You may have noticed that it’s kind of hard to find any new graphics card as of late, since supplies are limited for a whole variety of reasons. For the launch of its upcoming RTX 3060 GPU, which might prove to be a relatively affordable and capable upgrade for many, NVIDIA is going to try and do something about the shortage – by crippling the card’s suitability for cryptominers.
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.
[…]To address the specific needs of Ethereum mining, we’re announcing the NVIDIA CMP, or, Cryptocurrency Mining Processor, product line for professional mining.
CMP products — which don’t do graphics — are sold through authorized partners and optimized for the best mining performance and efficiency. They don’t meet the specifications required of a GeForce GPU and, thus, don’t impact the availability of GeForce GPUs to gamers.
It’s a good first step, I guess, but I feel the market is so starved at the moment this will be a drop in the ocean.
I am among those who are extremely annoyed at what’s happening with the GPU market. IMHO crypto currencies had a long list of cons even before this GPU mess. But now the crypto currency industry is actively displacing regular consumers, AI applications, medical applications, autonomous robots, etc. The stupid thing about throwing more and more hardware at crypto currencies is that are coded to automatically respond by increasing difficulty, thereby completely offsetting the new computational value of additional hardware, which wastes more power for no net gain.
Just think if all the crypto miners could have a gentleman’s agreement in place, they could make the same “coins” and profits using a fraction of the hardware and energy. All the GPUs they’ve been wasting could go back into the market. The cryptocurrencies would be no worse off. It obviously isn’t possible because the miners are all competing as opponents, but it’s just a shame that the world has to tolerate their waste while being deprived of high end technology.
It seems nvidia is under intense pressure to do *something*. Heuristics that halve performance…I can’t see any way this could go wrong. /sarcasm.
Is this going to be better than 3070-3090? If not, the miners are just going to continue to buy those cards.
Also, the problem with ethereum in particular is that they intentionally try to avoid specialized hardware. – in the past this meant specialized ASIC solutions, but in the future either ethereum or other crypto currencies might take the same stance for nvidia CMP by deliberately making nvidia’s specialized hardware useless and once again favor consumer GPUs. From their perspective, doing crypto on consumer GPUs is good.
https://www.bitcoininsider.org/article/22421/lets-keep-ethereum-asic-proof
I don’t know just how specialized Nvidia CMP is going to be, but the community might respond by changing the currency in order to keep standard GPU mining profitable. So I’m skeptical that the crypto miners will be on board. Even assuming they agree with nvidia CMP, how many resources is nvidia going to need to divert to build these crypto specialty cards? Give that they can’t meet demand today, how is adding a new sku going to help?
I don’t think this is really going to work until supply catches up to demand, but as long as crypto profits are increasing the demand is near limitless.
Crypto Currencies are hideous monstrosities. In addition to creating a materials crunch, they require hideous amounts of power to mine and transact. In an era of global warming what the hell are we doing inventing new, exponentially more inefficient, ways of paying for things?
It’s all about trust. No mining + fixed amount of money (1) = someone owns all the money initially. No mining + unrestricted amount of money (2) = fiat currency, someone owns all the money forever (can devalue existing money at will). Mining (3) = money are created and distributed gradually, without us having to trust a single person/organization with all the world’s money.
The main issue with mining is that practical implementations are all very expensive. They tend to divert human resources, energy, equipment away from the productive economy and result in unnecessary environment pollution. Despite all the imperfections, mining is currently the only solution we know to making money more fair (be a measure of value and not means of controlling people) and the only implementations that are available to us are gold and cryptocurrencies.
Ideally we would connect money supply to something fair (like 1 coin given to each newborn person) or useful (like productive work). Unfortunately, it is not possible to implement such system without making someone a gatekeeper of money supply, which is the same as option 2 (fiat currency). Also, we would quickly discover that “fair” and “useful” is less important than “richer”. With these two examples we could end up with people farms or companies being created for the sole purpose of earning coins rather than doing anything remotely useful.
