Nvidia announces mining GPUs, cuts the hash rate of RTX-3060 in half
blogs.nvidia.com
blogs.nvidia.com
I really wish Nvidia would do this across the RTX product line.
Edit: the "Syrian refugee crossing into Europe" trope might pop up as a counter-argument for crypto. I am Syrian and still have relatives there. Refugees' assets are a drop in the ocean of speculators' mining and HODLing.
I'd rather they have scaled up the demand. They've had years of this to take the higher profits and invest them. Even if people don't use them for mining, the demand for Machine Learning is there - instead GeForce cards have been banned for use in data centers for years. Not to mention that the crippling of 3060 would be trivial for mining operations to get around.
Also, my concern is not for data centers running creepy ML models that invade users' privacy, but for the average home user who wants to get started with ML and at the same time get decent frame rates when they play a game.
My argument is they should and could have scaled up the production rather than just shuffling around which group they hurt or help at the expense of which other (be it miners, ML folks or gamers). This isn't the 'right thing', it just helps the group you prefer on this round.
A sudden implosion of demand combined with a flooding of the used GPU marked. Better not plan on bringing out any new GPU architectures when your warehouses are still full of the previous generation.
Getting started with ML doesn't require a GPU at all. Realistically, any video card or even a modern CPU is more than sufficient to get started with ML learning.
It's a myth that $1000 gaming video cards are necessary or even helpful for learning basic ML.
Development of domestic chip manufacturing might even start to become a national security concern, leading to even more money flowing in on top of the money coming from crypto miners.
For reference, Visa does thousands to tens-of-thousands of transactions per second. The cost to run one of these transactions is negligible because it can be handled so quickly. Contrast with BTC, which struggles to hit 10 transactions per second on a good day. I used to work in a facility that monitored debit transactions. Our SLAs were ridiculous. Certain customers had requirements that if a single transaction took more than 100 ms, we were to open a ticket and contact their help desk immediately. This is the type of rigor it takes to maintain a financial payment network that consumers will have confidence in using.
This is not something that can be solved with current crypto approaches. There are information theory constraints regarding how long it would take to reach consensus across a distributed system and in such a way that transactions are serializable.
This defies your assertion that scaling cannot ever be achieved with current crypto approaches.
Not really a change in semantics here. It's still the point that crypto promised a LOT of really cool things for fintech that realistically won't materialize for a long time.
With the recent defi boom and the increased trust and popularity in oracles serving off chain data I'd say that you can find a cryptocurrency project for most promises.
Not being popular or adopted is another issue.
The Internet doesn't have built in mechanisms for regulation or fraud and nor should it. Cryptocurrencies obviously don't solve the problem, but can be used to create a stricter set of guarentees than the current sustem.
Later: data sharding, where you're doing the same thing but each Ethereum node only has to store a portion of that data. That will multiply rollup capacity by 23X.
Details from Vitalik here: https://vitalik.ca/general/2021/01/05/rollup.html
https://register.fca.org.uk/s/firm?id=0010X00004U9vVAQAZ
> fraud
Just to name two examples. Cryptocurrency like cash won't stop you from falling to fraud or illegal activity but communities and regulations could enforce a safe subset of projects with stronger guarentees than the current system.
Here's a list of Ethereum nodes that are staking 32 ETH, attesting blocks, getting rewards and slashes: https://beaconcha.in/validators
Roughly $5.6b in value staked, with no particular expectation date for being able to withdraw. That's 2.55% of the total available ETH.
I'm the top 25 market cap there's:
BNB (powering the Binance Smart Chain) DOT (Polkadot) ADA (Cardano) ATOM (Cosmos) EOS TRX (Tron) NEM
Bitcoin needs to be able to handle at least a few hundred transactions per second as a first goal.
The entire Bitcoin narrative has become buy and never sell (HODL). The more friction around transacting with your coins (other than buying) the better it is for the asset value.
* Accept transactions that are in the mempool * Use the lightning network
Accepting mempool transactions obviously has lesser guarantees and unfortunately lightning hasn't seen great adoption even from centralised exchanges.
Also obviously there is a human-imposed "timeout" too. Like, if you buy something with Bitpay they expect your transaction to be included in one of the next 3 blocks. If it's not, they will "timeout" the transaction and it bounces. If it is eventually mined for inclusion, they will voluntarily return it to you minus whatever fees they choose to charge (average transaction fee is $24 right now, so you will lose $24 sending it to them, then $24 for them sending it back to you, and $24 to send it again).
https://ycharts.com/indicators/bitcoin_average_transaction_f...
(of course they don't have to send it back to you, there is no network-enforced mechanism for this, they could bounce the transaction and then if it gets included an hour later they could not pay you back, the money is under their control once you've signed the transaction. They could even maintain a separate mempool of these "dead" transactions and use a friendly mining pool to process those transactions preferentially...)
>The existing Visa credit card network processes about 15 million Internet purchases per day worldwide. Bitcoin can already scale much larger than that with existing hardware for a fraction of the cost. It never really hits a scale ceiling. If you're interested, I can go over the ways it would cope with extreme size.
I can't tell you from personal experience if it would be the same if my amounts were a million plus, but they are not insignificant amounts either. I'm quite sure that attempting to deposit similar amounts of physical cash at a bank would be at least somewhat challenging.
That makes 3.6667 transactions per second. BTC needs a hard fork in order to increase this dramatically.
So you were actually quite generous with 10 transactions per second...
[0] "Blockchain Charts" https://www.blockchain.com/charts/n-transactions-per-block
Which was proposed, and shot down by core Bitcoiners.
They aggressively do not want to increase the transaction limit.
Miners want transactions to be expensive, because that's where they'll get their money when the mining reward falls again.
It's mind boggling that people will complain about banks conspiring to charge them hypothetical fees, then rush to use a cryptocurrency that has double-digit transaction fees when nobody is really using it for anything other than speculation. The reality is that Bitcoiners don't actually want people using Bitcoin as a currency, they just want you to buy and hold. No selling, no spending, just buying. That's the only thing that drives the price up, and that's the only thing they want you to do.
For example, Lightning is built ontop of BTC:
Once all of these cryptocurrencies and exchanges become fast, the price doesn't actually matter. If I can buy the crypto, transfer it, and exchange back out of the crypto to my target currency in seconds, it doesn't matter if SOL is at $0.01 or $100 per coin. If I'm only in it for a few seconds, I just buy and then immediately sell however many coins are needed to move my value from Point A to Point B.
The reason Bitcoiners don't want larger block sizes is that they don't actually want spending. They want Bitcoin to be an asset that only goes up in price. Spending and selling drive the price down, so those need to have high hurdles to discourage that behavior.
I might be answering the wrong q, but the separate currency represents the cost of a transaction in the network. I.e. the amount you're willing to pay to send a transaction in the network and the amount a staker is willing to receive in exchange for validating your transaction
How else would you represent that cost in a crypto network? Some kind of currency needs to represent the transaction cost.
Unlike enthusiasts they don't see it as a necessary step and haven't been following the long and arduous journey of developments and breakthroughs to get here.
By the same logic, trucks are useless and inefficient because my moped burns 1/100th of the fuel per km.
> This is not something that can be solved with current crypto approaches. There are information theory constraints regarding how long it would take to reach consensus across a distributed system and in such a way that transactions are serializable.
This is demonstrably false. Newer-generation coins offer better tps, and even bitcoin itself can manage due to things like segwit and lightning.
Ok, I’ll bite.
No, it doesn’t, but what is that actually useful for?
The best answer I can come up with is illegal activities and libertarian fantasies.
Even if one generously frames “illegal activities” in the context of financial freedom from a tyrannical regime, even that has limits.
Any regime with a banking system can implement a KYC-type of policy that will can make interacting financially with anything related to BTC difficult or impossible. Need a modern example? Look no further than online wallets (e.g., Neteller) and online poker sites after the UIGEA.
I will add that if any government views a digital coin as a threat to their fiat currency system, it will be trivially easy for them to make that currency an extremely inefficient store of value. The measures will be draconian towards coin holders, but most people (rightly or wrongly) won’t care.
This fear is greatly exaggerated by people who have bought into the false idea that Bitcoin is the only way to escape inflation.
The definition of inflation is that asset prices go up. You can escape inflation by simply buying assets, or basically anything other than cash.
If someone truly believed massive inflation was coming, they would want to take out USD-denominated mortgages and buy houses. Inflation would increase the value of the houses and reduce the value of your loan repayments.
It's silly that so many people are afraid of the government devaluing their cash that they've decided the best way to escape it is to pile into an asset with an extremely lopsided distribution in favor of early adopters. The average person stands to gain more from recent stimulus than they stand to lose from inflation in their cash accounts. Doubly so if they invest their cash in stocks or assets, which doesn't even require Bitcoin.
The definition of inflation is typically that consumer prices go up due to there being "more money". The only kind of asset that people consume is housing, and even then they don't really - housing as an asset (purchase price) can be disconnected from its consumption price (rent). Besides that, you can't eat stocks so you don't get poorer when the price goes up.
Assets can actually increase in price because of deflation. For instance, technology getting cheaper has made all the stockbrokers drop their transaction fees and add fractional shares, so now everyone can bid up the same stocks.
Ownership. It's the difference between "I own 20 quatloos because the Bank of Endor says so" vs. "I own 20 quatloos". It supports actual ownership instead of representative or proxy ownership.
In the same way that you can have $20 in the bank or $20 in your wallet. When it's in your wallet, it's actually yours and you own it. When it's in your bank account, then you don't actually own it, you must trust that your bank will do the right thing and give it back to you when you ask for it. That level of trust has been wavering for decades.
Remember that Bitcoin was started as a rebellion against the "second bailout for banks".
OK... so what exactly is "ownership"?
