Crypto miners reveal Nvidia's LHR solution to get GeForce RTX 30 series cards into gamers' hands was futile

[Redaktion]

An investigation into Nvidia's Lite Hash Rate (LHR) solution to prevent GeForce RTX 30 series boards from being bought up en masse by crypto miners has revealed that many involved in the crypto craze were not discouraged by Team Green's limiter. Third-party software is available to bypass the LHR mining lock and raise the rate.

https://www.notebookcheck.net/Crypto-miners-reveal-Nvidia-s-LHR-solution-to-get-GeForce-RTX-30-series-cards-into-gamers-hands-was-futile.596503.0.html

[Jonah Hensley]

I haven't noticed much of a performance decrease in mining on an LHR 3060 I picked up last month. I mostly use my GPUs to train AI algorithms, but mining is a good way to make use of the down time in between training.

Also I'm not sure what supply shortage everyone is talking about, I see plenty of stock on Amazon everytime I get on. It took me less than an hour to find each of my cards, and they were in hand a week later at the latest. It just seems people are unwilling to pay market price for such an in demand item.

These cards are designed to solve complex mathematical problems, and the need for that has skyrocketed in the past year or two, graphics are merely a by product of that ability. Eventually there will be enough for everyone at a decent price, but that won't be for years. So if you want one of the most useful in demand computing products for every industry, for something as wasteful as gaming, you're gonna have to pay the price like we did.

[_n]

... These cards are designed to solve complex mathematical problems ... So if you want one of the most useful in demand computing products for every industry, for something as wasteful as gaming, you're gonna have to pay the price like we did.

I'm not sure where you got the idea that these graphics cards are being sold as performant computation devices? In none of the web pages for any 30 series card was there a mention of its machine learning or mining capabilities (aside from the LHR disclaimer).
These graphics cards are designed, manufactured, and marketed as high performance gaming devices; the fact that you can mine or train ML models or do 3D renders is a byproduct of its design rather than the intended use case.

If you were talking about Quadro or Titan series cards then I would understand your viewpoint, but the contempt you have seems rather unfounded.

[RobertJasiek]

RTX 3000 (especially its tensor cores) are very suitable for machine learning or its applications whenever the VRAM is sufficient. If each sample is small, 4GB can be enough for their (example: Go game) application (when the previously trained net is already available) and all that matters is numbers of cores and bandwidth. Advanced ML (example: image recognition) might need more VRAM when training, e.g., deep neural networks, then RTX 3090 or A-series cards with more than 24GB can be needed. Only specific ML softwares (Igor's Lab has mentioned Siemens) profits from drivers etc. of A-series cards or their predecessors. Many ML softwares just need good numbers of tensor cores and CUDA cores, a few softwares even run faster on RTX 3000 than on mid- to high-end A-series cards. Some ML softwares profits from SLI while other softwares just profit proportionally from more GPUs due to the increased total number of cores.

At German Amazon, I have never seen Amazon itself selling RTX 3000 but only seen third party scalpers. RTX 3080 have cost 3x - 5x MSRPs even when some other retailers offer 2x MSRPs, which is still way too high unless one joins the mining Ponzi scheme or absolutely needs some card roughly in the price range of A-series cards. As somebody merely wishing to apply already trained ML, these prices (even the lowest 1,5x MSRP 10 months after launch for 2 weeks) are way too high and I skip RTX 3000. I do not join the Ponzi scheme and the same situation with RTX 4000 means I might just get an APU then.

[Jack Dawer]

Graphics a byproduct? Wasteful gaming my a**. That's what they were designed for, and it's in the name.
Just because you can use them for multiple purposes doesn't change what they were originally designed for, idiot.

[Jafar]

The only thing bad about LHR cards is that the energy use is less efficient than FHR cards. NVIDIA only succeeded in making their cards less environmentally friendly. I have 5x 3080ti cards and they are mining beasts. Even on Ethereum.

[Username]

Gaming is a pretty wasteful use of these cards when they could be generating cash.  As long as they can generate $100+ a month off $5 of electricity people will keep paying the market price for the cards. 


I wouldn't be surprised if nvidia raises the msrp again on the 4000 series cards like they did with the 3080ti.

[danwat1234]

GPUs have so many scientific and engineering uses besides gaming. CUDA computation is no joke.
For example when not gaming, is to crunch for Folding@home Distributed Computing project. I would never have them do proof-oby-work, busy work even though Folding is profit-less.

I was wondering about the energy of efficiency of mining eth on LHR models. Thank you I had a feeling it just made things worse rather than consume less power.
Reminds me of the Norton ethereum mining software.. Linus Tech Tips found that it consumed more power & mined slower than regular solutions.

[Joe]

All we can hope is that the crypto market continues to spiral downwards and the goverment steps in and regulates the market (i.e. over taxes it and makes preventative regulations that in turn destroys it) so that this hording of computer components stops and people can get back to having sctual jobs to provide an actual tangible asset to the country..

[p0nch0.eth]

All we can hope is that the crypto market continues to spiral downwards and the goverment steps in and regulates the market (i.e. over taxes it and makes preventative regulations that in turn destroys it) so that this hording of computer components stops and people can get back to having sctual jobs to provide an actual tangible asset to the country..

Did you loose some money in crypto or why you mad?

[Brian]

So why does Nvidia sell gpu to begin with? I think they need to contact me so I can be ceo. Maybe the manufacture should just do the crypto mining itself since it is apparently the most profitable use. Makes me wonder how many of those super computers the governments have are mining crypto.

[AlexS25]

Nvidia wants their boards to be bought in mass by crypto miners and is deliberately allowing bulk purchases.  When it looked like the shortage would end they shut down their chip factory for a few months.   They are asshats.

[Spunjji]

"...crypto miners found themselves forced into..."

Nobody forced any of these assholes to do anything. Miners suck.

[Dirty dan]

Maybe you should tell us things we didnt already know for the last year

[Anonym]

All we can hope is that the crypto market continues to spiral downwards and the goverment steps in and regulates the market (i.e. over taxes it and makes preventative regulations that in turn destroys it) so that this hording of computer components stops and people can get back to having sctual jobs to provide an actual tangible asset to the country..

That's a very long-winded way to say "I don't have the faintest idea what I'm talking about". If that was your objective then congratulations, mission accomplished!