Which brings the question: Do we want a decentralized currency? Aren’t governments supposed to have monopoly control over currency?
It depends what you want to do with it. If you are saving for retirement gold is a great option. Cryptocurrencies may also be good, it just a bit early for them, which makes them rather volatile. For shopping we don’t have much choice yet.
As for “supposed” – it depends to whom. Yes, to a government supplying fiat currency everyone is supposed to use it, they may even resort to stronger arguments at some point. But in many places in the world people still have a choice of what to do with their savings and the market trends are rather obvious.
The decentralized part is fine, but the cryptocoins don’t have any way of self regulating their own prices. They are basically guaranteed to always have massive hyper deflation and make early adopters rich. Add some way to regulate the price and stop wasting gigawatts of electricity, then it might be usable as a currency instead of a speculation vehicle.
@ndrw
I prefer to wear gold not use it as an investment vehicle. Sadly 99.9% of my jewellery is costume jewellery because investors push the price of gold up but it doesn’t bother me that much. In fact with jewellery the issue can be not what materials it is made of but its design. You can buy gold any time you like but finding a good design you like or replacing a design if you lose anything? That can be hard to impossible. One of my favourite matching jewellery sets I managed to put together by chance is made of plastic and utterly impossible to source.
As for traded cold there are more gold certificates than actual gold in circulation much like I doubt if everyone realised their Bitcoin assets they would get the paper value of them. In both cases there would be rush and a large number of people would find their assets are devalued and in some cases worthless.
@HollyB
Gold has almost no uses beyond being money and that’s a good thing, we wouldn’t want e.g. food to become money, would we. The only thing it is good for is coating electrical contacts, a fairly recent development. Jewellery is a derived use – its main function is showing off the owner’s status, and what material is better for that than the money itself. And yes, you can substitute or augment the material with a design, rarity, a brand or a story – many brands of luxury items have discovered it long time ago.
Gold is money, not an investment, it is expected to store the wealth, not provide any returns. If it does, it is only because of price fluctuations and currency depreciation. To be fair, currency depreciation is so strong now that I can understand why people consider gold or even normally depreciating houses an investment. Gold certificates are derivative instruments and as such they are as far from owning gold as any other derivatives.
Bitcoin is very similar to gold in the way it was designed to function as money but there is a huge difference between them. Whereas bitcoin hasn’t gone even through a single boom-bust cycle and is still at an early adopter’s stage, gold went through numerous crises, wars, wars fought for it, rise and fall of civilizations, all this over a period of thousands of years. This is how you get from a idea of “let’s make X our money”, through “let’s gamble against X in a corner of the marker”, to “X is owned by billions of people who store and measure their wealth with it”.
@ndrw
I’m liking the earrings I’ve been wearing this past few weeks and today bought a similar type of design off Ebay for less than lb5. Yes it’s another piece of “gold coloured” jewellery probably made out of some kind of alloy and the closest it has come to real gold is the flight over from China.
In all honesty I’m not sure men know how to value women’s jewellery or if men even notice half the time. Quite a few people Asians especially will put up with any old crap if it looks expensive and the Chinese are very into status enhancing material goods. Not that the West or anywhere else is fundamentally that different. I have my eyes on some Indian jewellery. It’s a touch on the expensive side for costume jewellery but the price reflects the amount of work involved in production and slow moving stock levels so fair enough. One thing I do know is modern jewellery in almost all its forms would have cost a ransom back in the day. That said the perceived value is less in the piece itself but more how everything is put together. There’s a halo effect from what it’s being worn with and who is wearing it.
Gold has intrinsic properties which make it valuable both aesthetically and for its electrical properties but I don’t care for it as a currency or investment vehicle.
ndrw,
That was the theoretical basis for P2P crytpo mining, but it quickly became institutionalized. Normal end users can’t even buy the hardware right now. There’ve been times when the network were so skewed that 50% of the bitcoin hashing capacity was in the hands of only two people. So that’s not really ideal. And the direction of the network is largely determined by a small number of developers who update the software.
