized can of worms.


Brokers have been required to track costs basis information since 2011. That doesnt really help for assets purchased before then, so estate executors would need to find records for transactions before then. The IRS will generally assume a costs basis of zero until proven otherwise.

The usual way this happens is accidentally logging passwords. Or even other cases where passwords happen to be included in something else. It can happen more easily than you think.

Like for example, if you collect server side crash dumps, are you really taking care that there is no sensitive information sitting in the memory image stored in them?


the word you should use is carelessly, as in carelessly logging passwords.

When working with data that you can reasonably expect to contain secrets, you should behave as if it does contain secrets. It worries me that you mention you're aware that server side crash dumps may contain sensitive data, but you also speak as if it's reasonable to not protect them knowing they do, or they might. I'd hope or expect anyone would mention or at least imply that it'd be negligent to behave so recklessly with someone else's secrets.


What you're saying makes sense for a small startup with 20 people. Have you worked in an org of 10000+ that's been around for 30 years? Where the people who built the systems no longer work there? Where the people who maintain them are 3+ career cycles removed? Things get so baked in "one doesn't simply make changes" because no one person understands how everything works, so you do your best and keep closing tickets. "Hey we should really investigate whether or not passwords are included in memory in our 2000 severs across the globe!" "Lol ok maybe when you close out the 30 tickets on your plate."


> Have you worked in an org of 10000+ that's been around for 30 years?

yes.

Would you still make this argument for something else known to be dangerous?

Hey this gigantic lathe doesn't have a safety shut off switch.

Ahh yeah the guy that did the wiring for that left years ago so we just don't touch it. It's fine as long as no one puts their hands near it when it's running.

Hey, isn't this freshly dug 20 foot hole supposed to have an escape route, and side wall reinforcements so it doesn't collapse on someone? lol ok new guy, let me know when you've finished pouring that concrete and then we'll look at that idea.

To go back to the example; Taking steps to protect them could be as simple as restricting who can access core dumps, enforcing they don't get stored unencrypted, and the only people who can copy and inspect them have are already in a trusted role where they can get root on that production server. Or an even better option would be restricting the service that can read clear text passwords. These machines don't crash, and when they do, we don't write a coredump. (but this trusted team can start if we see the system crashing suddenly)

> Things get so baked in "one doesn't simply make changes" because no one person understands how everything works, so you do your best and keep closing tickets.

This is a super disappointing attitude. It's hard, and it's the way we've always done it, so that means it's impossible, or not worth the effort? Yeah, I'm not likely to buy into that mentality. I believe I can fix things and have a positive impact. It's one thing to say it's impossible because you don't understand how to do it, that's wrong, but I guess if that's what you need to believe to sleep at night, I can pretend to understand.

It's another thing entirely if you don't have the autonomy at work to try to improve something you know to be broken, and a risk to users, and the company. If that's actually what you meant to describe, you might wanna consider if you can find another job. You're clearly not an idiot, and there's plenty of companies that wont treat you like a code monkey.


I can’t view a video at the moment, so I apologise if this has been answered in that.

If the server software a game uses requires a licence to a third party library, what is the developer expected to do about that?


If the EU decides to build any legislation around SKG, they will give developers ample warning before things go into effect. Apple had years to prepare for USB C. This is still years out. So the answer is: negotiate compatible terms, or don't use the third party library.

This is merely an issue to begin with for companies that are absolutely massive, like Sony or Activision. Smaller developers just don't do stuff like that in general: you download the game and then you have the game.

Since the ask is for a reasonably working game, maybe as a developer in that position you can just cut out the functionality that depends on the library or replace the library with something similar or mock it out or use a static cache of request vs response for all possible requests. The technological possibilities are endless.

It's not like as developers we're these helpless infants who have never solved a problem in our lives. It's a tech problem, tech a solution to it, that's why you're a professional and not bush league.