A buddy of mine mined ~160 BTC (currently about USD $8.1m) back when it was fairly worthless. He left it on a hard drive on one of his many machines. BTC became worth something. He look for said machine and HD. Couldn't find it. That BTC is effectively gone forever. So... who owns it?
There are a not small amount of stories like the one above for various levels of non-trivial amounts of money (tens of billions). https://www.nytimes.com/2021/01/13/business/tens-of-billions...
Situations like this point to a desire for many/most people to want some sort of custodianship so that they don't accidentally lose something that might be worth millions one day... or even hundreds or thousands one day. At that point, one is looking at banks and banking systems, so we're back at square one.
If one is looking at an alternative to banks, we sometimes/often get stuff like Mt. Gox and Bitfinex -- a situation in which a group of people who want to be independent of the banking system learn the lessons of the value of the banking system the hard way.
So I guess where I am at is that I appreciate a desire to rebel against banks -- they wield way too much power. That said, I don't think that it's particularly prudent (or maybe even possible) to throw the baby out with the bathwater on this matter. The banks definitely do some fairly nefarious stuff, but they also do a metric shitload of very pedestrian-but-useful stuff extremely well on a daily basis. Until a coin can replace that pedestrian-but-useful system with equal grace, I doubt that a coin will be able to replace our current system. Furthermore, somehow if it does, it will be because the current system integrates coins into the current system rather than coins usurping the current system.
So again... what is distributed consensus and verifiable transactions with no central trusted authority actually useful for?
My point is that most people seem to prefer having a central trusted authority, specifically for custodianship.
That’s why most people put money that is significant to them somewhere in the banking system rather than a shoebox or a tin can somewhere. They want that money to be be traceable, accessible, and recoverable.
There are exceptions, of course, but they are the exceptions that prove the rule, IMO.
Really? Ignoring the philosophic debate, ownership in terms of BTC deals with the capacity to modify the ledger using an agreed-upon protocol -- but you already know this, and I don't know why you're asking that question. What do _you_ think ownership means in terms of BTC? Do you think you can implement a public system for ownership of digital assets without a distributed consensus system? Or Verifiable transactions? Maybe you can, and if you do I'd love to see it. Good Luck.
>So... who owns it?
The person who has the key.
Yes. You sign one receipt at the store with a security camera pointing at you, and are handed a second, matching receipt by the vendor.
And don't come back with "you shouldn't want to avoid KYC and regulation" as that's a different topic. Others find those properties useful even if you don't.
My knowledge of this technology isn’t very deep. From my understanding, mining is essential to facilitate transactions, which in turn makes cryptocurrency a currency. Mining needs computing power, which needs energy. More energy translates to more rewards, which incentivises parties to keep mining and keep the network running. Even if you make the underlying calculations more energy efficient, miners are still incentivised to spend more energy to capture more of the reward(?).
I'm not a fan of using so much energy, just as I'm not a fan of guns. What does that have to do with discussing how they work and why people use them?
If you actually believe what you're saying, then short it[1]. If you really think that the practical value is non-existent, then put your money where your mouth is, and take out a short.
This isn't some academic theory where the best you can do is throw tomatoes at it, and ridicule it from a distance. This is an economic theory, where if you are correct, and everyone else is wrong, you can profit significantly from it.
Maybe once you're faced with actually taking a stand on a belief like this, you'll come to understand why cryptocurrencies are beginning to dominate the global economy. It's also totally possible that you're right, but I think it's far more likely that you just like to ridicule things you don't understand because you get some sense of power from shitting on people who you consider to be beneath you.
And thinking that bitcoin is a terrible idea is not the same as thinking you can predict its market value. When something is as volatile as bitcoin, shorting it is almost as bad as going long; eventually its value will likely drop but being early is the same as being wrong.
As I've said elsewhere: doing one transaction on bitcoin uses approximately the amount of electrical power that an average US household uses in 2 weeks.
About 700 KwH, far as I can tell. (Based on 76.87 terawatt hours annually in 2020, as an estimate for bitcoin). A house uses ~1000/month, ballpark.
PoW is also the most efficient way to do what Bitcoin is doing. Attempts to displace PoW have all failed, because the required trade-offs are always too high, and detract from the value of the network.
If you can figure out a fair way to seed value on a network that doesn't require burning energy, then please let the people know, and your new magic cryptocurrency will dominate the crypto-ecosystem overnight. No one would love that more than me. I could talk for hours about the various ideas people have had in the past, and why they failed, but I don't think this is the right place for that rant.
It's also important to recognize that there isn't a single watt of PoW power that's wasted. The power utilized reflects exactly amount of value that Bitcoin is providing to the world. Satoshi also made the mining rewards get cut in half every 4 years, so the value to power usage ratio gets better and better over time.
I could also go on about how mining facilities overwhelmingly are using excess power produced almost entirely from clean power sources. If the goal is reducing carbon emissions, there are so many serious offenders that are massively more wasteful.
Don't you mean reverse?
Maybe I'm willfully ignorant, but I just don't see any practical use outside of massive speculation game. Mind give us some examples? :)
It really doesn't matter if you don't recognize how big of a deal that is, but to A LOT of people, it's a complete game changer. If you don't need a service like that, well that's fine, but obviously a lot of people do, which is why the network is where it is today. Massive amounts of value is being moved through the network daily, and the fees that people are paying to utilize the network are not at all trivial. It's worth it to them though, or they wouldn't be using it.
If I could short bitcoin over a 20, 30, 40 year timeframe I absolutely would, unfortunately that's not possible. There might be utility in some cryptocurrencies for some applications, but bitcoin as it currently exists is worse than useless, it's actively harmful.
I'm speechless
It's to the point where I'm thinking about selling my "backup" 1060 3GB card, but if my GPU died then I would be in real trouble because lol at getting ahold of GPUs right now.
like, shockingly to some, there exists a market for GPUs for doing things other than mining. And not all coins require 8GB cards either.
ML is a bit different in that the applications are probably here to stay, and the applications are "legit". But, the ML market will not have the same brand loyalty that the consumer gaming market will. Consumers like brands and they stick with them. Companies spending money on ML are going to flock to the cheapest TCO, and a lot of this will also be abstracted through cloud offerings.
In short, they don't want to piss off what they believe to be their cash cow for years to come.
Some gamers are switching to AMD now that Nvidia cards are sold everywhere.
Meanwhile, companies are tied to CUDA, and AMD still doesn't have a similar API offering. There's some progress, but they're not there yet.
Whatever original hopeful goals the creator(s) of Bitcoin had with this idea, and decentralised currency...if there was any victory at all it is a pyrrhic one when you look at the externalities.
If you can call a digital gold rush a victory, anyway. The physical one in the US was pretty bad for a lot of people.
I'm gonna steal that one, thanks. :D
No? Then it's a calculation. Or an abstract and useless one in your terms.
I don’t have an opinion on this but if you’re making a point, you should be the one providing evidence, not telling others to look it up.
Yes, it's work to back up your claims with direct links to a source. But maybe that's a reasonable minimum bar to hurdle to stand behind ones statements.
This is different than people blithely spouting “it’s not my job to educate you” when asked to give some support for their controversial, unsubstantiated claim.
The energy usage of ethereum comes from the proof-of-work DAG, not the solidity execution. You can easily see this by running a (non mining) ethereum node, any decent CPU will churn through the transactions in blocks in a matter of milliseconds (~200 ms per block), not the 13 seconds it takes to generate that block
The upkeep cost for Bitcoin is, and by necessity must be, proportional to the value represented by Bitcoin.
Actually though, I wonder if there's an analogy that could be drawn between the real origin of gold (supernova) and bitcoin (mining a block). Both are created as dense byproducts of a large compression event (bitcoin: transactions; supernova: the rest of the star). ?
Unless you've got some dyson spheres lying around that you're not telling me about, I'm going to politely request that bitcoin advocates not expend more energy than we have in the entire world just so that they can diversify their investment portfolios.
This is a tangent, but it's more fun than the usual Bitcoin fetishizing and bashing.
There is building evidence that neutron star collisions, not supernovae, are the primary source of the universe's gold [1].
[1] https://www.quantamagazine.org/did-neutron-stars-or-supernov...
But let's be honest, Fort Knox doesn't require 0.5% of the world's energy consumption (and increasing) just to continue existing.
False equivalencies don't help this argument. Yes, we know everything requires energy. No, that doesn't make it all the same.
In other words, if you just mine a physical gold block, you don't really have an asset without the state's (or your own) continual upkeep that keeps it protected from attacks.
This completely misses the point, which was one of magnitude:
>> The upkeep cost for Bitcoin is, and by necessity must be, PROPORTIOANAL TO THE VALUE REPRESENTED by Bitcoin.
Also, that upkeep cost is 100.00% to protect the value of existing bitcoin.
C.f., gold:
> If you want to own some gold you need to protect it (usually with the state apparatus of violence, which has massive externalities).
I don't need continuous mining of bitcoin to protect my non-bitcoin stuff.
However, I do need the "state apparatus of violence" to protect my non-gold stuff, including things like my house and my body, and oh yeah, my fucking mining rig!
If all gold disappeared tomorrow, would anything about the "state apparatus of violence" change? No, of course not. That is absurd.
Although as a Brit myself it's amusing to consider myself some kind of atavistic archetype of humanity
This describes Bitcoin too.
I think it's facially absurd to claim that the existing financial system (including military, police, etc) costs anywhere near what bitcoin does, but even if it does, it's irrelevant because bitcoin doesn't provide the same things. Bitcoin is a virtual bar of gold or piece of currency, not a financial system.