I think that was the novelty of the idea behind it, but in real life it hasn’t turned out that way. The people with more real-life resources can make so much more wealth out of cryptocurrencies than normal users mining at home. Modern economies are pretty imbalanced (the top 1% take 50% of the wealth, but with crypto I wouldn’t be surprised if the top 1% of bitcoin miners take 99% of the wealth. So IMHO it’s been a failure in terms of democratizing money.
I know, that’s the difference between ideal vs real life. I do understand the appeal, especially at first. I participated in a joint mining teams because the hashing difficulty had become too great for individual miners, but it was a fly by night scam and they ended up stealing all the coins, it wasn’t that much for me but still it leaves a lot to be desired. And don’t forget there have been times when bitcoin was so backlogged it become impossible to process transactions in a reasonable amount of time, even for days on end. Even on a good day there are no real time confirmations; you’re looking at 20minutes-1hour to be reasonably confident it won’t get reversed by the network. And all that mining goodness is paid for through astronomically high fees > $13 today and this is only going to get worse if these currencies become more popular.
https://privacypros.io/tools/bitcoin-fee-estimator/
Honestly I don’t think it’s going to get better for crypto currencies. They’re going to continue to be institutionalized, incur high fees, impractical delays, and to top it all off they’re wasting our planet’s limited resources.
@subsider34 > hideous amounts of power to mine and transact.
That always puzzled me. One of the stated goals of the “cryptocurrency movement” (?) was to “replace ordinary fiat currenct and do away with the power of governments, central banks, meddling politicians, etc. But if the whole *idea* of CC depends on the “traditional economy”.. How exactly is it “better”?
You’re in luck! This situation isn’t likely to last much longer, and not because of anything Nvidia is or is not doing.
I suspect that the biases implicit in your first sentence heavily influence the rest of this paragraph.
Are AI applications displacing hardcore gamers as well? Or perhaps medical applications are so vital that it is actually greedy gamers displacing those important uses? Who gets priority, and why? It is only because you place a low value on cryptocurrencies in general that you could make such a statement, and go on to characterize the energy used as waste. But why is it for you to determine what uses of hardware are valid or not valid, or what constitutes waste? For many people, the concept of a maintaining a global ledger not under the direct control of corruptible human interests is a valuable enterprise. Many people consider that alone to be a net gain, whatever the costs.
Leaving aside the continued disparagement of the industry, as well as the horrifying optics the phrase “gentleman’s agreement” brings to mind, there are alternatives to the currently favored model of Proof of Work (POW), which is the model that requires intensive computation and therefore GPUs/ASICs that consume large quantities of power. The most promising, Proof of Stake (POS), has different kinds of problems that come with it, but has so far shown promise in protecting the integrity of small to medium size chains. You can run staking on any general purpose CPU with minimal resources, and GPUs and ASICs do not help in the slightest.
The real test for the high-demand, large chain viability of POS will actually be the cryptocurrency that is causing the current brouhaha: Ethereum. It is undergoing a transition to version 2.0 as we speak. Once completed, the Ethereum chain will be entirely secured by POS, rendering all hardware mining schemes for the platform obsolete.
https://cointelegraph.com/news/eth-miners-will-have-little-choice-once-ethereum-20-launches-with-pos
Once this happens, we still have Bitcoin, but it has long been unprofitable to use anything but ASICs (lots of them) for mining Bitcoin. Continued POW mining by Bitcoin still carries the burdens of power consumption, but it is unlikely to ever be an issue for GPU supply again. Even smaller ASIC-only pools have found it difficult to compete and stay profitable. The few remaining alt chains that use POW and that can be profitably mined with GPUs are smaller and far, far less popular than Ethereum, and will likely not produce this level of demand for GPU hardware for the foreseeable future.
Given the pressure already on the industry regarding power consumption, I suspect many of these chains are watching to see what happens with Ethereum. If the transition to POS is successful, and Ethereum can withstand attacks against chain integrity under the POS model, it is possible that some of these other chains might consider making the switch themselves, further reducing demand for GPU based hardware.
