A Word Of Warning On Used GPUs

[IInuyasha74]

With new GPUs selling for outrageous prices, people are looking everywhere they can to find a graphics card that doesn’t break the bank. For some, a used GPU may be the answer, but you need to take caution when considering this option.

A Word Of Warning On Used GPUs : Read more

 

[Sakkura]

"The constant high level of power running through the card, coupled with the taxing workload and heat, can damage the onboard circuitry and lead to the card failing far sooner than expected."

That's nonsense. A constant level of power is, if anything, beneficial as it avoids thermal cycling that induces physical stresses on delicate solder joints. And electronics, if properly cooled, can run for many years at full load.

Inadequate cooling could perhaps be a concern on cards used for mining, as could the wear on the fan bearings.

 

[2Be_or_Not2Be]

"The constant high level of power running through the card, coupled with the taxing workload and heat, can damage the onboard circuitry and lead to the card failing far sooner than expected."

That's nonsense. A constant level of power is, if anything, beneficial as it avoids thermal cycling that induces physical stresses on delicate solder joints. And electronics, if properly cooled, can run for many years at full load.

Inadequate cooling could perhaps be a concern on cards used for mining, as could the wear on the fan bearings.

But didn't he say a "constant high level of power" - "high" being the key word? So could mean that the card has been overclocked & running non-stop at a high overclock. That, in turn, means at the minimum, the cooling function has been affected as it has to run non-stop in order to keep temps down.

It also could mean components that weren't designed with higher tolerances could then be stressed more heavily than initially designed/expected. So component failure could also result earlier than desired long-term.

 

[2Be_or_Not2Be]

With new GPUs selling for outrageous prices, people are looking everywhere they can to find a graphics card that doesn’t break the bank. For some, a used GPU may be the answer, but you need to take caution when considering this option.

If it seems too good to be true, then it probably isn't.

 

[redgarl]

It is much more related to the quality of the card than the fact it run at full load 24/7. Basically, it is coming back to the HS design. Most of the time, GPU are undervolt to reduce the powerbill because GPU are usually not well tuned for mining.

 

[AnimeMania]

If you buy a card that was used for mining, what things do you have to change to get it to run as a gaming graphic card, BIOS, CPU clock, RAM clock, voltage, fan profile? Is there a simple way to reset it back to original specs? Is it possible that malware might have been introduced?

 

[ccook009]

"The constant high level of power running through the card, coupled with the taxing workload and heat, can damage the onboard circuitry and lead to the card failing far sooner than expected."

That's nonsense. A constant level of power is, if anything, beneficial as it avoids thermal cycling that induces physical stresses on delicate solder joints. And electronics, if properly cooled, can run for many years at full load.

Inadequate cooling could perhaps be a concern on cards used for mining, as could the wear on the fan bearings.

You're right about the thermal stress mostly however; there are other aging effects in semiconductors that are negatively effected by constant power. In particular, Electromigration is a huge problem in modern advanced semiconductors. Essentially, uni-directional electron flow causes stress to build-up due to momentum transfer from the electrons to the metal lattice in the chip interconnects. While companies try to design the wires in their chips to be "Electromigration-immortal" they usually have to allow some wires to be susceptible because they don't really expect their customers to have the chips running 24/7 at full load for years at a time. It should be noted that there is a relaxation effect that mitigates electromigration if you let the GPU rest for a while ( But this, as you eluded to, could have consequences as it could result in some mechanical stress). There are other aging effects too, e.g., Time Dependent Dielectric Breakdown, Hot Carrier Injection, and Bias Temperature Instability which can degrade the performance, cause faults in the data, or even permanent failure of the chip.

 

[Destari]

This is a ridiculous assumption. Miners don't run cards at 100% load and they certainly don't overclock them. Mining is entirely about efficiency, in fact most miners will UNDERCLOCK their cards to reduce thermals and get lower power draw. When mining the primary adversary to profit is power usage, so saying that a miner overvolts or runs a card screaming fast 24/7 is insane.

 

[jaber2]

As someone who runs a couple 6 GPU rigs I have to say none of this is true, first of all I don't run above 65%, and when you do that you always are running cool between 40 to 60 degrees, the video cards are designed to run hotter and on and off more often, also wanted you all to know I don't plan on selling my gpu's, they will continue running until they die

 

[raspberrypiguy]

Smart miners will decrease the power limit to 85% and Max out the fan to keep the cards running cool. Although the hashrates are lower the Hash/Watt ratio is more profitable. As long as you don't mess with the voltages or jack up the power limit a modest core and memory overclock is acceptable for mining and the risk of burning out the card is relatively low. Don't put any thing over +400. And don't buy Chinese GPUs that are to good to be true. Also another person commented on thermal cycling. Hot and cold swings are bad for solder points over time. A constent temperature is safer than a swinging one.