"Of the existing 18.5 million Bitcoin, around 20 percent — currently worth around $140 billion — appear to be in lost or otherwise stranded wallets, according to the cryptocurrency data firm Chainalysis"
"“Even sophisticated investors have been completely incapable of doing any kind of management of private keys,” said Diogo Monica, the co-founder of a start-up called Anchorage, which helps companies handle cryptocurrency security. Mr. Monica started the company in 2017 after helping a hedge fund regain access to one of its Bitcoin wallets."
"“This whole idea of being your own bank — let me put it this way, ‘Do you make your own shoes?’” he said. “The reason we have banks is that we don’t want to deal with all those things that banks do.”
Like remembering where the money is!"
https://www.bloomberg.com/opinion/articles/2021-01-12/don-t-...
Mutual funds, the most legible way to get an estimate of the holding cost of securities (since they are required to transact at net asset value) routinely charge between 0.2-1% fees, depending on size and asset class.
Clearinghouses, transfer agents, and cash management firms don't work for free, which is the problem that cryptocurrencies explicitly solve; neither do compliance, KYC, fraud, etc., which most cryptocurrency protocols elide altogether.
Oh wait.. You don't want that functionality, and you're still using up the energy footprint in the event you are correct. One of these delivers significantly less versatility than the other for the same energy investment.
I was curious about this. What would "significant" be? What is it based on - the number of outstanding blocks? Is this vulnerability mentioned anywhere in the protocol spec?
[1] https://en.wikipedia.org/wiki/California_Gold_Rush#Longer-te...
If a billionaire hoarded manufacturing plants and kept them shut down, people would probably be mad, since resources were being wasted, even if it was a net positive on greenhouse gas output.
It is almost exclusively reflected back out into space, and wasted. We have nearly no incentive to capture it, so the energy is being wasted at a mind boggling scale. I've been mining cryptocurrencies with solar power for about 6 years or so.
The great thing about crypto mining is that it creates a price floor for energy usage. This gives people real incentive to build the tools we need to harness the energy on our planet.
[1] https://www.businessinsider.com/this-is-the-potential-of-sol...
Solar? It heats the atmosphere, because it captures solar heat which would bounce back.
Wind? Obviously it slows winds down, and who knows how it affects the existing ecosystem. We know how important the gulf stream is, you wouldn’t want to slow that down..
Etc..
I am obviously not an expert but I looked at the magnitude of the energy itself vs the greenhouse effect and it's about 10-100x less potent than the greenhouse effect. But that means that even if we had 100% clean carbon-free energy, we can't continue increasing our energy utilization forever. And we still have a huge portion of the world that is below a first-world living standard and will eventually want automobiles and air conditioning and vacations using jet travel too.
IF we do not level off the energy consumption of the "first world" standard of living sometime within the next 50-100 years then we are on a course to cook ourselves to death even with 100% clean energy.
Whether or not that limit is actually reachable and something we should actually worry about I don't know. i.e: is there enough physical space for that many people on earth where each person uses X amount of renewable energy where the sum of X is greater than the amount of energy the earth can actually capture without being negatively affected.
As was pointed out upstream, capital and resources are required to get a solar panel, and if its output is directed to something frivolous like crypto mining, those inputs are being wasted. The fact that there remains unused energy from the sun is completely orthogonal to the waste that is happening.
To be clear, I'm less passing a judgement about the efficiency of crypto than I am pointing out that crafting objectively absurd arguments about why it is not actually wasteful helps nobody.
A) the environmental impact of manufacturing a solar panel is 0 and carbon neutral itself (it's not),
B) there's an excess of solar panels available too (not really),
C) the bitcoin miners aren't often running on coal power during the nighttime hours (they often are), and
D) there's nowhere better for mankind to deploy the capital used in deploying miners and renewable power production than for cryptocurrency mining (dubious).
The infinitesimal portion we do manage to convert with the currently available technology shouldn't be pissed away in the wind so lightly.
>Markets facilitate trade and enable the distribution and resource allocation in a society.
as long as a majority of bitcoin mining happens from non-renewable resources, the network as a whole is damaging.
That's not true at all.
You can enter a power cost into this calculator: https://www.nicehash.com/profitability-calculator/nvidia-rtx...
It comes out positive using energy prices from virtually anywhere.
The idea that mining is powered purely by renewables that would somehow otherwise go to waste (that is, not take fossil fuel plants offline) is pure FUD.
(It's only an upper bound because energy isn't perfectly fungible.)
Renewable energy is also better than fossil energy in terms of environmental damage, but it's not free from damage. A dam is a massive localised environmental impact, and a huge amount of concrete. A solar farm requires (holes-in-ground) mining and significant industrial processing to build. The lower bound on environmental harm is the cost of building renewable energy sources that are used for something so unproductive.
Or is there energy being "stolen" from other uses?
As I wrote: is there any reliable data?
It's a myth from the pro-cryptocurrency crowd.
With recent price spikes, mining is profitable even if you have to pay relative high prices for electricity. You don't need to have cheap energy.
Also, it's another myth that we have an abundance of hydroelectric power that would otherwise go to waste if not for Bitcoin. With modern power transmission it's trivial to send power through the grid to other locations, and nobody is building power plants in the middle of nowhere without a way to send that electricity somewhere else.
Finally, Bitcoin's power consumption is additive. Energy that goes into Bitcoin mining is energy that isn't being used to power cities and other infrastructure, which must still get its power from somewhere.
Don't believe the whitewashing campaigns to downplay Bitcoin's energy usage. It's very bad.
Happy?
Bitcoin is not humanity using energy. It is humanity flushing energy down the drain.
"But look at the sun man! That's the real waste"
Are you high?
A more useful perspective is "humanity uses more and more energy for computation". The many devices which enable us to post these comments burn energy. The astronomically high video-traffic these days burns energy. Modelling early universe, protein folding, advanced chemistry... every damn thing we compute has the cost of burning energy.
As an aside, let's not forget how much electricity gets used simply for heating. There's no shortage already of "smart heaters" with a nice side effect of mining you some coins while they keep your house warm. Anything environmentally wrong with that?
See, it's not the endgoal of cryptocurrencies to burn energy; it's simply a cost. Now, thanks to bitcoin, we can have economies of scale drive optimization of that cost: improving the core compute-per-joule ratio of our technology. With all the collateral improvements across the globe with regards to environmental impact.
Could you say that porn distribution industry "kind of by design" lead to people watching more and more porn? Regardless of the answer, thanks to that we can have fast internets now. Similarly with bitcoin: thanks to it, we'll pack more & more bitflips into every joule, up to the thermodynamic limit.
There’s real uses for Merkel trees in finance - Cryptographically secure history is valuable. I’ve yet to see any problems that blockchains solve that merkel trees do not[1]. Finance and interchange works just fine with just Git - Interchange done via simple published balancesheets with countersignatures by both parties, then merged into “head” on the regular by any form of rollup. It doesn’t have the (well overstated) pseudonymity advantages of bitcoin, and that’s a feature in a well regulated financial market.
[1] With one ridiculous exception: In Iain Bank’s “Excession”, a mothballed war-fleet is resurrected by a traitor who presents enough cryptographic signatures of coconspirators to burn the keys of the real C+C chain. Blockchain would make this much harder - But not necessarily enough harder, even if they had to effectively recreate the military history of an entire total-peace (through superior firepower) civilization.
First they came for the miners, and I did not speak out.
I hope they don't, given that there is limited upside and a lot of push back from dev community.
I've actually been wondering about this -- are there any major ones that meaningfully provide other kinds of value for all of the computational resources they're using?
My understanding of blockchain as a concept is pretty surface-level, so maybe I'm just missing something obvious here. But at least at first glance, it kind of baffles me that so many of them are using seemingly arbitrary hash algorithms to calculate proof of work, when they could instead be tied to distributed projects -- folding@home, for instance -- and measuring actual work.
[0]: https://www.newsbtc.com/news/ethereum/ethereum-on-pace-to-se...
There's a few others that take different approaches like Golem Network Token, Nym, Oxen, etc.
I'd understand if you wanted all economic activity to have to figure in the negative externalities before a profit could be realized by stake/shareholders, but your comment implies to me that crypto should have some kind of special tax because you're upset you cant buy a gpu.
This is exactly the kind of antagonism that is detrimental to everybody. Read op comment again:
> Mining trashes the environment and, for the majority of crypto assets/volumes, does not provide genuinely valuable economic activity besides speculation
I am not "upset I can't buy a GPU", mining uses a lot of power[1] and its environmental impact is not being taken into consideration. So yeah, if mining is damaging the planet and if we can assess the damage quantitatively and link it to mining without a doubt, it should have to be taxed accordingly.
Replace 'mining' with 'anything' and I fully understand as I'd said.
If I'm running 3,000 watts worth of GPUs at home to heat my house, and you're running 3,000 watts worth of space heaters at your house, I'm certainly not damaging the environment anymore than you are.
We should both pay the FULL cost of the toll the generation of the required electricity cost to run our equipment. Why I should pay taxes above and beyond what I already do because my heater makes money is where you totally loose me.
However, it is important to note that mining does do something useful: it secures the network from attack. To my knowledge, there is no known alternative solution to the distributed trust-free consensus problem.
Just becuase your use case is more noble doesn't change the fundimental issue of supply and demand. Nividia also already has artificial restrictions on their GTX and RTX lines to produce a product for your use case: the Tesla series.
Anyway, seems easier for them to obtain crypto than getting a commodities account and buying gold, since presumably one would have to deal with KYC/AML in that case, and one can continue to use crypto as a currency within the country (though I doubt Bitcoin, since the transaction fees probably amount to more than a typical Venezuelan earns in a month).
Not yet, but don't presume current economical status quo is working flawlessly. The current monetary system is kind of a joke, it no longer is backed by gold, in fact it's not backed by anything. So if you think about it for more than a minute why should it be that currency in the US should have any value? Creating an alternate store of value that is not just printed or made up and can be proven to exist at the very least provides something not too dissimilar to a database backup in case of a complete meltdown of fiat currency. It, however, does need to be allowed to get sufficiently big for a future restore to work. That is worth a bit of electricity. Yes, a bit, as a percentage of the GDP.