You have this exactly right. They are under pressure, and to formulate proper anti-mining countermeasures would take time. So this is a triage solution. There is the definite possibility that there will be overlapping computational patterns between other applications and cryptocurrency mining, causing false positives that result in lower performance on these other applications. File this under Be-Careful-What-You-Wish-For if you must, but within the scope of my limited understanding of the tech involved, I’m not sure what else they could do from the tech side to effectively address the situation quickly.
My guess would be that they know this, and I would expect it to reflect in the product specs. Omitting the logic and circuitry necessary for image display, along with other components not needed for mining, should allow them to do things like add more cores to beef these cards up and make them more attractive, without substantially increasing the cost. Long term, they could also add specialized parts that can speed up cryptocurrency computations. However, adding special parts carries additional implications that I describe below. Given the impending move to POS for Ethereum, I think this would be a foolish avenue for Nvidia to pursue.
I really don’t think this is an issue long term. Ethereum transitioning to POS effectively puts this whole problem to bed, and it is in process. But even allowing that another POW coin causes the same situation down the road, the problem set is different from the one Nvidia is faced with now.
ASIC resistance cannot be understood as a set-and-forget proposition or binary condition. It would be better described as an arms race between the coin devs and the ASIC devs/users, who naturally do not like being excluded from the mining equation. So far it has worked out, because of the highly specialized nature of ASICs. But general purpose GPUs are a whole different animal. This is why you can jump between mining different coins on a GPU, something that is usually impossible with ASIC equipment.
To do blocking on specific kinds of equipment, you cannot simply detect the hardware – someone will find a way to spoof this, probably on Day One of release. So you’re back to heuristics and, unlike ASICs, making specific computational paths more “expensive” won’t help you with GP-GPUs. The qualities that give ASICs a speed advantage over GP-GPUs prevent rapid, iterative warfare against the coin devs, in that the introduction of new countermeasures to a coin’s code demands a new generation of ASIC hardware be developed. Not so with GP-GPU based equipment.
Really, they only have a shot at blocking these cards if Nvidia starts adding specialized hardware to supplement the GP-GPU. This is the problem you alluded to earlier. But to start designing and adding special parts will take time and resources, and puts Nvidia in direct competition with ASIC manufacturers. I don’t think it is a market they really want to be in – at least they haven’t demonstrated any great enthusiasm for it. I certainly don’t imagine it being very profitable for them in the big picture, especially since specialized hardware could only target specific blockchains, like with ASICs. Additionally, if they were to pursue this, the cost of production would go up, reducing the probability that people could use them profitably for mining anyhow.
Well, cross your fingers that POS continues to work as designed when it is scaled up, and that the Ethereum 2.0 rollout proves to the industry that POS is a viable alternative to POW. There are certainly naysayers inside and outside of the industry. But despite what you appear to hope for, I don’t think cryptocurrency is going away, even if it was 100% outlawed across the globe, which does not seem to be in the cards. The horse has left the barn and, for better or worse, this technology will be with us for a while.
Nvidia is making the right call here, in my opinion, but not for the reasons I’ve seen many people give. Gamers, or the other market participants you mention, have no special claim to this hardware. Nvidia owes these people nothing, despite what some people appear to believe. They are a hardware vendor that serves the general public, and nothing more. Provisioning supply, especially in times of shortage, so that all market segments can be served to some extent is a responsible and intelligent business decision, not a moral obligation. I would like to think that this is the rationale behind the move, rather than this being a case of Nvidia succumbing to the cesspool of entitled moaning and performative moralizing being hurled at them.
FWIW, I can understand the frustration. I do video production & 3D rendering, and use Nvidia hardware for this purpose. I’ve directly felt the sting of the recent shortages, as I had been looking to add another card to my farm. But that’s not going to happen in the present climate, so I’m making do. This will pass, and I’ll add my card then. Life goes on.
