News LinusTechTips: Buying Second-Hand GPUs Is a Good Idea, Even From Miners

[3ogdy]

"several triple AAA games". That's AAA AAA AAA games. Man, that's like diamond level quality!

Joking aside, I will NEVER buy a GPU from a miner. They were the reason why I coudn't buy a GPU for such a long time, because they were buying them massively to resell them at unbelievably high prices on ebay.

Miners: "Ok, you dumb buyer, I'm using scripts and bots to buy new GPUs as soon as they become available. Then, I'm gonna ask you to work more hours and earn the kind of money I want to make. That way I can make other people's lives harder."

Duck'em!

Also, every electronic component has its own lifespan. Cards were not designed to be used constantly and intensively. That alone, more likely than not, causes premature wear on a GPU that's sold at prices higher (not lower, despite being used) than MSRP.

So, it becomes a circle. I don't buy overpriced cards, they keep the stocks. They'll have to lower the prices to get rid of them (that applies to AIB partners as well), and THEN, I will buy at MSRP or less.

After all, it was us (the consumers) that allowed all this to happen.

 

[russell_john]

Linus from LTT has provided testing that shows buying a used mining GPU is safe to do, amidst the growing trend against buying mined-on GPUs because of reliability concerns.

LinusTechTips: Buying Second-Hand GPUs Is a Good Idea, Even From Miners : Read more

19 is not a proper sample size to draw any real conclusions from ...... There is a gentleman over in Germany who used a sample size of 60 used GPUs and 43 of them completely failed to the point many of the dies cracked because they were mining cards. Should I draw a conclusion from that 3 times larger sample size that mining GPUs have a 70% chance of critical failure?

 

[Pollopesca]

I picked up two used EVGA 3070 XC3s a while ago for about $300 each. I can’t confirm if they were used for mining or not, but they were in decent shape, just needed some cleaning and re-paste for good measure. I’ve got some buddies who have confirmed 3070s from mining (riser adapter was included and still had aftermarket firmware) but after re-flash and clean up they’ve been running solid for months. With GPU prices these days, its worth the dice roll if its cheap enough IMO. Would have gone with an Intel A770 if the DX9 support wasn’t trash at the time. ¯\(ツ)/¯

 

[logainofhades]

19 is not a proper sample size to draw any real conclusions from ...... There is a gentleman over in Germany who used a sample size of 60 used GPUs and 43 of them completely failed to the point many of the dies cracked because they were mining cards. Should I draw a conclusion from that 3 times larger sample size that mining GPUs have a 70% chance of critical failure?

It wasn't because they were mining cards, it was because of improper storage, and/or handling of said cards.

 

[Deleted member 431422]

This wouldn't have something to do with them preferring you to buy new, would it?

That's an odd statement of theirs. Either is how you said, or they're using cheaper components that are just right, without headroom for aging.

 

[Maebius]

Almost everything in life is a gamble. I just don't want to give a person that hoarded 10-100 graphics cards my money. Even if it's good.
He deprived me of being able to buy it at the initial target price. Most importantly, Nvidia's appetite went way way way up and now I don't even think of getting a new card any time soon.

 

[Deleted member 431422]

19 is not a proper sample size to draw any real conclusions from ...... There is a gentleman over in Germany who used a sample size of 60 used GPUs and 43 of them completely failed to the point many of the dies cracked because they were mining cards. Should I draw a conclusion from that 3 times larger sample size that mining GPUs have a 70% chance of critical failure?

A significant sample would be 10 000 mining cards and from around the world not from a single market.
Also, you should watch the 2nd part of that video. KrisFix explocitly stated his belief is those cards were stored improperly in a place with high humidity, hence the cracked dies.
I haven't seen, and likely will never see, any significant study of mining cards lifespan. Mostly due to the cost. It'd be thousands $ for the cards and possibly even more for the study. No one is going to waste money on something for fun. Not relevant to either AMD or Nvidia's business.

 

[LolaGT]

If you are listening to GPU advice from a guy that bought a couple of a few various GPUs then you get what you ask for and deserve. Hey, you might get lucky. ha

Buy a mining card if it is deeply discounted, and you can get a good look at and test it with the option to return it.
I would not risk much money.
But: We already know much of the consumer market are suckers, they paid the ridiculous prices through the crypto boom, and they are going to pay more than what they are worth for these too.
Scalpers and con men exist because there is a pool of people willing to give them money.

 

[Deleted member 14196]

Whenever a salesman, like Linus, says, trust me, I turn and run the other way

when I buy hardware, it will have the ability to be returned If it doesn’t work—and replaced for free. Or else I’m not buying it. Buying any used electronics is a huge gamble and I guess if you’d like to gamble then go for it, but I don’t like to throw my money away by gambling.