Good luck with this one NVIDIA but you're playing a losing game. As soon as one un-restricted version of the driver exists, this is as good as useless and demand will be the same as ever for these cards unless they're able to ship CMP cards at a very high rate, but my understanding is the restriction here is silicon.
Let's not forget that last time they produced "mining only" cards, people managed to crack the drivers to render graphics too... this is no exception.
https://videocardz.com/newz/zotac-geforce-rtx-3060-early-gpu...
There's speculation that the limiter might actually live in the cards VBIOS, which is much harder to modify than the driver as newer Nvidia cards require a signed VBIOS to boot.
My Gigabyte motherboard is doing that. Ain’t too old, has AMD’s X570 chipset.
I'm also curious if this "mining detection" is going to kick in "accidentally" on other compute loads. It would be really crappy if the CUDA performance of these cards is bad because they don't want miners using them.
Pure speculation but I'd imagine that Nividia's drivers can properly differentiate and already had the limitation in the current release.
Given that I was doing this as only a rather casual miner with three GPUs, I have no doubt that this new restriction will be attacked ruthlessly.
That sounds pretty interesting. Mind elaborating on what changes you achieve with your modifications?
Basically just lets my Plex server transcode more streams at once without having a Quadro.
And... why do you need your Plex server transcoding multiple streams? Just curious!
https://techgage.com/news/nvidia-bumps-nvenc-stream-limit-on...
It's a sort of accidental bug bounty and I wonder if Nvidia have really thought this through?
On a side note all those virtual currency are just plain wrong on the environment side, it does not make any sense in 2021 to have a currency that need heavy computation thus increasing power consumption. When people designed those currency they def did not think about that.
They've been "moving to proof of stake" for something like three years.
The miners are making money off of them not moving to proof of stake. The miners stop making money when they move to proof of stake. The only people who can actually allow the chain to move to proof of stake are the miners.
Meanwhile, xtz and ada have been proof of stake since their inception, 3+ years already for both.
They've been _planning_ to move to PoS for ~3 years, with a much too optimistic timeline. In the meantime a lot of companies working towards that have collapsed or refocused on other efforts, delaying the whole transition. It will probably take another ~3 years (my personal estimate) until the transition is reasonably finished and core parts of the ecosystems have migrated, but at least it's moving into the right direction.
PoS has always been on the roadmap even before their ICO. But yes, with a MUCH more optimistic timeline.
2-3 years ago is when the Casper FFG paper was published and subsequently the architecture of Ethereum 2.0 (and the concrete 3-phase roadmap towards it) were published.
The miners can keep running the legacy version, and try to convince users and exchanges to stick with it. Given the massive community support for PoS, they're not likely to be very successful.
They claim they are going to fork ETH to make a POS fork of it. Now the people claiming this are going to want you to believe that’s the “real” eth- but the miners may not agree. They have no incentive to do so.
And yes, they can absolutely keep running the real chain and let the chips fall where they may. Given that course of action benefits them, and the other benefits Vitalik and his whale friends...
It doesn't matter what the miners believe, if the exchanges, applications, and users have moved to the PoS chain. Which is very likely to happen, given the community support for PoS, the scaling coming on that side at around the same time, and the fact that exchanges are already letting users earn returns by staking ETH, while the exchanges take a percentage of profits.
Vitalik et all have been talking about pos for years. Meanwhile a huge portion of the cryptocurrency space went there already (years ago) and left eth behind. POS isn't some new technology.
Fees don't come from whether it's pow or pos, they come from volume and the need to prioritize transactions.
The migration is a decent amount of work but it's not the difficult research task that the beacon chain was. Someone did a proof-of-concept in about a day, though the exact plan has changed since then.
Likely closer to: they'll create a mirror on the new chain and a group of unknown size will keep ETH 1 fork alive.
Kind of like how Ethereum Classic (ETC) is subject to 51% attacks every other week and requires multiple days' worth of confirmations after deposit to any exchange.
Once you make changes to a chain its not easy to fix without making a huge mess. Hopefully the migration with be done by Q1 2022.
And they probably didn't factor in ASIC miners and large mining pools, either. Proof of Work is legacy technology at this point, it's a good thing Ethereum is moving to Proof of Stake along with most of the ecosystem.
So yes we do have alternatives to gold. Things such as Bitcoin and crypto, which keep going up in value, while gold speculation has stagnated and is even going down. Weakening dollar and massive debts due to the pandemic should mean gold skyrockets, but people are buying the alternatives. It looks grim for gold. Tesla didn’t buy gold, they bought Bitcoin.
It sucks that gold is so desired just for its look. If it were cheaper it would be used a lot more for technological devices.
I can't wait until asteroid mining crashes the gold price and hopefully puts an end to the ecological disaster that is gold mining.
Ethereum is instead solving it by having an initial proof-of-work period and then moving to proof-of-stake after some amount of mining and wealth distribution has already happened.
ADA and XTZ did pre-mine their initial distribution, and now they distribute coins through stake rewards.
Five months later I still have my eBay searches saved, and the exact model I bought is now regularly selling for 40+% more than it cost me. There's one right now with a few hours left on the clock, currently going for 50% more than I paid, with 75 watchers and 51 bids. It's wild.
How much co2 is emitted in the clearing of dollar transactions vs the clearing of bitcoin for example. That’s a whole lot of data centres.
I’m not persuaded that we can put costs like legal (or even military) protection under just the dollar column - if you steal my $CO2COIN i can leverage the legal system to pursue you.
About 40000 times less I think? That's the figure I saw for a SEPA transfer or Mastercard transaction vs the bitcoin equivalent
So I don't think we can lay the military's oil consumption at fiat's feet.
Also, a minute with Wikipedia and a calculator will show you that transmission losses for the longest possible straight distance in China are <5%.
So what's your point?
There's an emissions cost to producing the equipment required to construct the hydro plant.
There's an emissions cost associated with constructing the hydro plant.
This is multiplied if you are building it in an area without heavy industry (no existing demand, remember?) - because you have to construct your own support infrastructure and road network to build and service the plant.
Not to mention the impact of the now flooded 'wilderness' (which is highly likely to contain carbon sinks).
Or the impact to forestation and vegetation downstream.
You're arguing with a discussion that didn't happen.
You then proceeded to excoriate me for not addressing everything in every discussion that has ever happened in your mind about bitcoin.
https://www.theguardian.com/sustainable-business/2016/nov/06...
There are others that are.
In the long run miners who do not use untapped (cheaper) renewables are going to be priced out, it just takes some time. That said other protocols much less computationally intensive, such as "Proof of Stake", are starting to be tested.
Anyhow, good idea NVIDIA!
So where exactly did you read that in the Bitcoin whitepaper? Because everything else is just guessing.
"Gold mining is a waste, but that waste is far less than the utility of having gold available as a medium of exchange.
I think the case will be the same for Bitcoin. The utility of the exchanges made possible by Bitcoin will far exceed the cost of electricity used. Therefore, not having Bitcoin would be the net waste."
I think we're long past this point and we have better alternatives to PoW now.
I dunno. Suppose Acme, Inc. starts selling a giant red button. When pressed, it means my boss will decide to give me the rest of the day off, no strings attached. There's a little disclaimer in the fine print that warns that whenever the button is pressed, Acme, Inc. will roll a 6-sided die. Whenever a 1 is rolled, a small child in Kenya will be executed.
Individual users of the product can claim that they aren't directly responsible for the externalities, because "I'm not the one performing the executions!" or "my dice roll was a four!".
Or, perhaps there are business users of the product, who are motivated by cost and nothing else. Good luck convincing a company that ethics are more important than its bottom line.
Sometimes users of a product are so detached from its consequences that they're willing to accept them, even if the true result is horrible. We should certainly hold companies accountable for exploiting human nature in this way.
Given the existing investment in Ethereum HW, it almost seems like it’s just a carrot to deflect criticism about power consumption, but not something that will actually happen.
Once you have that, you can get consensus on literally any data just by putting a hash of that data in each beacon chain block. In this case, a hash of the EVM state, while the EVM runs as before.
There's still some discussion on the design, but that's more about deciding the best of multiple options, rather than figuring out how to do it at all. As I mentioned in another comment, someone built a running proof-of-concept in about a day.
The hard part is figuring out how to execute arbitrary code and synchronize it across all shards, without removing the benefits of having shards in the first place.
Around the same time is the migration to data sharding, which multiplies data storage by 23X initially. That's without code execution on the shards.
Zkrollups in production now do the execution off-chain. On chain, you store compressed state data and a compact proof that everything executed correctly. That's good for a couple thousand tx/sec today, and 23X that after data sharding. Given that, Vitalik has argued that shard execution may not even be necessary; it wouldn't add nearly so much transaction capacity as we'd have already.
Here are the papers on how Ethereum consensus actually works:
https://arxiv.org/abs/1710.09437
https://arxiv.org/abs/2003.03052
And here's more detail on rollups:
There are approximately 2 million employees working at FDIC insured banks in the United States.[2] There are approximately 330 million people in the United States.[3] United States food production uses about 10 quadrillion BTUs per year (about 332 gigawatts = [10 quadrillion BTU / (365246060) sec to gigawatts]). 332 gigawatts / 330 million ~ 1kW. That's approximately 1 kilowatt per person in the United States. There are probably some good arguments about exports etc but we will hold that aside for now. In 2019, total U.S. primary energy consumption per person (or per capita consumption) was about 305 million British thermal units (Btu).[5] = 305 million btu / (365246060) sec = 10kW. So, we get around 11kW per person for the total food and primary energy consumption rate.