(Note: I’m a long time lurker on this site (15+ years) with a healthy respect for your contributions. If any of this comes of as snarky or combative, I apologize. It is a sore subject for me. But I do think people should hear the other side, which is why I made an account today.)
If POS crypto algorithms indeed prove to be practical and secure dramatically lowering infrastructure costs for maintaining the ex-changeability of the currency, wouldn’t Gresham laws cause all the value to migrate from Bitcoint to Ehereum (or some other POS contender) in moderate time period?
Given the rabidity of many Bitcoin enthusiasts, probably not. For better or worse, I believe Bitcoin as we know us will be with us for awhile. Many of the people supporting that standard do not acknowledge the validity of other cryptocurrencies. They intend to replace the USD with BTC, and that is that. And they have convinced a lot of people that Bitcoin is “The Way” so if what you describe is to happen, it would take a good amount of time.
What it would do is drive adoption of POS by other, competing cryptocurrencies, reducing the footprint of the industry as a whole.
dan_sickles,
If you’re a mining farm, you buy all the hardware and electricity you possibly can to increase profits. The economics are extremely different from typical applications where hardware & energy costs need to be offset by selling goods/services to consumers. Crypto mining is done purely for financial greed. Obviously other companies are greedy too, but at least they produce output in the forms of goods and services that benefit society. Unlike crypto miners, their own economic incentives are to minimize waste. Crypto mining has encouraged wide-scale waste with very little productive output.
Yes, I am fully aware that a gentleman’s agreement isn’t possible, because greed makes it so we can’t trust anyone. Regardless, my point was that if not for the wasteful incentives, the needs of crypto currency users could be met with a tiny fraction of the resources that are currently being used. The inefficiency is astounding.
Humanity should at least get something back for all the resources being used. For all the computational resources that have been burned on solving absolutely useless hash problems, we should have something more to show for it, In principal these resources could have gone towards protein folding and curing real diseases. Again, I know it’s not the way our world works, but it’s just such a shame.
I really hope so. Proof of stake has it’s own issues, but on the sustainability and resource wasting front it can’t get much worse than our current proof of work crypto currencies.
It’s not clear to me how specialized these are going to be. Are they going to be cuda compliant? If so I already want one for non-mining applications (I’ve been on waiting lists since last year for new GPUs to do cuda work on). For them to genuinely say these are not displacing other markets, they would have to be specialized and not be capable of generic computation, but that’s kind of what an ASIC is, and given ethorium’s feelings on that – they may not accept nvidia’s move to replace GPU miners with application specific hardware.
Yes, it’s true that GPGPUs are more flexible than ASIC, but they also waste more resources versus an ASIC. For people who actually need a GPU, it’s depressing to have to get in line behind crypto mining applications.
Yes, we’ll see what the future brings. I never had the expectation it was going away any time soon. In the past falling crypto prices helped to bring relief to normal GPU consumers. Maybe that needs to happen before things will get better, who knows.
You’re fine, I have to apologize for my own grumpiness, haha. I have trouble being optimistic.
According to https://ethereum.org/en/developers/docs/consensus-mechanisms/pos/,
this is what I understood for “proof-of-stake”:
– Instead of all miners mining, a random subset will be selected and they are asked to get a consensus
– Those willing to be part of this need to put in a state 32 ETH to be considered
– And if they chosen, and are offline they will lose part of the stake
I am not sure I understood all the details. However this means not all of the network will be mining at the same time, and most of it can stay idle. However even when idle the can be called to action at any time.
Doesn’t that mean, they would still waste electricity (albeit less), and miners will still just run the machines 24/7, but GPUs will be idle?
A big fat “Amen” from my side. This is stupid on so many levels. Miners will find ways to patch it out, mining apps will obfuscate their behaviour to get around it, it could cause bugs in legitimate software, and NVIDIA’s driver development team has to divert resources away from more useful stuff (they are essentially being asked to write an antivirus for your GPU).
Alfman,
I think the crypto is just a convenient scapegoat here. For example we can’t find CPUs either, and mining has (almost) nothing to do with the CPUs.