 

[Wendigo]

On one hand, it's fair to say that the odds are on the buyer side to get a perfectly good card. On the other hand, 2 cards with issues out of 19 also means the odds of having a bad card are also much higher than with a new one.

Since LTT (or Tom's for that matter) have tested and used a lot of cards in new or almost new state, it would have be interesting to know what failure rate they observe with these non mining cards. Although the sampling of mining cards was low, it could still be significant if compared to a high enough number of non mining cards. Said otherwise, a large enough group of non mining group cards allows to calculate a good estimate of the failure rate. Based on this, it's then possible to verify if the 2/19 of the mining group falls within this expected failure rate or is significantly outside it.

EDIT; Just done a quick check and Tom's or LTT could randomly make a group of 35 cards (or more) they received for testing and check if they were working fine or not. If they were all fine, then having 2 cards having issues in a group of 19 is significantly higher than what would be expected. And if the control group is 60, you'll have to be very unlucky to get one bad card in a group of 19.

 

[palladin9479]

Yeah LTT ... missed the boat and is likely being paid to do this.

The issue with mined cards isn't so much them being DOA, but them having power / thermal issues in the long run due to them having been hard for so long. So only buy one if you can get them really cheap, then don't OC and if possible underclock / undervolt it or disable any sort of auto-boosting to reduce chance of memory / compute error.

 

[PEnns]

19 is not a proper sample size to draw any real conclusions from ...... There is a gentleman over in Germany who used a sample size of 60 used GPUs and 43 of them completely failed to the point many of the dies cracked because they were mining cards. Should I draw a conclusion from that 3 times larger sample size that mining GPUs have a 70% chance of critical failure?

Exactly! And it was even posted here!!

But hey, The Linus groupies don't want to dwell on such more interesting findings from someone who doesn't get paid by somebody to peddle mining GPUs!

 

[Eximo]

Always a set of things to consider.

Casual gaming, turning it on, a few hours of play a week.
Gaming 8 hours a day
Mining at home 16 hours a day + gaming 8 hours a day
Smoking/Vaping while gaming/mining
Non-ventilated computer, running the GPU throttled 90% of the time
Pet ownership
Plugged into a poor power supply
Running overclocked

VS

Sitting in a data center or data closet with filtered air, steady power, and a constant load.
Sitting on a haphazard shelf in a dusty warehouse
Plugged into an unreliable power source/power supply
Modified BIOS
Running overclocked
Running underclocked.

 

[Deleted member 14196]

Yes, and I don’t like the sound of either of those options

 

[Deleted member 431422]

Always a set of things to consider.

Casual gaming, turning it on, a few hours of play a week.
Gaming 8 hours a day
Mining at home 16 hours a day + gaming 8 hours a day
Smoking/Vaping while gaming/mining
Non-ventilated computer, running the GPU throttled 90% of the time
Pet ownership
Plugged into a poor power supply
Running overclocked

VS

Sitting in a data center or data closet with filtered air, steady power, and a constant load.
Sitting on a haphazard shelf in a dusty warehouse
Plugged into an unreliable power source/power supply
Modified BIOS
Running overclocked
Running underclocked.

I hope people without PC hardware knowledge read this and consider what the best option is for them.

Thinking about the current state if the market there is either you pay €250+ for a new card even though you top at 1080p and play older games or you go to 2nd hand market and buy one or more generations older card. I am yet to see a proper replacement for RX5500 or GTX1660 Super. RX6400 or RX6500 are not worth the money, GTX1660 Super i actuallyl too expensive. It's better to buy the cheapest RTX3050.

There's nothing sub €200 to choose from and I don't believe RTX4050 will be any cheaper than RTX3050.

 

[Eximo]

Intel A380 and GTX1650 Super/GDDR6 version is about it sub 200 that isn't a 4x card.

But the argument could be made for picking up an RX6600 or A750 despite the extra cost, so much more price/performance.

I agree, the RTX4050 is going to be a nightmare. At a certain point I would just start looking at gaming laptops that are on sale.

 

[d0x360]

Most cards used for mining would be perfectly fine to overclock and game on. It doesn't matter how long they were run.

The issue now is the price for cards tanked so these people are resorting to nast tactics on order to sell them for as much as possible.

Linus would have had someone test the cards for a decent amount of time before making the video.

He's also not for kids he's just for a for mass market audience than something like gamers nexus. Most people don't understand wth Steve is talking about or why. Linus makes it easily digestible but he could definitely go in deep like GN if he really wanted to but why would he? He has his audience and GN has theirs.