11kW / person * 2 million people = 22 gigawatts
So, we come to the conclusion that if you look at the people employed by FDIC insured banks in the United States alone, they have a larger impact on the environment than bitcoin. If my calculations are correct, maybe we don't need this complaint to be the same first post in every thread on crypto.
Maybe I'm missing something or have done an incorrect calculation, if so, please point it out.
[2] https://www.statista.com/statistics/193286/number-of-employe...
[3] https://www.census.gov/popclock/
[4] https://www.chooseenergy.com/blog/energy-101/energy-food-pro...
We’re running two banking systems right now. It’s inefficient and stupid.
I mean, how do you do for example mortgages or corporate finance with bitcoin - without hiring all the people and using all the resources current financial system does?
Frankly, more proper comparison would be the resource consumption of using gold in the financial system.
I fail to see how this is relevant to the question of how bitcoin compares to the banking system as a system for storing and transferring money from an energy perspective.
The relevant comparison would be to compare the costs of doing a similar volume of similar operations using bitcoin and the bank system.
Ok.
I don't know how can you say that after volatility in banking systems crashed large parts of economies of at least a few developed countries leading to resurgence of nationalist leaders that fuel and use those sentiments in people damaged by banks gambling on stuff they can't be trusted to count properly.
We'll see how well your wealth gets preserved when China stops buying US debt for whatever reason and oil is sold for euros or yuans. I think there will be proper mess in global economies and you might use a thing that can't be fudged by people.
And I am not particularly for any energy conservation measures. Except for isolating buildings because most dirty stuff we burn is for that purpose and not isolating your home is just negligence.
Wasting electrical energy is only bad because it came from burning stuff. But I don't think any fossil fuel power plant was closed because people used less. Maybe some were not built because of that, but then some renewables and nuclear power were not built for that reason as well and I believe that's very bad.
> Wasting electrical energy is only bad because it came from burning stuff. But I don't think any fossil fuel power plant was closed because people used less.
Coal plants are indeed closing because they are no longer profitable. Natural gas is the main “on-demand” power source, and plants raise and lower production based on demand. Increased power consumption will delay the transition away from coal, and keep natural gas plants running more often.
But not to drive up demand there are better ways to do that in case of housing. There are places where more efficient housing can't be built bot because lack of demand but local regulations that serve local wealth owners.
Waste food... Not so much bacause food preduction doesn't really suffer as a result of low demand. We pretty much grow food as efficiently as it is possible and in places we don't we do it for different reasons than just lack of demand. There are food producers that destroy some food they produced to drive up the prices by increasing ratio of demand to supply.
Coal power plants close because there are cheaper ways to make electricity now, not because demand dropped.
Crypto mining operations can provide load on-demand which is almost as good as power production on demand while balancing power grid.
Nuclear power plant plus huge crypto mine can provide as much flexibility as coal power plant if not more.
The price for power is set by supply and demand. Increased demand for power increases the price. This means that coal plants can afford to operate depending on the amount of demand.
Here’s a story about coal plants closing due to lower demand: https://www.google.com/amp/s/www.wired.com/story/the-covid-1...
Here’s another article about how natural gas plants are replacing coal and nuclear: https://www.mckinsey.com/industries/electric-power-and-natur...
Any increase in demand for power will increase carbon emission by keeping coal plants from shutting down, building more natural natural gas plants, and keeping those plants online more often. Renewables are important, but only satisfy a minority of demand for electricity.
In CO2 free future we'll have all of home heating running on electricity and all of our transportation too. Power generation needs incredible amount of investments over the next 50 years. I believe crypto mining might drive those investments so we are ready for that future.
Even now there are instances where flexible demand of crypto miners helps balance the grid now: https://www.bloomberg.com/news/articles/2020-09-01/bitcoin-m...
Because it makes so much sense to dedicate equal or if not vastly greater resources to playing games?
Mining needs more resources to solve a brute force problem whose difficulty is increased just for the sake of making the problem even harder to solve.
So no, not equal, not greater, in fact lower. And I think the Graphics Processing Unit should process graphics, don't you?
It's going to take days for mining operations to reverse-engineer the driver and undo this. So commercial operations won't be affected at all.
Worse, their detection isn't going to be 100% perfect. What if After Effects does something that looks "kind of" like mining. Is it just going to be crippled? What will nvidia do then? Suggest I buy a Quadro for AE support?
This will only penalise average users.
I wish AMD had a working product to kick their teeth in...
Vendors and demand have effectively driven prices to an insane amount, and you just can't get them, unfortunately. Short of AMD doing something like Nvidia is doing here, I doubt we'll see them on store shelves any time soon.
[0] https://www.mindfactory.de/Hardware/Grafikkarten+(VGA)/Radeo... [1] https://www.alternate.de/Grafikkarten/Big-Navi
Edit: 1100$ actually.
RX 5700 will hit 55mhs and the RX6K series will do ~68mhs, but the cost of the 5700 is typically like half that of a 6800.
There's a reason it's easier to find the new AMD stuff in stock than nvidia stuff, they've got a terrible mining roi.
If you're in the market for a new card and you have the option of a powerful, purpose-built card that works out of the box or one that needs endless tweaking, few people would choose the more complex option. Especially if it means every new driver that gets released is going to require days to make work properly.
Constant GPU driver updates are the norm for gaming GPUs because GPU drivers include a huge pile of workarounds for buggy games, and there are new buggy games hitting the market all the time. But if you just want to run a specific straightforward GPGPU task like a miner, you only need to get one driver version properly cracked to unlock full performance, and keep using that until the hardware is no longer profitable to mine with.
And it seems pretty likely that a countermeasure bolted on to the drivers this late in the process will be defeated fairly quickly, and miners will go on a second shopping spree after clearing out the stock of the new purpose-built mining cards. (Assuming those even ship before the 3060 drivers get cracked.)
Miners don't care about driver updates so it's really a one-off, not endless, once it's running they don't have to touch it.
This just seems like a feeble gesture at appeasing gamers. What's interesting is that these "mining-crippling" changes will only apply to the 3060 GPU, not the rest of the RTX 3000 series. If it was a legitimate gesture, then Nvidia would've made it universally applicable.
Color me cynical, but this seems like a strong gesture at increasing margins on GPU computing by upselling using cryptocurrency and gamers as PR pawns/scapegoats
if Nvidia is selling more cards than they can produce/distribute at desired target prices they should produce more cards and improve their distribution networks..
Maybe it will offset the days it takes for a graphics card to pay itself by a few weeks? (e.g. previously RTX 3070 took x weeks to pay itself by mining 24/7, maybe it will change to x + 2 weeks?)
How?
Currently, average users can't get their hands on any of these cards because they're being bought in bulk by crypto miners. As soon as mining becomes unprofitable, the market will be flooded with cards that have been run 24/7 for months or years, which isn't a recipe for a good experience.
The average user can only benefit from getting access to these cards again. nVidia knows they can't afford to burn goodwill with their customers, whom they depend on to buy card upgrades every few years.
> Worse, their detection isn't going to be 100% perfect. What if After Effects does something that looks "kind of" like mining. Is it just going to be crippled? What will nvidia do then?
nVidia isn't a resource-limited startup. They have a lot of smart people. They're obviously going to test the firmware against common workloads.
Regardless, don't buy a card unless you know it does what you want it to do. These cards are all mining-limited from day 1, so it's not like the performance will regress suddenly.
They can only test against pre-existing software. If they start penalizing uncommon workloads, this kills innovation in many areas, videogames included.
In one of my projects I'm using GPGPUs for complicated optimizations in CAM/CAE. I don't use any libraries/frameworks. From nVidia's point of views, my compute shaders are extremely uncommon workload. Unwanted too, because I skipped CUDA and their vendor lock-in.
I'm not too worried about that. Nvidia has track record of releasing driver updates to address issues in end-user applications. If a legitimate application got accidentally flagged, people will report it and nvidia will whitelist them.
That approach doesn’t work in general. Look at “Manage 3D settings / Program settings” in nVidia’s control panel. They only have presets for very popular software like Adobe After Effects or Premiere, benchmarks like 3D mark, or AAA videogames.
But I'm concerned that a company can tell you what you can or can't do with your hardware. Where is the spirit behind the hacker community, Jailbreaking iPhones, the right to repair, the ban of tattoos that prevented the use of Linux on PCs,... That's a very concerning move from me.
Also, for a company that does morale :
(1) Nvidia's marketing is one of the best for making old products look obsolete and replaceable. Most gamers would do great with their old gen card, they just don't need to upgrade already. But they reserve cards for them so that they can create as much e-waste as they want
(2) they could just have prevented supply issues if their supply chains were diverse and resilient. But it turns out they aren't. And it's not something that was asked by the gamers
(3) for a company that's focused around gamers, they have already prevented powerful "crypto" GPUs from being repurposed into gaming gear, as featured in LTT https://www.youtube.com/watch?v=TY4s35uULg4
That said I'm not sure how effective this measure will be as in my understanding all that miners need to do in order to circumvent this throttling is avoid downloading the newest NVIDIA drivers (?). In addition, I'm sure miners will quickly come up with a crack that renders the throttling moot, so not sure if this will fix the supply side. Offering specific hardware to miners that doesn't compete with that for gamers seems to be a better solution, but again if NVIDIA isn't able to satisfy the demand for such hardware miners will keep buying regular GPUs and try to circumvent the software throttling.
I'd rather they have scaled up the demand. They've had years of this to take the higher profits and invest them. Even if people don't use them for mining, the demand for Machine Learning is there - instead GeForce cards have been banned for use in data centers for years. Not to mention that the crippling of 3060 would be trivial for mining operations to get around.