 

[2Be_or_Not2Be]

This is a ridiculous assumption. Miners don't run cards at 100% load and they certainly don't overclock them. Mining is entirely about efficiency, in fact most miners will UNDERCLOCK their cards to reduce thermals and get lower power draw. When mining the primary adversary to profit is power usage, so saying that a miner overvolts or runs a card screaming fast 24/7 is insane.

I like the points you make about miners & their usage of GPUs. However, caution about used GPUs isn't strictly related to GPUs used by miners; it can be about gamers overclocking like crazy & then trying to trade with someone else.

 

[raspberrypiguy]

Nicer miners will fit the factory BIOS back on the card so that the average person can buy it. My mining card has the stock BIOS on it as I am casual about my hobby mining. If a card happened to still have a mining BIOS on it you could reflash it using tutorials online. As for overclocks they are usually controlled by software and not hardware.

 

[delaro]

There are always risks to buying used anything but with computer hardware it's about sticking to certain brands and communities. I avoid Ebay for GPU's since lately most sellers seem to be smoking crack and you find to many scammers. There are many forums that have fairly tight restrictions on the how to sell using "http://www.heatware.com/"and "http://www.beerology.com/" as seller reference and using PayPal as the method of cash transfer. For the most part sellers in these communities are pretty honest about what their hardware has been used for, to many complaints means getting "IP" blacklisted and losing that method of selling.

Now for hardware paying $250+ for a 1050 TI vs $180-$225 for a 970 GTX used seems like a better deal when your looking at $/performance. It's not hard to find used cards with warranties left and most of them will transfer, this is why you stick to certain brands. I avoid PNY, Zotac and Sapphire since the warranty is attached to the user but EVGA,XFX,ASUS,Gigabyte and MSI apply their warranties to the serial number of the card and not the purchaser.

 

[Paul Alex Cazacliu]

That's so true, I just got to replace in the previous December my GTX 970, which was the best of it's kind, the MSI 100 Million Edition that was running alone in a big cooled case and at just 7-8 hours a day under heavy load, not as the miners do, stacking cards like canned sardines running 24/24 at Max capacity; my has lasted just two years, a miners one obviously is lasting way less than that. Previously I had a MSI Radeon 7870 HAWK Edition that also lasted just two years at just 7-8 hours at heavy load. Yes, I got cards that I just retired without being damaged, but those were low end ones when I couldn't afford more than that, purchased over ten years ago when the cards were low power consumption and needing just one cooling fan for the GPU. So yes, an used card is a waste of money, it has very little life left if any and no warranty.

 

[TJ Hooker]

@Paul Alex Cazacliu your couple of anecdotes don't prove anything about the typical lifespan of cards, whether they're used for gaming or mining or whatever.

 

[TJ Hooker]

"The constant high level of power running through the card, coupled with the taxing workload and heat, can damage the onboard circuitry and lead to the card failing far sooner than expected."

That's nonsense. A constant level of power is, if anything, beneficial as it avoids thermal cycling that induces physical stresses on delicate solder joints. And electronics, if properly cooled, can run for many years at full load.

Inadequate cooling could perhaps be a concern on cards used for mining, as could the wear on the fan bearings.

I partially agree.

But in addition to what has been said above about electromigration etc., there are also the components on the board other than the GPU itself. For instance, take the VRM capacitors. The more card is used, the faster the capacitors will deteriorate, potentially impacting their performance or eventually causing them to fail.

 

[quallen]

yep very true about electron migration. this also happens in the transistor junctions over time. given enough time and stress the migration will punch a hole in transistor junctions. Ironically, this is one way how EMP from a nuke destroys your electronics.

 

[RCaron]

It doesn't seem like the writer has a lot of experience in electronics as he made a few mistakes.

1) semiconductors operate better over consistent workloads as opposed to being reved up to max and turned off, and having that same process repeated over and over putting the board through thermal expansion.

- This is actually what you're looking out for. Unfortunately you'll most likely see this from used gaming cards as opposed to from mining cards, because miners tend to run the cards at a high memory clock but at a low voltage in order to mine more efficiently.