 

[SyCoREAPER]

Most cards used for mining would be perfectly fine to overclock and game on. It doesn't matter how long they were run.

The issue now is the price for cards tanked so these people are resorting to nast tactics on order to sell them for as much as possible.

Linus would have had someone test the cards for a decent amount of time before making the video.

He's also not for kids he's just for a for mass market audience than something like gamers nexus. Most people don't understand wth Steve is talking about or why. Linus makes it easily digestible but he could definitely go in deep like GN if he really wanted to but why would he? He has his audience and GN has theirs.

- Most cards used for mining would be perfectly fine to overclock and game on. It doesn't matter how long they were run.

False. Miners don't care and often have the cards open to air elements, not cleaning or maintaining them and running them 24/7 without monitoring heat or for other issues. Dust, Humidity and other less ideal conditions included.

- He's also not for kids he's just for a for mass market audience than something like gamers nexus. Most people don't understand wth Steve is talking about or why. Linus makes it easily digestible but he could definitely go in deep like GN if he really wanted to but why would he? He has his audience and GN has theirs.

Correct, he's not for kids, he's for the uneducated, gullible and flock mentality types who take his word as gospel. Nothing he does is scientific or even thoruough in nature, it's to push out a video and push it out fast. He's quantity over quality.


Don't defend his BS tactics because it makes you no better than him. And comparing Steve to Linus at all is insulting, I don't even mean that in a funny or metaphorical manner, it's comparing a trade specialist to a professor. There is no doubt Linus has experience but he's clouded by dollar signs and metrics.

Steve and DerBauer are super scientific and thurough in their tests and don't just jump to conclusions. If there are two Youtubers to listen to, they are it.

 

[cfbcfb]

I watched this video when it aired. The brief piece people should focus on is how they procured these gpus. Because it made the rest of the video useless.

They bought cherry picked, ideal cards from a variety of sources. Clearly not the pressure washed and warehoused cards that so many are having trouble with.

Most of these youtubers who make big coin, like LTT, ETA Prime and others gloss over things by saying "It's not that bad!" or "Better than I expected (but I didn't expect much)" or they fly right over the big neg that would have led to no monetized video.

 

[edzieba]

NIB 'used' card - could be a victim of the bathtub curve just as much as any other card.
Used gaming card - could have been babied, or could have been overclocked on a dodgy PSU and clogged with pet hair
Used mining card - almost universally underclocked (because power efficiency > outright performance maximisation), most will have been in datacentres.

Buying used is always a a gamble, mining use does not appreciably tip the scale. Buy based on price and seller return policy.

They bought cherry picked, ideal cards from a variety of sources. Clearly not the pressure washed and warehoused cards that so many are having trouble with.

I've yet to see any of these supposed 'powerwashed' cards outside of tiktok ragebait clips.

 

[logainofhades]

I haven't seen such a thing either, but it is possible that those cracked die GPU's were due to such treatment, as it is believed moisture is to blame. High humidity is also a likely reason for moisture buildup.

 

[horsemama]

I'm actually going to be grabbing a 5600XT or 5700XT next week. I currently have an RX 6400 that is meant for a secondary/emulation PC but haven't found anything retail that isn't a bottom of the barrel model(they run hot and loud) for a decent price in Canada. I don't care about RT, high resolutions/refresh rates either.. Upper model 5600XTs are going for 180 or so.

I can't support these new GPU prices so it looks like I may just go used and be a generation or 2 behind going forward, which is fine.

 

[helper800]

Whenever a salesman, like Linus, says, trust me, I turn and run the other way

when I buy hardware, it will have the ability to be returned If it doesn’t work—and replaced for free. Or else I’m not buying it. Buying any used electronics is a huge gamble and I guess if you’d like to gamble then go for it, but I don’t like to throw my money away by gambling.

By your definition everything is a gamble because anything can be purchased new and damaged or non-functional, and companies regularly do not honor warranties, or will make you pay to return their product.

 

[helper800]

Yeah LTT ... missed the boat and is likely being paid to do this.

The issue with mined cards isn't so much them being DOA, but them having power / thermal issues in the long run due to them having been hard for so long. So only buy one if you can get them really cheap, then don't OC and if possible underclock / undervolt it or disable any sort of auto-boosting to reduce chance of memory / compute error.

Oh yes, a hunch claiming someone has no integrity with no basis when Linus's entire brand is propped up by their integrity. When they do screw up, they own it. I generally agree with your sentiment, but like you said about Linus, you missed the boat on this one...

 

[Deleted member 14196]

Nonsense. Buying from Established good vendors will always honor warranty.