There are car factories shutting down due to lack of chips. The silicon for the systems that I work on has its lead times increased from the customary <30 to 50 weeks. The supply issues is worldwide, for all fabless semiconductor issues.
How do you suggest they scale up demand short of building their own fabs? (Which I hope you aren't suggesting...)
But in the context of "I'd rather they have scaled up the demand", where "they' is Nvidia (or almost any other fabless semiconductor company), it's not a solution because it's impossible for them to build fabs or to foresee and plan for the supply chain disruption that was caused by COVID-19.
This has been happening for half a decade, it is not remotely a COVID issue. And no, they can't build their own fabs but they can certainly invest (either directly or indirectly by securing larger orders at a higher price even more years in advance) all the premium so there's at more fabs at an earlier time.
Securing even more years in advance is a recipe for repeating that.
Unfortunately it happens all the time. For instance Nvidia had Quadro lineup, with nicer multi-display settings that to my knowledge was always just driver thing. They never ported those to GeForce line up, and people suffered with displays that wake up too late and all windows were messed up when you waked up your computer.
It hasn't happened to me lately so they might have fixed it, problem was that if one of the displays took longer to wake up then Windows didn't recognize it until it was too late. With Quadro you could force the EDI settings so that it faked for operating system that the display was there.
Everything does it - from Intel chips being binned by performance, Windows versions and # of CPUs, and pretty much any big software products.
Another way to see it is: it costs X to develop a line, and a vendor can charge everyone the same, no matter how they use it, or they can design and partition the line so users with different needs pay more or less. So, for example, a home PC with 8 cores is not paying for software that handles 128 cores, and those needing 128 core software pay more. When Intel does a run of CPUs, and tests them, and finds some can run at 4 GHz, and others only at 3 GHz, then they can either throw out the 3GHz ones and charge everyone much, much more for the single sku, or they can partition the same chip into bins for different prices, effectively lowering the chip cost for every single sku.
The alternative to pricing in tiers is to charge the 128 core user less, and the 8 core user more, then the majority of users get charged more for things they will never use. Or only sell the top of the top of the chip yield, making the per unit cost astronomical, since now your yields dropped through the floor.
Personally, I like market segmentation, since it allows me to only pay for the things I want, while others can buy the things they want. I prefer not being charged to support development of things I don't want or need. And in some cases it lowers the cost for everyone.
The most extreme instance of this that I'm aware of was when Intel sold a ~$50 piece of software that would unlock the rest of the L2 cache on a specific low-end processor model. They didn't say it might work on your chip; every instance of that model had apparently passed QA with the full cache enabled but shipped with a fourth of it (IIRC) disabled. So Intel wasn't saving any money or harvesting any otherwise-unusable parts with that model. They were just price fixing.
Product segmentation for software is also a different issue, because all the scarcity is artificial to begin with, with no natural underlying supply curves because there's no marginal cost to making another copy.
Marginal cost is not the only cost. It's disingenuous to treat it as such. It's also incorrect - making a copy for someone to access is most certainly not free.
And the economics is sound - partitioning markets allows sales at lower prices to many purchasers. There's simply no way around this, no matter how you feel about it.
>That's a pretty clear cut market failure
No, the same economics above apply - partitioning the market lets those less needs obtain product at a lower cost than those with more needs, no matter how the device/software is created.
Market segmentation replaces a single, ham handed consumer and producer surplus with a multi level one, allowing each side to capture more of the surplus for their benefit, leaving less of the summed surplus on the table.
Not every purchaser has the same surplus to trade. By splitting the market some purchasers that would not otherwise buy are now willing and able to do so, and the tighter or possible non-existent margins there are offset by selling to those accepting higher price points, almost always offset by extra capability.
And no matter how you slice it, commodity chips and software are so incredibly cheap compared to the value they provide it's hard to claim these prices are somehow not reasonable. For less than a week's pay I can purchase an incredible set of tools that I can then turn into orders of magnitude more value for me. Pretty much any professional making a living based on computers is in the same position.
Well, it's a good thing I didn't do that. I just said that the lack of any real marginal cost means that artificial market segmentation for software should be analyzed separately from artificial market segmentation for physical goods. And only one of those two analyses is relevant to this thread. The software stuff you brought up is a red herring.
This is akin to economic voodoo. There is no reason to do anything different for either one - they are products, customers buy them.
Once you want this artificial split, I guess based on your (incorrect) claim that there is no marginal cost (original claim) or lack of any real marginal cost (backpedaled claim) as the deciding factor, then when do you start to add a continuum of how you analyze based on a continuum of costs?
If there is a marginal cost 3/4 between your two cases, do we now need to interpolate your forms of analysis? All because you personally don't like how markets work?
Or why not simply use the same method, standard economics, to analyze them? And realize all these products have the same benefits to consumers via varied pricing?
>The software stuff you brought up is a red herring.
No, it shows how markets work. That you move your goalposts in the middle of a thread is the fallacious reasoning.
And you've failed to address that segmentation like this allows people to buy that could not have otherwise. Do you dispute that segmentation allows cheaper pricing for some consumers, and in some products, all consumers? Are you arguing for higher prices or just ignoring this empirical fact?
What you're saying is that due to some irrelevant production related factor, it's totally fine for a software company to price their product based on what the market is willing to pay, but as soon as a piece of hardware is involved, that principle doesn't apply anymore.
What about selling software that comes with a hardware copyright protection dongle? Under which category does that fall? The marginal cost of producing that dongle is not zero after all.
I didn't say it's fine. It's off-topic, because it's a different issue with fundamentally different economic forces at play.
I see that as a market success. People with lighter needs can purchase a neutered chip at a lower price. While people with greater needs can purchase an unlocked chip for a premium.
An alternative is to only have one chip, which is priced higher than the neutered version. Because of lost sales, fewer people overall buy the chip, so the price per unit goes up. This is a net loss to the consumer.
Another idea is to build two chips for each market. This increases design and production costs. And is a net loss for the consumer.
Or they could always just avoid making the more powerful chip at all, and stick with the volume seller.
This is a really common strategy in consumer products because it makes so much market sense. It sounds unfair when you think of it as, "paying for a product which you can't access the full potential". But it sounds pretty nice when you frame it as, "getting a discount for not using a feature."
If you're able to take e.g. an 8 core cpu, artificially disable 2 of the cores and sell the 8 core for $300 and the 6 core for $250, and still make a profit/break even on both of them then you could be selling the 8 core for $250 and your competition should be punishing you.
Intel are only able to offer a "discount" because their profit margins are so high because of lack of competition.
It doesn't work that way. The $300 premium chips are where all of the margin is, and is what finances the whole operation. If they production costs of in your hypothetical chips is $235, then Intel would be making $15 on the cheap chips, but $65 on the premium ones. Which means that one premium chip is worth the 4.3x the margin of a cheap one.
Losing that premium segment is a big deal. Even if it is only 10% of the volume, because margins are so high, it ends up being like 40% of profits.
This isn't new. Just about every consumer goods market works like that, from peanut butter to cars. You need the middle tier to have enough volume to minimize production costs, but you need to find that premium offering to actually make money.
A fundamental principal of business is that a successful one needs to earn a profit. And that profit needs to be high enough to continue operation.
No business needs to earn a profit. By definition profit is surplus to operating costs (and I include a reasonable emergency fund in the category of "operating costs").
This is the first time the ends have even remotely justified the means.
In a way it is, the extra capacity gets used to manage wear. So it can distribute load over the cells to make the overall battery last longer.
Same thing we have in flash storage where is more flash storage than you are able to use but as parts die, the reserve gets used to replace them.
It actually is not, because those cards sold to miners don't just magically vanish, when crypto prices finally fall the miners will flood the market and the prices for new cards will have to be depressed as a result.
Mining sales should largely be thought of as a "pull-forward", they represent revenue that you're not going to make in some future quarter when everyone is buying used miner cards for 40% under MSRP. And NVIDIA themselves don't crank the chip price for OEMs (all of that was locked in a year ago) so they aren't making more money in the short term to make up for it.
That is specifically what NVIDIA is trying to address here by banning mining on gaming cards and creating mining-specific cards that can't be gamed on. Now cards that are sold for mining can't flood back and hit their future revenue in the gaming market.
IMO people obsess far too much on turning games up to the top setting when the actual result is hardly different, and certainly not worth buying a top card yearly for.
The 3060 Ti was already NVIDIA's mining efficiency leader (per W and per $). I think the 3060 would be in the same ballpark, if not better, which I assume is why NVIDIA is doing this.
Presumably they will do nothing to the existing cards (in part because it would solve very little if you could just use old drivers to get the unthrottled MH/s).
The behavior on display in this thread is highly reminiscent of the behavior of people demanding that open source authors do free work for them.
> RTX 3060 software drivers are designed to detect specific attributes of the Ethereum cryptocurrency mining algorithm, and limit the hash rate, or cryptocurrency mining efficiency, by around 50 percent.
Yes, it is. The hardware has capabilities, and Nvidia is artificially limiting them in certain circumstances based on what they consider "acceptable" use.
But that is not taking anything away form you.
If you buy something, knowing what software is on that thing, you are getting what you were sold.
You were told that this would be the software on this device, and you bought it anyway.
Just don't buy something if you don't like what that product is. You were told what software is on that device!
No, they are simply selling you a product that you are voluntarily choosing to buy. You are perfectly within your right to not purchase something that you don't want.
> And are granting themselves the ability to alter the value of my purchase after the fact.
False. You seem to be unaware of whats going on. All of this stuff is being told to you ahead of time. Your software is not being modified after the fact.
You are simply buying a product, that comes pre-installed with certain software. If you don't like the software that is installed, up front, beforehand, then don't buy it.
Nothing is being modified after the fact.
> What's to stop them from deciding to update the driver
Thats not what happened. No updates are out. Instead, this is about new hardware, that has not been sold to anyone yet. Nothing is being updated.