I wrote it somewhere else. But this will just start a cat and mouse fight. Miners, especially larger industrial ones, will develop software to avoid detection, and it will make things worse. (i.e.: small miners will no longer be able to afford those software, which will most likely be closed).
Silver lining: we will have a controlled experiment to definitely say once for all: do miners cause GPU shortage?
sukru,
I haven’t seen any long term shortages of intel CPUs, Everything seems to be in stock and prices haven’t gone up compared to what I’ve paid in the past. AMD on the other hand is experiencing shortages. This is probably partly due to a shift from intel to AMD given AMD’s performance gains over intel. But also that intel has it’s own fabs whereas AMD has to get in line at the same fabs that apple, nvidia, etc outsource to. Supply chain consolidation has lead to disruptions having a much bigger impact across brands.
It depends on what nvidia CMP actually is under the hood, and nvidia’s press release is vague. Conceivably nvidia could develop a specialized chip that performs the ethorium hashes far more efficiently than conventional consumer GPUs can. In this case using consumer GPUs would become unprofitable (as has happened with bitcoin versus ASIC), that could be ideal for the consumers who want those GPUs. But ethorium’s developers could react negatively to specialized CMP hardware and take new steps to ensure that GPU mining remains profitable.
Clearly they do, the question is by how much. I don’t know whether there’s any publically available data on this? We can do some basic math to get an idea of how many GPUs are in use…
Total hash rate = 430.18 TH/s
RTX 3090 = 106 MH/s … 4.06M units
RTX 3080 = 80 MH/s … 5.38M units
RTX 3070 = 61 MH/s … 7.05M units
RTX 3060 = 45 MH/s … 9.56M units
RTX 2080TI = 61MH/s … 7.05M units
GTX 1080 = 39MH/s … 11.03M units
VEGA 64 = 34MH/s … 12.65M units
RX 6900 XT = 62MH/s … 6.94M units
Obviously I don’t know the actual hardware breakdown, but it does give us a reasonable range for the number of GPUs in use by the second most popular crypto currency.
I could not find any data on how many new GPUs are sold annually, apparently they’re not publishing this information. But I did find this article that suggests scalpers sold 39K 3000 series units on ebay. Scalpers get tons of critisism because everybody hates what they do, but if these numers are even remotely close to the true figures, then the scalpers are nowhere near as problmetic as miners.
http://www.techspot.com/news/88446-scalpers-sold-over-53000-new-nvidiaamd-cards-worth.html
Sources for the hash rate figures above…
coinwarz.com/mining/ethereum/hashrate-chart
nicehash.com/blog/post/nvidia-geforce-rtx-3080-mining-hashrate#!
nicehash.com/blog/post/nvidia-rtx-3060-mining-hashrate
nicehash.com/blog/post/nvidia-geforce-rtx-3090-mining-hashrate
nicehash.com/blog/post/nvidia-rtx-3060-ti-mining-hashrate
coinhashreporter.com/2020/12/gtx-1080-mining-ethereum-hashrate.html
legitreviews.com/amd-radeon-rx-vega-64-vega-56-ethereum-mining-performance_197049
cryptoage.com/en/2292-amd-radeon-rx-6900-xt-ethash-mining-hashrate,-ethereum-cryptocurrency.html
reddit.com/r/NiceHash/comments/jsxd0q/my_rtx3070_fe_hashrates_61_mhs_155w/
I came across this article…
NVIDIA Allegedly Sold $175 Million Worth of Ampere GeForce RTX 30 GPUs To Crypto Miners, Could Be A Contributing Factor Behind Immense Shortages
https://wccftech.com/nvidia-allegedly-sold-175-million-worth-ampere-geforce-rtx-30-gpus-to-miners/
I wonder if these GPUs for crypto miners were sold at retail prices or even bulk discounted? That’s probably 300k-400k of high end cards earmarked for crypto farms, and the hashing numbers suggest that they already had several million running before that. I think crypto miners could probably suck up all the entire GPU supply for quite long time if we let them.
