Thoughts on a “HDD shortage” from an industry insider
reddit.com
reddit.com
Now Chia wants to eat all the hard disks, and I am pissed. I declare war on hegemonizing proof-of-waste swarms. Burn the coins. I want that disk space back.
https://www.fool.com/investing/general/2013/12/11/amds-graph...
Anyone curious can verify that google translate doesn’t even understand “il chia” and google translate is quite good on French.
Someone cares enough to have compiled the conjugation: https://la-conjugaison.nouvelobs.com/du/verbe/chier.php
and it's a really old word, used frequently by Rabelais, for instance.
Not to say it won't grow to become a problem but even at 1EB of storage, it's not a problem:
> Data center storage growth is round 2-2.5 EB/day. The 1.4 EB used by Chia so far is less than one day's data center growth.
I'm not defending yet another wasteful use of cryptocurrency ... first it was the GPUs now its the SSDs.
* lower price points ($/GB);
* speed;
* size;
* availability / supply.
Of course, it will suck in the short term if HDDs become much harder to find, but that should only last as long as the manufacturers believe this to be a short term trend.
The demand from the cryptocurrency users will be extremely high until they've spent a particular amount in total. They can do that with 1000 units, or with 5000 units.
Obviously auctioning off units would be bad for having predictable prices for retail customers but that is irrelevant if aggressive miner buying is buying up all the supply.
Moreover, shareholders at the manufacturers should be demanding it: they're leaving money on the table.
Also, remember when Martin Shkreli increased the price of an anti-parasitic drug by 5000%? That upset a lot of people, and arguably it's what put him in jail. Of course, HDDs are not quite the same as vital medicine.
Yeah it’s bad for regular customers, but if there’s no supply then what would they be able to get anyway?
But for them, availability will also be limited, and next-term prices will get higher.
Also, increasing production of a highly-demanded item is not unheard of! But it will take months.
Auction sucks because it drives the price of goods up, but you know what's worse than prices going up? Prices going to infinity because they're out of stock.
Should the community have divested from Nvidia's proprietary format and spread its resources on multiple platforms? Concentrating effort can benefit everyone, but it requires trust.
In addition, part of this community investment was funded by government research grants, adding a legal/public policy aspect to the issue. Should taxpayer money be spent on creating private, abusive, monopolies?
The alternative is that the product is completely unavailable. This is currently the situation across a broad swath of GPUs. Every unit that becomes available is instantly snatched up, in some cases by people who've bribed the people doing stocking at the distributors.
That is asinine. Yes, auctioning the limited supply off is also obnoxious but it is a strictly no worse than being entirely unavailable. Auctioning it of also has the benefit of depleting the uncontrolled demand faster (which is often taking the form of 'buy $x million in parts, however much we can get'). It potentially helps expand the supply (though unfortunately lead times and tooling make it hard to expand supply for surge loads), and for some cases where paying an astronomical price would be justified it would preserve some access to the product.
But the main motivation for my suggestion is to deplete the resource of the parties buying 100% of the stock as quickly as possibly so that the market can get back to normal. If that doesn't work it's because the money fountain is simply too large, and at some point the increased revenue will justify increasing the supply (even though doing so is risky).
> Business models evolve toward efficiency, and there is a good reason no one is doing this.
Sorry, but that isn't how 'evolve towards' works. Sure, eventually. But this kind of demand is unusual and surprising, it presents new challenges. There may be good reasons why what I suggest wouldn't be a good idea but those reasons are not "if this were a good idea they'd already be doing it".
I've worked with Big Retail execs before. They're old school and conservative (in the business sense).
In a smoke filled CES hotel room they shook hands on a deal of X for $Y, and that's how it's going to stay for the next decade.
2) Risk. These shortages and cryptocurrency demand spikes are a blip on the history of these companies. It'd be stupid to suddenly change their entire business models and retail relationships for something that could disappear next month.
As much as I believe cryptocurrency and hence mining is here to stay, I can't fault these companies for not believing the same and acting accordingly. Especially given the cycles of boom and bust that cryptocurrencies have gone through.
Do you not know how professional journalism functions? You can verify decades of history with very limited retractions. There's good reason why people trust the New York Times.
There is the whole Gell-Mann amnesia thing. There is the basic selection of which news is important and which facts are relevant for it. There is the bias in reporting. Each of these can make an article untrustworthy without needing any retraction.
What should I read to prepare myself to take advantage of the cheap used high end hardware that will hit the market do build a Windows or Linux system for my own pleasure.