> they were sued over this and ultimately settled.
Well then it is a good thing that nothing is being modified after it was purchased then isn't it? This only applies to unreleased products, that you have the choice of not buying.
No, they are imposing restrictions on use after it's purchased.
> False. You seem to be unaware of whats going on.
You seem to be deliberately misrepresenting the argument. No one is claiming that this has actually occurred yet with this product. However, this is their stated policy, that they will artificially restrict what owners of the product can do via software drivers. Since it's via drivers, they can update them to further modify the capabilities of the devices after purchase. They have already done this in the past, when they modified drivers to disable capabilities if they detected more than one display adapter present (also to try to limit use in crypto mining).
> Thats not what happened. No updates are out. Instead, this is about new hardware, that has not been sold to anyone yet. Nothing is being updated.
No one is claiming this is currently happening. But they have a history of trying to control what people do with their hardware after it's been released and sold. So nothing has been updated YET.
> This only applies to unreleased products, that you have the choice of not buying.
Yes, it's good they're announcing their intention to change the functionality of devices after they've been purchased.
No they do not! Nobody has any 3060s yet! The cards become pre-installed with this software. If you don't like this software, don't buy the card. It is not an update! It is pre-installed software that you can choose not to buy.
> No one is claiming that this has actually occurred yet with this product.
So then you agree that nothing has been changed after the fact. Got it, that is my point. The cards are not sold yet. And you are being told, ahead of time what software is on those cards.
If you don't like that software, don't buy it.
> they can update them to further modify the capabilities of the devices after purchase.
Nothing has been updated on these specific cards after the fact. The cards are not sold yet. This is just about unsold devices, that you can choose to buy or not buy.
> No one is claiming this is currently happening.
So then you agree that your device has not been limited after the fact, and that the drivers are pre-installed? Got it. Thats my point.
> their intention to change the functionality
No. It is pre-installed software, that comes on a device, before you purchase it.
The cards do not come preinstalled with drivers. Drivers are installed in the OS and are updated regularly while the hardware is in support.
You have a very odd obsession with future/past here. These are statements that Nvidia has made about their plans for this hardware. Unless they change their mind, they will do this. It doesn't matter that it hasn't actually happened yet, it's their stated policy for these cards, as well as something they've done in the past without warning. And even if you buy the card knowing that the hash rate is restricted, there are a myriad other uses that they can now arbitrarily restrict at their whim.
> No. It is pre-installed software, that comes on a device, before you purchase it.
No, it's software you install separately. Since the drivers are required to have the card function, you must install them, and they can be updated at will by the manufacturer to remove capabilities.
When you are buying this kind of hardware, you primarily look at the hardware specs to understand the capabilities. In this case, those specs can't be trusted, because Nvidia is deciding what sort of code you're allowed to run on hardware you've purchased.
But this isn't about you being sold something, and then it being change later.
Instead, you are being sold something, knowing full well what software is available.
> Drivers are installed in the OS and are updated regularly while the hardware is in support.
I have misphrased things then. What I should have said, is that you are being sold these cards, and you know ahead of time, that it is only this software that is available.
That is very different from buying something, under the expectation that other software, that has already been released, can do certain things, and then that software changing.
You know full well what software is going to be available, and you voluntarily choose to buy it, given that knowledge.
That is extremely different than if I bought something, and was given software X, and then that software X turned into software Y later on.
Not really. Even if I buy a card and am content to live with the 50% reduction in hash rate (assuming that's why I wanted to buy it), Nvidia still has the ability to change the drivers and further restrict the hash rate at their whim.
> That is extremely different than if I bought something, and was given software X, and then that software X turned into software Y later on.
The problem is that we can't know if Software X will turn into Software Y later on. If I buy a card because it has certain specs, I expect to be able to exercise those capabilities. I don't think Nvidia should be able to try to restrict what I choose to do with those capabilities.
(2) Big Mining operations actually write their own drivers
(3) Huge Mining operations even use custom boards
(4) I heard that that Nvidia has been selling silicon directly to the big boys (https://wccftech.com/nvidia-allegedly-sold-175-million-worth...)
(5) With this action I expect its only the little guys getting removed from the game here (1) a huge benefit for the big players (2)(3) benefitting the corporations biggest customers (4)
I am curious though whether they could be used as "second cards" for systems so that I can speed up Blender render times...
This is a great video: https://www.youtube.com/watch?v=sg__YdXbRtI
Strong claims require strong evidence.
What this tells me is that they just didn’t care enough to lock this down hard.
What are the practical reasons supply hasn’t kept up? Obviously, a fab is expensive but they have to be just printing money with these things in the current market. This really goes across the board with semiconductors as CPU supply issues are outrageous right now as well. Why aren’t more fabs being built? Is this just plain and simple greed?
It's a further shame that they would artificially hinder their hardware for such purposes.
I will avoid nvidia in the future as a result of this sort of attitude, and I don't even mine cryptocurrency. I have no dog in this race other than a philosophical one.
(... assuming the miners can't trivially patch to the drivers to re-enable the functionality)
And don't get me started on DVD region locks.
I somehow doubt these drivers will be enough to stop it, however.
What, to you, is the difference between a scalper and a dealer?
Regardless, even if I were losing money, 100% of the waste heat from my computer is doing actual useful work in heating my room. No electric space heater would be more energy efficient at that, although it would be quieter and put out more heat per unit area.
1. https://www.notebookcheck.net/Nvidia-GeForce-RTX-3060-laptop...
Though not impossible, NVIDIA has been doing application detection for ages for the purpose of tuning the performance of the latest popular games. So as long as your scientific calculations don't calculate too much Keccak-256 hashes, they should probably be fine.
The case for this kind of video card "DRM" is actually stronger because miners either have to gamble and buy cards they know they won't be able to use until (or even if) a "crack" is available, or wait until a solution is available before buying them.
If the miners wait, this is a win for gamers and Nvidia PR because the cards will be bought by gamers since cards have a limited supply unlike video game licenses.
This is also a fantastic move.
Personally I am very very VERY much okay with this. This will keep consumer cards in correct demand. And if miners want to throw away money for mining, they can.
People talking of modding, its a high effort and high risk. Again, I'm okay with that. The cost of modding will likely be the same as just buying a specialized card. Overall mining is a horrible environmental impact.
(I am an ignorant layperson on this topic.)
If Nvidia has the capability to make a differentiated mining product, shouldn't that be enough to get miners to go for it? Why halve the hashing rate on the "gaming" GPU artificially? If it is a supply issue, shouldn't Nvidia, as a company, try and solve for it directly (maybe it is really hard?), instead of making their products unattractive to a certain set of customers?
The same criticisms were given last time nividia tried to make a mining card and obviously it didn't work as last time they didn't need to use artificial limits.
This makes the problem of designing specialised mining cards equivalent to designing fast and efficient DRAM, an area where the consumer product is the cutting edge. This was by design to reduce hashrate centralisation.
Even the latest ethereum ASICs are only 10% more efficient, cost much more and can only mine ethash and its variants.
So as it stands I see no evidence that suggests mining cards will be cheaper with better efficiency, especially when the cryptocurrency community slowly moving away from mining.
I you refer to the infamous Error 43 on Nvidia, it can be worked around by setting some QEMU flags.
“RTX 3060 software drivers are designed to detect specific attributes of the Ethereum cryptocurrency mining algorithm, and limit the hash rate, or cryptocurrency mining efficiency, by around 50 percent.”
All decisions have tradeoffs, and that is a small one.
That’s on the order of years for this problem. I would imagine that Nvidia doesn’t think it could survive the reputational costs of charging market clearing prices for their cards for years - effectively proscribing their core gamer market for the entire period.
Sure, that’s what’s happening now, but it’s the fault of scalpers not a business decision by Nvidia. The scalpers get the ire of the core market, and Nvidia gets to look like it’s doing its best (while guaranteeing all the inventory it can make will sell)
It’s the same reason your favorite band charges a low face for tickets that doesn’t reflect demand - they don’t want to be the bad guy excluding poor people from the experience, they just let the scalpers do that.
Apple outbid everyone (and given their budget, will continue to do so.), so the rest of them are fighting for whatever capacity is left.
TSMC is expanding, but new factories take years to build.
So the issue here is mostly the overreliance of whole world on TSMC and Samsung.
Everybody else, is in the same boat as Nvidia. So this is the norm not the exception.
And NVidia is not in a mess, they are having record profits.
I think Nvidia could borrow or raise basically any amount of money if they had capital-intensive projects to invest in that would solve current or future supply constraints. Retained earnings are not the only way a company can invest in itself.
The problem isn't capital, its fundamental inelasticity of supply for semiconductors. Fabs (or capacity in fabs) can't be spun up or down quickly, nevermind cheaply.
People who insist that prices are the only control point tend to be those most unfamiliar with the practical dimension of finite resource distribution.
https://www.theverge.com/2020/7/2/21311332/bmw-in-car-purcha...
Its an opportunity for someone to hack the drivers
Business opportunity because other miners will pay for the driver (or the mining software you make has the 2% mining fee to your address) or you can be the only one that has the driver and buys up all the rtx cards yourself if you have access to capital
vertically integrated is harder of course but could have the biggest payoff
For example, for Ethereum, mining calculators assume the block reward of 2 Eth, but for the past year each block has had like 5 Eth on average because the transaction fees are so high, extended periods of time with 7-10 Eth per block. So its double and triple the estimated payout if you are just taking a passing look.
Now that is a great question. If they were purely doing this for their supply issues and not for control they would be stating that this is just temporary.
> RTX 3060 software drivers are
> designed to detect specific
> attributes of the Ethereum
> cryptocurrency mining algorithm,
> and limit the hash rate, or
> cryptocurrency mining
> efficiency, by around 50
> percent.
I'm saying this as a person who wants to use computers for something useful, like playing games.
Pivoting and creating a mining division also doesn't make sense - mining companies take on a lot more risk from volatile market prices and need to have a lot more cash on hand to sustain the long periods of unprofitability.
They could subcontract the operation of the mining equipment to third parties so they make a profit immediately selling the mining hardware, and skim as much mining profit as possible while their subcontractors assume all the risk.
(Just using nvidia as an example, but this model would work well for anyone making hardware that's capable of mining at scale).
https://www.tweaktown.com/news/76468/cant-buy-new-ampere-gpu...
> if crypto [mining] is really viable and long-term profitable
This is still a big if, with the most profitable gpu mining crypto moving to proof of stake in the next few years it's unclear if gpu mining will be as viable.
What you suggest already happens, companies like Bitmain who can invest millions to develop custom asic designs and buy directly from the silicon manufactures, contributing to the silicon shortage. Miners are buying commercial hardware becuase the allocation at the factories doesn't meet their demand, if it did graphics cards would just be sold at higher MSRPs.
I mean, imho most people are using cryto as an extremely complicated casino, and it's not even that fun. At least games are honest about their purpose as a form of entertainment and singularly focused on that.
You wouldn't have to monkey around with hardware freedom and even if a person decides to re-sell it at least it had the chance of being used by a "real gamer" first.
Really, any consumer who isn't buying the card for mining is the intended beneficiary here.
The popular gpu algorithm is ethereum, who's heading algorithm (ethash) is memory hard, so muhh of the performance relies on the speed of commercial DRAM chips so the difference between ASICs and GPUs are not as great. Along with the lower incentive as ethereum plans to move to proof of stake ASICs for ethereum are not as popular or common.
Disregard everything below...
I hate mining, I love my (old-ish) Nvidia GPU, and what this will do for the availability of cards and reliability of cards on the second-hand market is great.
But, FUCK this. Nvidia just removed a feature from a product customers paid for, and then sold it back to them in a new product. This is like if Apple disabled Wifi on old iPhones to force you to buy a new one.
It doesn't matter whether the reasoning behind this particular decision is something you agree with, a company should not be allowed to get away with doing something like this.
I hope a class action lawsuit comes out of this and Nvidia loses big, because a world where a company can do this by adding sufficient spin to the decision, is terrible for consumers.
Isn't the 3060 a new product?
This isn't like Sony removing Linux on the fat PS3 post launch.
This will be like Quadro: Same chip, a couple of feature flags flipped, then sold at several times the price to price-insensitive market segments.
Nvidia have noticed miners will pay $1000 for an MSRP $500 card, and they want to take a cut without alienating the gaming market too much.
Many people in the community don’t like ASICs in crypto because they believe their high cost and rarity creates a hurdle for mass adoption. Every kid playing games on their computer has a GPU, but it’s much harder for that same kid to get their hands in an ASIC. Mass adoption by all kinds of people, not just already rich corporations, is important to the security and the mission of crypto.
In response, Ethereum was designed to have an algo that stresses your RAM rather than your processor. This reduces the advantage of ASICs tremendously. You can’t drive RAM much harder than GPUs already do. And so, Eth and many other memory-hard cryptos are still mined on GPUs.
Is it likely that this affects something more than just mining? Or is this hobbling specific enough to leave the card unhobbled for every other need?
As for these mining cards, without the resale value they might as well be ASICs.
Limiting to justify a new product. Like they do to fp64 rates and virtualization support.
>RTX 3060 software drivers are designed to detect specific attributes of the Ethereum cryptocurrency mining algorithm, and limit the hash rate, or cryptocurrency mining efficiency, by around 50 percent.
Nice, one more reason to get the open source driver moving.
edit: formatting
(via https://news.ycombinator.com/item?id=26180342, but we merged the threads)
This happened for bitcoin. If you don't have latest ASIC you can't even dream of mining bitcoin unless your electricity costs nothing.
And even then you are still better off investing in ASIC as opposed to gaming hardware.
The fact that people waste electricity gaming drove the development of GPUs that serve researchers. Maybe more general mining software will end up cheaply in hands of scientists as well?
ASICs are ASICs, application specific.
Mining crazes appear quickly and dissappear quickly leaving little opportunity for the much slower silicon pipeline to catch up. The demand for cryptocurrency ASICs in the last mining boom was much higher than the production. Thus even if they could meet demand that would just result in less allocation to gaming hardware.
The reason it has been successful for bitcoin is becuase sha256 utilises only half of the gpu whereas memory hard algorithms like ethash utilise almost all of it. Designing a specialised, power efficient ASIC only matters when the return isn't ludicrous.
During mining bubbles it makes a lot more sense for miners to purchase general hardware to use during the craze and sell afterwards. Until these bubbles aren't as crazy that's not going to dissappear.
The graphic card manufacturers would, of course, have to optimize their hardware and software for these situations, to make the gamers happy.
So, a software solution to a hardware issue? Well, this is going to work exactly like expected. Can somebody spell "cracked games"?
I’ve never mined Ethereum, but this really upsets me.
Imagine a new driver, for hardware you already own where you couldn't watch cat videos (extreme example) in 4K but only in 360p :-)
On the other hand if you're well informed before buying such a card, that it will have reduced performance for crypto (or whatever) I guess it's ok and you can select something else.
We have to be realistic here and understand that on one hand NVIDIA hates miners since they're creating a shortage of cards for gamers but on the other hand they're the ones buying all new cards straight away for mining.
I think this is a marketing trick so that both gamers and miners are happy, RTX line for gamers and CMP line for miners, win-win for NVIDIA :-)
But they won't magically have twice the amount of chips on hand, will they? So either they make a lot of CMP cards and too little RTX cards, and gamers will still struggle to get one, or they'll make sufficient RTX cards and too few CMP ones and miners have a big incentive to get around those restrictions (and will, in all likelihood). I guess they can mark up those CMP cards and include some mining-friendly features in the future for some additional revenue.
This is not the start of NVIDIA dictating what you are running, just a new symptom of it.
One "excuse" miner had for buying GPUs was that they could sell them 6-12 months to gamers.
> RTX 3060 software drivers are designed to detect specific attributes of the Ethereum cryptocurrency mining algorithm, and limit the hash rate, or cryptocurrency mining efficiency, by around 50 percent.
Why just Ethereum? This cannot possibly limit the sales of 3060 alone because there are a zillion cryptocurrencies. People will just use it to mine for other crypto currencies right?. Also 50% reduction. They wouldn't hardcode this percentage. The number obviously needs to be able to go up to 100% (no limit) and down to 0 - ie block it completely.
This effectively means 2 things:
1. Their detection algorithm is "upgradable". They wouldn't do that just for Ethereum. 2. That they can block any kind of workload they can detect.
Should we allow them to have the power to do that?
No, I think the belief is that they need to keep at least a minimal base of support among gamers and developers, for when the inevitable crypto crash comes and they need to go back to their original customer base
But then how do they capture the "just gamers"? They would have to additionally sell a neutered card just for them, and that is exactly what they are doing here
Nvidia is kind of throwing gamers a bone while figuring out how to profit from cryptocurrency mining.
I guess it makes business sense, but I still hate the externalities of cryptocurrency mining.
It'll be interesting to see how games perform on the newly gimped cards people already own. Can they firmware-gimp a card to reliably stop mining without affecting games? We'll see!
I shouldn't be surprised - this is what Nvidia already does with their Quadro vs. consumer line. Heck, I owned a Sound Blaster Live! which was functionally equivalent to the EMU APS and used the hacked drivers for years back in the olden days.
While their move is questionable (I don't judge, but recognize that it's questionable), there is a situation out there that is 1. specific and 2. concrete (gaming cards that can't be purchased for gaming), which is unlikely to apply also to machine learning.
Maybe? It would make sense because ML use cases are professional and there's a lot of money to be made there.
This isn't without precedent. Quadro cards have always been 99% GeForce cards with some special bios magic to de-gimp them for CAD or VFX applications. They charge 4x the price because they know that workstations running $50,000 a seat software can afford the premium.
What a waste of time for everyone
The people that are doing the best right now are almost certainly the ones with fully depreciated equipment.
This would be better than crippling the product via software drivers.
LE. I see you meant NVidia should be making the ASICs. In that case, that's exactly what they're doing, isn't it? This CMP line is exactly that.
As for a reason, why should they be able to dictate what the hardware they release is used for?
Limiting mining performance on a consumer GPU is no different than limiting certain professional or machine learning specific performance features.
It's perfectly fine to offer a different product that's designed for better performance but to artificially limit product A so that people are forced to buy product B (that's probably more expensive) is messed up.
This is no different.
As long as you know what you're paying for, segmentation is fine. It allows consumers to pay for what they need.
https://www.tomshardware.com/news/intel-xe-hpg-graphics-unve...
Linus was absolutely correct.
What next, class action lawsuit against Apple when they announce the next iphone won't have any external ports?
first, this is very obviously because they noticed cloud operators are offering these RTX cards and they do not have a 10x-the-price-enteprise-quadroRTX yet.
what they think will happen:
they will blame miners for their production issues, make RTX cards less attractive for cloud operators, sell a truck load of quadro cards to them while offloading RTX cards to actual gamers who then generate revenue in other ways (they do not have those free game bundles deals for nothing)
What will happen:
nerf cards, lose sales for gamers and ML startups.
impact production lines even more with multiple models, lose sales for cloud operators.
cause actual miners (not their strawman 'oh the prices are up' for their production issues) to either reverse engineer the drivers, or just keep the old drivers running and buying the now-cheaper RTX cards, compounding the production capacity.
game studios that started to use those features to speed up crowd AI etc will just slap a "AMD Recommended" on their games.