Alfman,
Thanks to the wacky hidden inflation and non-increase in power costs, some older mining options became viable again. I am not sure how many of the new rigs are just old ones coming back online.
That being said, yes there is a waste in crypto. But I remember the 2018 craze, and I was able to find hardware easily at that time. This time no GPUs, no high end AMD CPUs (can you find any 5000 series in stock anywhere?), no consoles (definitely no mining on those, …yet), once again not even car makers can find microchips.
I would guess a mix of many factors:
– Increased demand due to pandemic
– Decreased production due to pandemic
– Expectation of downward trajectory for desktop CPUs, so less production was allocated
– Mining, yes of course it is one of the factors
– Competition of 5nm/7nm production pipeline with phones, consoles, etc
– Lower yields in 5nm? But I am not sure about that
I think main issue is the new production lines. nvidia started offering older GPUs, which means 12nm+ is working fine. As you said we can find CPUs, but those too are 12nm+. I can also find Raspberry PI, for example much more easily compared to last years.
So maybe we jumped to 5nm/7nm at the wrong time.
sukru,
For bitcoin, GPUs were already on their way out due to bitcoin ASICs outperforming GPUs. But there was a moderate shortage that exactly matched the peak of etherium prices in 2018…
https://digiworthy.com/2018/03/19/gpu-prices-go-down-2018/
The prices and shortage were short lived, so a lot of people didn’t notice if they weren’t buying hardware in that period. I had no issues buying stuff in october of 2018.
Maybe, but on the other hand you don’t need these high end GPUs for teleworking and most people I know who were working from home have already been asked to return months ago. Also nvidia has stated that production has actually increased despite the pandemic.
I do think the pandemic had an impact though: crypto is a lot like gold, the price goes up when people have an uncertain outlook about the future.
Not necessarily, I think the main problem is that there’s only two fabs in the world that do it and everybody has to take a number. All our companies have outsourced everything to the point where the whole planet is extremely vulnerable to supply chain problems.
I think the trends that have lead to this consolidation are systemic and it’s going to be extremely difficult and costly to create new competition. Intel is considering becoming fabless, I haven’t a clue what the numbers look like on paper, but I think it would be a disaster in terms of putting too many eggs in too few baskets.
Alfman,
For crypto I think we disagree on how much of an impact it has but agree on it having *at least some* impact.
For outsourcing: Yes, I think that is a good point. Intel for example, failed to modernize their fabs, but they are still churning out chips, and I did not see any shortages of those.
But AMD “took a number” as you said, and cannot ramp up even if they wanted to. That also led to console shortages (both Xbox and PS use AMD Zen), and ironically Nintendo Switch is the most available console at the moment.
And I am writing this on my 14nm i7 desktop.
Avoiding detection would be far more complicated than just hacking the drivers to disable the check, and then installing them as unsigned drivers; which is probably the first thing they’ll do with these cards.
It depends.
For example, if the detection is simply based on looking at memory vs core usage, then they can just add some “dummy” work, and trade the 50% drop, for say, 10%.
If the detection is based on access patterns, they can just reshuffle the loops.
Of course, there will be trial and error, and it looks like detection kicks in several minutes after the fact, so nvidia has already implemented some smarts against basic attacks.
Nevertheless it will be entertaining to watch.
I hate to break the news to everyone but NOBODY is gonna buy the CMPs for one simple reason…ROI. When the crypto bubble bursts (which it will, most likely sooner than later) the miners can get between 50-75% of their money back for the gaming cards on the used market. Go look up what the previous generation mining cards from AMD and Nvidia fetch on eBay…spoiler alert they are lucky if they get even 10% and since Nvidia recently put out a patch that makes it impossible to use the old P108 miner card for gaming? The vast majority of those mining cards will go straight from the mining rig to the dumpster.