I'd go a step further regarding the fan, Don't just check it, replace it. Which means you should be looking for a second hand card where the fan was built to be swappable. If the fan fails your card gets cooked. Fans also undergo wear overtime while in constant use, unlike the silicon.

 

[quallen]

ccook009 made a good point. but this also occurs in the transistor junction too. eventually it will punch a "hole" through the junction. Ironically, this in similar to one of the ways EMP from a nuke destroys electronics. Also, perhaps not miners run their cards to the wall, but there are those who do. and you cant tell who's who.... just a thought.

 

[RCaron]

Writer doesn't seem to have any real experience with electronics.

1) mining cards are better. They're run at lower voltages and are run 24/7 (which is a good thing) as the card does not undergo thermal expansion and contraction from being turned on and off. A constantly used card adds much less stress on the electronics.

2) His fan recommendations is also going to land you in trouble. Don't look at the fan, replace it. Fans are the single item on any card that will lead to the cards failure. Fans also age while being used, they get dirty and slow down as well.
Which means that you should be replacing these tiny inexpensive items on every video card you buy used. So only buy a video card where the fans are built to be swappable.

It's simply not worth it to buy a video card only to have the fan fail on you. So replace it, it's cheap.

 

[Sakkura]

I partially agree.

But in addition to what has been said above about electromigration etc., there are also the components on the board other than the GPU itself. For instance, take the VRM capacitors. The more card is used, the faster the capacitors will deteriorate, potentially impacting their performance or eventually causing them to fail.

If we were still using electrolytic caps, then that might be a realistic worry. But we aren't, those things are basically confined to PSUs only by now.

And as for electromigration, I doubt that's a real problem. Bear in mind a lot of chips are co-designed for use in both consumer desktops and enterprise workstations or servers. The latter can have workloads just as constant and demanding as crypto mining, and those customers want hardware that's reliable and durable. I'd still be much more worried about worn out fans (also because the cooling solution will often be specific to the consumer model).

 

[Giroro]

Is there any actual evidence that a GPUs used for mining are more likely to fail than ones used for gaming?
Even if the miner was irresponsible with heat management in a tightly packed case, how well does modern thermal throttling protect the circuitry? I would like to think that the cards would start throttling before they even come close to discoloring the circuit board.

 

[TMTOWTSAC]

I think it's a mistake to assume that every miner is a professional who makes the best decisions. Everywhere you look there are people late to the party suddenly interested in mining. This forum is full of them. And yes, half are looking into efficiency and undervolting and profitability and ROI. The other half have completely unrealistic expectations and make poor decisions.

And sometimes it is in the best interests of a large scale professional mining operation to push their hardware. When prices are high, it makes sense to maximize production in order to take advantage of it. When prices are high enough, efficiency becomes secondary to time until the inevitable fall/crash.

 

[Zaporro]

"The constant high level of power running through the card, coupled with the taxing workload and heat, can damage the onboard circuitry and lead to the card failing far sooner than expected."

That's nonsense. A constant level of power is, if anything, beneficial as it avoids thermal cycling that induces physical stresses on delicate solder joints. And electronics, if properly cooled, can run for many years at full load.

Inadequate cooling could perhaps be a concern on cards used for mining, as could the wear on the fan bearings.

Classic miner propaganda right there

Buying a Used Graphics Card? What to Watch For...............

What to watch out for? Watch out to not buy a card that was used in mining. Ran 24/7, with modded unofficial bios and who knows how else modified.

Not to mention that its the miners who ruined gaming/PC building by driving prices to insane levels, buying GPU that was used to mining is a straight support for miners. They buy gaming GPUs specifically because they know some naive people will buy them any time, they use that as their escape plan if something goes wrong and they want to get their money back.

By buying GPU from a miner you support them and their cancerous behaviour, you become part of the problem.

 

[hoofhearted]

I sure hope the video card makers turn this from subjective to definitive. Some sort of hours / history / odometer baked into the silicon. But not just time, also load factor. Some sort of integral or summing calculation. Then reasonable thresholds to deter "heavy" miners vs gamers. It could only help their warranty dept as well as the used card market adding a level of predictability vs uncertainty to their market.

I wonder if they already do this, but just keep it secret. Any wireshark or jtag (or whatever the acronym is for the behind-the-curtain stuff of video cards) experts out there?