Ultimately if someone can't figure out how to do their business without scamming people out of ownership then that's a skill issue. If they're not creative enough to figure it out, the business is doomed to begin with. Legislation often has the additional positive effect of ridding the market of people who shouldn't be there to begin with, like food trucks infested with cockroaches and pizza places that use fake cheese.


What I don’t like about this argument is that this was originally pitched in the initial video as “All you have to do is release the server you already have, you don’t have to do any extra work”.

And I took that to mean that you just have to provide the server binaries and no support for them.

And fair enough, as a developer of a large online game myself, could get behind that.

But the moment this extends to needing to find solutions for people to be able to actually run it, I would withdraw my support.


sorry, if you can't sell your game without stealing it back 10 years later then you shouldn't be selling it in the first place

> It's a tech problem, tech a solution to it,

Except it's not - it's a business problem. SKG would essentially ban the use of Oracle as an example. Or it would likely kill games like Rock band which have licensed audio. You might be ok with that, but why are your preferences more important than mine.

> This is merely an issue to begin with for companies that are absolutely massive, like Sony or Activision. Smaller developers just don't do stuff like that in general

This is a naive viewpoint IMO. Another way of looking at it is that only large companies will be able to conform and this will squeeze out the possibility of small developers having multiplayer games. This sort of red-tape stifles innovation.


Small developers have multiplayer games all the time that aren't affected by issues like the ones SKG is concerned with. Ad-hoc multiplayer and dedicated servers that players can self-host are long-established solutions. A common argument in bad faith is that SKG demands perpetual upkeep of presumed infrastructure that will somehow harm small developers, and it just isn't true.


To be fair they didn't make that argument, but thanks for the support none the less! Agreed, ad-hoc servers are a staple and a worked out problem. 99% of the time you have to go out of your way as a developer to make things not work in this way.


> Except it's not - it's a business problem. SKG would essentially ban the use of Oracle as an example

why are you using Oracle for video games? what's wrong with you?

> Or it would likely kill games like Rock band which have licensed audio.

Rock Band DOES work offline. Licensed audio in Rock Band is licensed in such a way that once a copy is sold the license allows the use of that copy in perpetuity. 100 years from now I'll still be able to pop in my Rock Band disc and play it, because that's how ownership works and the developer didn't get in the way of my ownership of my own property.

But when I said "It's a tech problem" I was answering someone who mentioned a tech problem.

Coming up with a different, non-tech problem as a counter-point to a whole discussion exclusively about a tech problem is not as smart as you think it is, and the examples you bring up aren't very good at all.

> Another way of looking at it is that only large companies will be able to conform

No, that's unmitigated nonsense. Just your first paragraph showed you have no idea what you're talking about, but now you're just stringing words together. The reason why only the largest companies can have these problems in the first place is because of their legacy technology integrations and pre-existing technology supplier agreements which they would have to re-negotiate. Remember, this is a scenario AFTER the initiative gets a million signatures, which is a year out, and AFTER the EU has legislated, which is another year at least, and AFTER the warning period which is several years. Even with the fastest possible timeline it's probably like 5 years of warning that things are going to change. And at that point anyone entering the space from the bottom as a new player is free to negotiate a deal which conforms with the market regulations going forward; if the technology suppliers don't want to negotiate realistic terms, they go out of business. While we're at it, large companies are also free to renegotiate their contracts to make them legal in the eyes of the legislation because contract survival terms are a standard staple in any technology supply agreement and if changes to market regulations make a contract unfit or illegal then renegotiations commence as a matter of course. But given the timeline of this going into effect they'll have renegotiated YEARS ahead of the deadlines.

This isn't a twitter poll. It's not going to go into effect 5 minutes after it's been posted. There will be AMPLE time for everyone to figure stuff out and change their paperwork, and the only companies really affected are the ones that already spend $1M+/year on legal anyways.


> why are you using Oracle for video games? what's wrong with you?