While I personally believe crypto mining should be banned (the LAST thing we need during a shortage of silicon, a global pandemic, and AGW is someone buying thousands of GPUs to blow through power like sh*t through a goose 24/7 mining) the CMP is a solution in search of a problem as miners know they will quickly be made useless for mining as newer better GPUs and ASICs come out and unlike a consumer GPU they can’t even sell them to recoup some of their costs. Yeah these cards are DOA.
This could be interesting if it’s a consumer version of the T4 which is GPU compute only. There are uses for raw GPU compute without all of the video hardware, and if this reduces the power envelope to something more reasonable, that would be a win.
I don’t like Nvidia, and I might be interested in one of these as a basic GPU compute card. I might get a couple to run folding rigs if the price, power, and performance works out.
This move will only push forward an open source driver to unleash the power of the cards for crypto mining.
Not !
A technical fix for a social problem. Taxation is not a solution so ban the class of things we call cryptomining. It was a joke or an experiment or scam which got out of hand. It’s dangerous to the environment and a complete waste of physical resources. I daresay the “tech” industry will continue trying to be “clever” and try and ride both horses so they can monitize it every which way but it boils down to a choice. You give a shit or you don’t.
This, again.
A software fix to a business problem. The problem: Nvidia wasn’t making enough money.
Nvidia has a history of nerfing their products in order to force people to use a higher priced version, and this is the latest maneuver. They nerfed consumer cards via the driver when people started using them in clusters to force people to use their server cards. They lock FreeBSD out of CUDA via the driver, and now forcing people to buy a CMP card via their driver.
Nvidia has a long history of abuse here and keeping people from using the hardware they purchased. It’s good they’re using chips which might have been tossed otherwise (Hopefully?), but wielding the driver like a blunt weapon to extract money from people who aren’t doing things they approve of is old.
Whats the over/under on years before Nvidia starts uses their driver as the basis for a subscription service? They only make one card, and the capabilities are controlled buy the license in the driver. I say 5 years until the driver won’t install without a credit card number. XD
I agree with you about cryptocurrencies. Mainly for environmental reasons since we’re not all on renewable energy and for weakening the password algos which are used. I think they’re an interesting engineering exercise, but I’m absolutely appalled at the damage they’ve done to crypto.
@Flatland Spider
I’m not a fan of Nvidias business practices. Their drivers allow sloppy API code and cheat to achieve plausible visual performance gains plus as you say have played cute to differentiate products via drivers to extract more money. Yes they have pushed good technology but there are lots of downsides.
Sony acquired a patent for tiered graphics processor provision allowing processing to cascade between local and cloud based hosting invisibly. It’s a good idea but not one I belive deserves a patent. I hadn’t thought of this but given Nvidias business practices and what technology enables (Nvidia have cloud based GPU processing) turning the driver into a subscription based service does seem like a logical follow through.
Some people have described cryptocurrencies as being misnamed and they are really transaction protocols which use the load of processor intensive encryption as the price of a transaction. It doesn’t have to be encryption but could be something actually useful. Even if it were useful the price-performance would still be hideously bad. I will leave it to the academics to crunch the numbers but the simple headline figure of energy use being equivalent to a small country for minimal gain when compared to established transaction systems does not look good. In fact in Iran bitcoin mining has caused blackouts. Regulating or banning systems organised for the sole purpose of generating profit versus the everyday energy needs of people in Iran and now Texas are not I believe unreasonable. This makes cryptocurrencies not just a human rights and environmental issue but also a national security issue and international security issue.
I think the basic engineering exercise has merit. As an example if a big enough problem came along it is not outside the realms of possibility that the entire output of the worlds supercomputers and distributed systems were turned to that problem the cost-be-damned. Covid-19 has been proof enough of this as on the quiet many organisations and individuals went “all in” on a none profit basis as part of the international and national efforts. Collectively we achieved in under a year what may previously have taken a decade or multiple decades or never to achieve. This is notable.
I think it all comes down to are you a builder or a destroyer. I know the world is more nuanced and complex than this and lots of people have no choice in the matter but these are the essential questions.