Your assuming that proves a point, but it doesn't.
>I've never built a PC.
Definitely worthwhile. Basically lego for adults. Tons of good info here
No guarantee, but it’s more of an anti signal.
"Jump on the bandwagon" Meaning: to support a cause only because it is popular to do so.
Sounds like a bubble. I encourage more people to jump on the bandwagon. I'm not in a hurry to get my new computer and the bigger the bubble gets the more cheap hardware will be available when it pops.
https://www.ecenglish.com/learnenglish/lessons/why-do-we-say...
Capacitors notably derate exponentially with increasing temperature as do many other things.
The 3080 and 3090 are the singular exception. The GDDR6X runs extremely hot and frankly I think it's going to cause a bunch of early failures for even non-mining cards.
(I wouldn't necessarily trust parts that have been worked to death btw)
That's why crypto isn't going anywhere and this will only get worse.
I don't understand what you're trying to say. To me this sounds like political rhetoric not an argument.
Are you saying that cryptocurrency can replace central banks? Are you saying that cryptocurrency is some kind of revenge or punishment against central banks and wealthy? Are you saying something else??
> The Times 03/Jan/2009 Chancellor on brink of second bailout for banks
That aside, to those of you that are scanning the comments here without having actually read the OP's reddit link - I'd urge you to read the linked page.
It is an incredibly insightful summary of the HDD supply chain.
(Whether it was indeed written by an "industry insider" or not is irrelevant; their analysis of the HDD market is highly informative)
I am not sure what grad student Deep Learning researchers are going to do?
It's really not much of a stretch to see HDs go to exactly that same pressure point.
I suppose getting a card in July in a prebuilt is still better than not getting a card on its own.
While I was digging around, it looks like some of the laptops with 30 series in them are even at the "ETA: One day soon!" point.
If you only had space for 1 drive, you did have to choose between small capacity with high speed vs. High capacity and slower speeds.
Plus (real or mythical) reliability issues with new tech.
The narrative was "for 300 bucks you can make your computer so much faster, best investment ever". No mention about increasing HDD prices, nobody cared.
Is it possible SSD or RAM have a similar bottleneck like read/write heads? Yes, but I don't know the answer.
Yes. You could replace "read/write heads" with "chips" and it'd apply just as easily to SSD and RAM. Texas's freak weather shut down Samsung's Texas fab which caused a shortage of SSD controller chips, for example. After the February outage, I don't think they started to tape out again until April.
Contracts are set up in similar ways, but HDDs in general have more constraints baked in those contracts.
Since it was due, in the last couple of weeks I was scanning prices in Romania daily, in case I get lucky with a discount. In the last week, prices surged 60-100%, and stock disappeared for 4 TB+.
I panicked and after a lot of scanning I managed to buy a slightly smaller one than I needed with just a 20% surcharge.
Prices on Amazon UK are also out of control.
So now we are out of chips, CPUs, GPUs and hard drives. What a time to be alive :)
PS: I would love to hear Backblaze take on this shortage.
It's the perfect time to set up manufacturing capability, so nationalists can be happy.
It's a good time to push for hire wages, workers should do it.
If there is some alternative industrial process, now's the time to try it.
This is no 70s oil moment, because there's no indication fewer widgets are being produced than prepandemic.
>If there is some alternative industrial process, now's the time to try it.
wouldn't this take years?
>It's a good time to push for hire wages, workers should do it.
This is just speculation on my part, but I suspect hard drive manufacturing is much more labor intensive than capital intensive. Therefore I doubt this will translate into higher wages.
Who decided that all computing must be for a sustainable or meaningful purpose? Cryptocurrency, as wasteful as it might appear, is like a drop in the ocean compared to the data centers of Google, Facebook, Microsoft, etc. Likes on Facebook provide much less value to society than a distributed currency that circumvents central control over the financial system.
Second, Crypto doesn’t exist in a vacuum. How much “energy” does it take to disassemble a mountain and suck out precious metals or diamonds? How much energy does it take to manufacture all the coins and paper money the world over? And how many innocent people have been gunned down for the petrodollar or some other dispute over fiat currency?
[1] https://www.visualcapitalist.com/visualizing-the-power-consu...
[2] https://www.deccanherald.com/business/why-bitcoin-uses-10-ti...
It isn't, as 'mytherin points out.
> Likes on Facebook provide much less value to society
That's absurdly untrue. For all the warts of big social media players, they still provide tremendous social value - rich long-distance communication, organizing ad-hoc groups of people (including, in particular, meatspace ones), improved exchange of news worldwide, making things easier for local commerce, ... I could go on. And while these services can, and are, being misused, they are still at least doing something useful.
> than a distributed currency that circumvents central control over the financial system.
Except it doesn't. But even if it did, this isn't necessarily a good idea in the first place.
> How much “energy” does it take to disassemble a mountain and suck out precious metals or diamonds?
Less than it takes to run Bitcoin, and it's also a one-time expense.
> How much energy does it take to manufacture all the coins and paper money the world over?
Not that much on an ongoing basis. It's also O(n) with respect to the amount of extra money you need in circulation. A trivial cost compared to what the upkeep would be if we tried to run the world economy on Bitcoin.
> And how many innocent people have been gunned down for the petrodollar or some other dispute over fiat currency?
People kill other people for profit. That's a fact of life, independent of the way money is represented. That they don't kill each other over crypto says only one thing: that right now, cryptocurrencies are completely irrelevant to the economy. If crypto ever sees any serious adoption, a proportional body count will follow.
(Also, I imagine the number for Bitcoin is already greater than zero, given that cryptocurrencies are disproportionally popular with criminals compared to general population.)
This is a fascinating insight I had not considered.
I think this isn't quite correct, but is even more in favor of fiat currency. Minting money costs in proportion to the number of bills/coins. If more value is needed, the denomination being printed can be changed to represent more value with the same cost.
On the other hand, cryptocurrencies require wasting resources proportional to the total value represented by them. Any less and they are vulnerable to attacks. Therefore, (as I'm sure will come as a shock to nobody) this is another way that cryptocurrencies scale worse than fiat.
Supplies of most fiat currencies are effectively infinite, and the currencies depreciate in general, see virtually anything from USD to the currency of Zimbabwe.
Bitcoin requires energy expenditure proportional to the value represented by bitcoin. Anything less, and it is vulnerable to attack. Bitcoin fundamentally does not scale.
I do think there is a lot of value in cryptocurrency research. The theoretical insights gained into ways of provably ensuring consensus in an untrusted environment, or tamper-proofing information, are already useful.
But that could all be done in a research setting (as it mostly already is) - we don't need every piece of such research to be immediately deployed in the form of a get rich quick scheme. So I'd strongly support an official, government-backed moratorium on the use of all cryptocurrencies, until researchers figure out a way to make one that isn't an exponentially growing, insatiable power sink.
We can test out the research incrementally in form of systems that don't pay people to grow chains. Like some of the project ideas I've seen thrown around big, non-financial companies - e.g. making it easier to ensure supply chain integrity. While in general I'm strongly suspicious of anything that even mentions "blockchain", some of these ideas don't look that stupid (but then, I'm not a domain expert in these areas, so the blockchain-backed solutions may all just be hot air).
https://www.eenewseurope.com/news/first-seven-customers-5nm-...
Are you sure? Bitmain is competing for 5nm wafers against the largest electronics manufacturers in the world.
In my case I was having my eyes on 5 TB 2.5" drives. I don't need them, I'd just maybe like them, so I haven't pulled the trigger, but I have them set aside on Amazon France. The price has stayed within a few euros of 110 € for the last 2-3 months.
Anyways for my data point: there was some external hard drives on sale for 5-10% higher than "optimal" prices.
>and stock disappeared for 4 TB+.
That's strange because those drives have terrible TB/$ compared to 10+TB drives.
Probably only 4 after you pay the appropriate capital gains taxes, I'd hope. :)
This one is easy to fact check.
/s (or is it?)
...they are not. Sale last week for 8TB at £14/TB. I would say that prices are marginally higher than they were a few weeks ago...average prices are probably £16/TB at the top consumer end but supply is fine, this price is not particularly high historically, it is not unmanageable for individuals, SSDs prices are cratering (hint: buy stuff that doesn't have massive TBW, these consumer drives are unusuable for Chia...prices of this stuff has dropped by 20-30% in the past few months), and there is ample supply on other sites too (I believe the issue with Amazon is that they just ran some huge discounts on the external drives popular with consumers..they will be restocked in a week or so).
Also, you can pick up used stuff too. The supply of used drives with lots of life left is basically endless. And no-one is mining with 4TB drives.
They rented a truck and hit up all Costcos (I think it was) from Florida to I don't know where buying all of the drives they could. There was a limit on how many so it was a mission to do it.
Edit to add a link and change "Around 2010" to 2012,