Because my previous project used it. It's an example. There are plenty of others. And I think this sort of attitude is unfair towards people like me who genuinely want to preserve video games, but are concerned that an ideological battle is going to negatively affect the industry.

> Licensed audio is licensed in such a way that once a copy is sold the license allows the use of that copy in perpetuity.

Licensed audio can be licensed in such a way. GTA being a great example of something that doesn't have perpetual licenses to their music.

> Coming up with a different, non-tech problem as a counter-point to a whole discussion exclusively about a tech problem is not as smart as you think it is

> no, that's unmitigated nonsense. Just your first paragraph showed you have no idea what you're talking about, but now you're just stringing words together.

In three paragraphs, you've attacked me three times, when there's no need to have done. If you can't have civil discourse, I'm not interested in discussing this with you.


> It's an example

no, it's not. An example is something that happens. What you brought up is a fantasy.

> GTA being a great example of something that doesn't have perpetual licenses to their music.

that's wrong, because even if Rockstar removed some songs from some versions of their game, if you bought the disc version of the game, then guess what - the songs are still on there.

it is neither the consumers' nor SKG's fault that the richest company in the richest entertainment industry is unwilling to negotiate terms that don't scam the people purchasing their products.

> you've attacked me three times

if you're going to make stuff up and bring up things that don't stand up to the simplest scrutiny then that's going to be brought up. that's not an attack on you, but it definitely is a comment on the quality of points you bring to the discussion. if you want to make better points, it's as simple as: before typing "X" google for "X?" and then read the top result. otherwise it's just whataboutism.


The concept of doing something at runtime that could be done at compile time is anathema for c++ programmers.


In theory ReFS is the replacement. It hasn’t seen large uptake though.


People were saying things like "Why does twitter need 10,000 employees?" for years and years on end.

We are seeing a reversion to the mean.


Turns out it didn‘t. Microsoft is beginning to get the same idea.

The idea that you didn‘t really need a large overhead of management, DEI employees and fairness committees was true all along.


Most large tech companies especially those with entrenched monopolies(FAANGS & Friends) don't need the huge number of workers they've amassed in order to achieve the same product performance and profitability, but they over hired during the years of tech hype and zero interest rates, and now it's difficult to know which people are those pulling most of the weight in the org and which are disposable and just cashing in while only engaging in empire building and theatrics to look busy and important, so the layoffs tend to be more or less random, but it also doesn't affect their profitability even if they let some of the valuable people go through the random layoffs, since like I said, they're entrenched monopolies.


> The idea that you didn‘t really need a large overhead of management, DEI employees and fairness committees was true all along.

Yes, and it's also true that you don't really need two lungs. That doesn't make it a good idea to have one removed.


The metaphor fits best if you make it an appendix. It may or may not provide some unspecified benefits, but there's marginal difference of having it removed. Sometimes it swells to an uncomfortable size, which is when you're best off cutting it out before it kills you.


It’s worse in a lot of implementations because often SMS is often used as part of a recovery flow in cases where you lose the first factor.

I find it more secure in some contexts to never give a company my phone number at all if possible, so that it simply can’t be used as any kind of authentication no matter what.


Yeah, I'd draw a hard line between "SMS 2FA is better than no 2FA" and "SMS should never become a single-factor recovery method."

I agree SMS should never be an option for single-factor recovery.


If we can’t even reliably know what the speed limit is, how can we even be close to self driving cars?

Not only is it a way simpler problem, but it’s actually a requirement for self driving also.


That. I am thus convinced we are decades away from self driving cars capable of taking you home via a dirt road. And unfortunately, the entire value proposition of self driving cars only unfolds if you really can go anywhere, like a cab driver could.

“Sorry m’am, I can go any any further on this road, please get out and carry your luggage over the remaining mile”. And then of course it’d be raining under the scorching sun, I guarantee you that.


I think most databases have some method of turning your Reads into Write locks without needing to actually do a write.

Like SELECT FOR UPDATE.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: