Should you buy ex-mining GPUs?

[@Mickulty]

00:00 Intro 01:00 Radeon VII 05:26 RTX 3090 09:49 Other HBM Cards 11:10 Other GDDR6X Cards 12:16 Other GPUs (GDDR6/GDDR5) 12:46 Fans 15:55 Dust 17:48 Cleaning 26:00 BIOS 31:15 Conclusion

[@savage8536]

I love you

[@vilimtustanovski1270]

A samaritan in the wild

[@seamon9732]

Lifesaver... AHOC really needs to learn to SUMMARIZE.

[@xinfinite8698]

dont buy gddr5 gpus

[@Tigerhearty]

vega 64 died on me after 1.5 year of mining. Unrepairable, These hbm modules are pretty bad.

[@louisfriend9323]

TL:DR; 1. avoid HBM memory GPUs like Radeon 7, Vega, etc. 2. avoid GDDR6X memory GPUs like 3080, 3090, etc. 3. consider GPU fan can break due to running at 100%, can replace. 4. consider deep cleaning the GPU, and replacing thermal paste. warning: thermal pads, try to keep them. 5. cards got weird mining BIOSes, try to find BIOS for your model. warning: avoid high clock BIOS or wrong memory brand BIOS.

[@1vend7]

The problem is we never know who owns the cards, I have a 3080 that the memory doesn't go over 70ºc in the summer because I use copper instead of thermal pads

[@xinfinite8698]

avoid gddr5

[@dat_21]

@@xinfinite8698 TL:DR. Don't buy ex-mining GPU.

[@Atheist7]

@@dat_21 WRONG! 6. TL:DR. Don't buy ex-mining GPU. Ahhh. That's better.

[@mehdidamfs3112]

in the crypto crash of 2018 , i bought a mining Vega64 for 200$ , i cleaned it , repaste it . it ran fine until 2021 where i had a short circuit in the Vbios switch which made the car tetected as MI25 ( buildzoid helped me diagnose the card ). now the card is still runing well :D

[@Drdrew1234]

Hey BZ. Radeon VIIs perform like a 3080 when mining eth (100 mh/s) which is why they are such a good mining card. The algorithm for mining eth absolutely slams the memory controller and makes the memory chips hot. VIIs tend to practically kill themselves while mining and give the generic error of code 43 when on a windows pc, overall they’re super delicate cards and I wouldn’t recommend buying one either

[@intesstine]

думаешь, он про это не знает?

[@InsertValidName]

Happy its not 42

[@briankleinschmidt3664]

That was my thought. On SSD's if you max out the memory, It will degrade fairly quickly. The principal should apply to gpu memory.

[@rigglestad8479]

@@briankleinschmidt3664 No, Flash Memory is not the same as DRAM...

[@DeeGeeFi]

I run folding@home on my Radeon VII. I hope it isn't as harmful to the GPU as Ethereum... Although I already had to "upgrade"the fans to Noctuas, failing bearings...

[@zoomzabba452]

Bought an ex-mining RX580 8gb for my nephew in Dec 2019. Being 1000 miles away, i made sure it was the Sapphire with easily replaceable fans. Sure enough, one died after a month. Seller sent him a replacement for no extra charge. Still running fine today.

[@dgurevich1]

Props to that seller. I bought a r9 280x long time ago. It ran well for 4 years and is now dead In a box somewhere. The fans on it failed after 2 years but replacements were very cheap.

[@purpasmart_4831]

Was it a sapphire dual-x? If so that could of been mine lol.

[@Wasmachineman]

I bought a bunch of cards which included a HD 4890 TOP and 7950 Dual-X, both with fucked fans. Said 7950 has two Arctic P14's tiewrapped to it now lol

[@SovietN1nja]

Also AVOID 3060 Ti, it has ongoing problems with Hynix GDDR6 memory failing, unless you're completely sure that it's not on Hynix memory. Media is somewhat silent about that, but some are already reporting about this issue, same as it was with early micron batches on rtx 2xxx cards couple years back.

[@Sva010]

Samsung or Micron are mostly fine

[@ThunderingRoar]

is it only 3060ti? Because 3070 uses pretty much the exact same memory layout

[@SovietN1nja]

@@ThunderingRoar yep, but 3070 for some reason is mostly on samsung or micron, 3060Ti is mostly hynix for some vendors so it's the most affected. Regular 3060 uses 2Gb chips, so it's unaffected.

[@sipansibabdreddknot5179]

@@SovietN1nja how about budget gpus neither amd or nvidia i was looking for 3050 or 3060 or amd's 6600 do you have any idea what vram they use?

[@ALostTerrapin]

Is it all 3060tis or just some? Is there a way to tell which memory it has before buying?

[@robreich6881]

I replaced the thermal pads on my RTX 3080. When mining the VRAM was going over 105C and throttling. In equivalent scenarios I got a 20C reduction by replacing the pads and adding pads to the back of the VRAM modules and the backplate.

[@longjohn526]

A good way to get rid of cigarette odor on about any device is clean them as best you can and put them in a sealed bag or box with 2 or 3 drier softener sheets and let them set for a week. I don't know exactly how it works but it does work to eliminate the odor. Somehow the drier sheets will absorb the odor

[@Jeter1230]

not smoking in the first place works too

[@phillgizmo8934]

@@Jeter1230 Duh, the safest way to live is not to live at all, so what? I'm not killing myself.

[@macicoinc9363]

Ozone machine. I believe you can go and take it to local places that have them for pretty cheap.

[@PlayinWithMahWii]

@@Jeter1230 I would assume OP was talking about getting rid of cigarette odor on *used* cards, not on a personal card.

[@matrixyst]

@@phillgizmo8934 i mean, if we're splitting hairs here, by smoking you kinda are :P either way it's not like you can control whether or not a stranger on the internet happened to smoke while at their pc before selling you a used GPU

[@LawrenceTimme]

Totally agree on Radeon vii. They are old and have been mined on for a long time now. They aren't even reliable for mining, you can just turn the off and back on again and they get code 43 error.

[@Heinz76Harald]

yep, just like that :D

[@wertywerrtyson5529]

Good warning regarding the thermal pads. I remember destroying a used 4890 back in the day when I changed to an aftermarket cooling solution and lost 100 bucks. I can’t imagine doing something like that on a 1000 dollar card.

[@Ascendor81]

Also use "NVIDIA GRAPHICS FIRMWARE UPDATE TOOL FOR DISPLAYPORT 1.3 AND 1.4 DISPLAYS" on some older GTX cards, as they might give black screen under UEFI with display port cable, HDMI should work. Older CSM works fine.

[@-eMpTy-]

Yep, had to do this on every 10-series card I owned yet.

[@matrixyst]

​@@Nighterlev checked the firmware tool out of curiosity after reading this thread, i was already running the update thankfully but i never even knew this was a thing

[@Spawny3]

Just a tip when looking for a defective GPU you might be able to fix: I'd avoid any RTX 3000 cards with "has video output, but can't install drivers" or "has video, but can't change resolution" or "not working properly, code 43 error in device manager". These issues are probably caused by one or more broken VRAM chip, broken/disconnected solderpads underneath one or more VRAM chips, broken or shorted traces from VRAM to GPU die, or defective GPU die. These defects are usually caused by a combination of heat and GPU sag, causing mechanical failure.

[@baraka629]

if you have a reballing die and a soldering station then many of those are quite straightforward fixes. for shorts just taking a few probes with the multimeter is more than enough.

[@KrazzeeKane]

​@@baraka629Jesus Christ lol, what an asinine comment. If someone has a reballing die, soldering station, and the skill to reflow the entire thing, they don't need advice about this on KZbin comments lol, they already know this

[@minus3dbintheteens60]

Vega HBM gets mental hot, I'd hate to know how hot Radeon 7 HBM gets, plus HBM tends to get overclocked to all fk with no sympathy. Even GDDR6X gets as hot as the sun, that's why all these 3080s and stuff have memory sitting at 105c out of the box. It's not mining that kills cards, its the stupid heat these chips produce and card makers doing stupid AF cooling for them like 2mm steel mid plates to cool them. Meanwhile Polaris cards are how old? Mining just fine. My 290x 8GB been mining 24/7 since March 2017. Actually, I have had 2 cards with dead fan speed controllers, out of about 15 cards. Just put 2x 120mm fans on them and enjoy the upgrade

[@mattiasolsson2499]

Agreed, 290/290X with reference coolers died like flies from mining as well back in 2014>2018. While cards with AIB coolers using the same PCB did much better.

[@liviuganea4108]

Bullshit. Mining IS the thing that kills the cards. If they weren't used for mining, the memory would be fine.

[@marcogenovesi8570]

@@liviuganea4108 mining is loading the card 24/7 vs gaming that loads the card for a few hours per day at most, but the problem is still that the heatsinks are inadequate. If you were gaming 24/7 (in an internet cafe or doing rendering work), the cards would still die th same

[@minus3dbintheteens60]

@@liviuganea4108Let's hear all about your mining experience, how many cards you use, what type, how many died, in what way, what brand of memory etc. Because the answer is fk all isn't it.

[@waifuhunter9709]

@@minus3dbintheteens60 lets hear about a scumbags integrity while a scumbag miner tries to defend it. Because answer is a scumbag does not have any integrity 🤷‍♂️ I dont need to have experience in stealing to point out someone stealing something. Miners or better i should say, scumbags are reason for current gpu condition. Mining does kill gpu. Not because it runs 24/7 but because of process. Be aware of scumbags trying to defend themselves and validify their action🤷‍♂️ and never buy from a miner. Their words means nothing, their whole identity is to ripoff others for money

[@minedustry]

Even before mining was a thing. The used video card market was full of cards with excessive fan noise.

[@moes95]

Yea duh just reapply thermal paste and pads

[@kernelpickle]

I think you called it on the Radeon 7’s, that sweet bandwidth for mining is the only reason anyone would buy them. The professional users who might’ve wanted them were better served by faster GPUs, and the only use for high bandwidth cards where core clocks don’t matter is ETH mining. I mine on my GPUs and it’s only like 4-6GB of VRAM in use for that, so on an 8GB 3070 I’ve got enough VRAM leftover to watching KZbin in HD and play older games that would run on older 2-3GB cards. The only thing I do when I’m gaming and mining is flip my settings back to default from my memory OC, lower core clocks and undervolting. Any of the Ti’s are going to be LHR models, so only the first run of non-Ti cards that were sold at the end of 2020 or early 2021 are more likely to be good mining cards. The LHR bypasses were only just released in the last couple of months or so, so don’t even worry about it. If BZ says HBM cards don’t hold up to the heat or heavy use avoid those and 3090’s because all the mining forums are filled with guys having issues with 3090’s due to memory overheating issues and power draw. My 3070 and an overkill heat sink on it so my memory is super cool, and the undervolting helps keep the heat down as well, so only thing I’ve put more wear on in any real way is the fans, but it runs silently most of the time anyway when it’s not running my custom settings. If it runs and you get a good deal, then it’s a good deal, and you shouldn’t worry about it. If the cards were gonna fail from mining they likely would have by now, because usually it’s sketchy power supplies and janky setups that are more likely to kill the cards than the mining itself. Only other thing I would look out for are used RX580’s because there’s a chance the VBIOS is modified and there’s a chance they fucked it up and if it wasn’t a dual BIOS card, then it might be a brick or it’s going to run like shit for games. I don’t know if other cards were known for running modded VBIOS, but RX580s were beasts with a custom flashed VBIOS.

[@mahmutdikcizgi9773]

I mined vega56 back in the day for 3 years without a single break, hbm and cpu temp never went above 60c, i sold them afterwards to a good price stating that they were mined, i even had logs of temps, and the buyer is still using it happily i think after 3 years. But it really depends on who is mining. I know lots of people overvolting and overclocking to mine, which makes no sense for the profitability specificly hbm based mem bandwidth mining and the lifecycle of the card will also degrade. So youll never know for sure

[@brulsmurf]

yeah, its a crapshoot. if you have free energy you just blast them at max power

[@mahmutdikcizgi9773]

There is no such thing as free energy. There is energy that someone else pays for you.

[@brulsmurf]

@@mahmutdikcizgi9773 solar panels. I cant return it to the grid.

[@mahmutdikcizgi9773]

@@brulsmurf in that case, your hash per watt will decrease, so youll lose/waste, money/energy. The important part is efficiency in mining

[@Fiztex553]

Man, I HATE thermal pads too! But with this abomination design of 30-series cards from NVIDIA partners, getting a reasonably cool VRAM on a 3080ti or 3090 was outright impossible without changing the pads... And because of the shortage you couldn't just pick some specific GPU you know is built well.. I've lost count on how many times I've dismantled the cards to balance out the temps and how many different pads I've tried. Ended up with a soft gp-extreme pads of different thicknesses and got VRAM temps down by whopping 20C on my cards. Such a lazy design these cards are for such a crazy price tag, this is insane.. I'm curious what's waiting for us with 40-series and their power appetites. Current cooling solution they have is at the limit already.

[@DatzBunz]

I had to do the same as well on my 3080 ti from MSI. dropped my vram by almost 15c and the core went down almost 3c too.

[@Fiztex553]

@@DatzBunz In my case core temp actually increased by 2C, I suppose it's because significantly more heat started to go from memory into the heatsink. But core temp was initially low and it stayed low, so it was a good trade-off, I think.

[@DatzBunz]

@@Fiztex553 definitely normal and good trade off. My hotspot went up by like 3c I believe on mine. Had to undervolt mine as its really hot here in the summer time and man just a simple undervolt to it dropped temps like crazy as well. These things consume so much power for such little extra gains its crazy. I came from a 3070 ti and that never ran hot or used insane amount of power. Mainly went with the 3080 ti for the vram headroom but man I might do the opposite of most people and find a lower clocked vbios for it lol. My case too has like 10 fans too so im maxed on cooling ;/

[@Fiztex553]

@@DatzBunz Indeed, a proper voltage/frequency curve is a key in setting up the 30-series. One can end up with a boost frequency "overclock" which is actually hurting real life performance as the card starts to drop voltage and thus frequency when reaching a power limit. I'm doing VR simming, so I'm pushing my 3080ti very hard and I found that an effective overclock strategy is to find a lowest possible voltage at which the card still holds >1900 MHz on the core combined with highest power limit available on the card. This gives me stable performance with higher average fps and less stuttering in VR than stock settings, even though the boost clock is technically lower.

[@T.K.9]

Imagine the lower end AIB GPU's. (Zotac, Inno3d) as example. Specially Zotac where the GPU design is even worst than the founders/reference. Cutting corners to maximise profit. They use the crappest materials and the most stone age power delivery design. Designed to run as hot as it can possibly can but still "good" enough to pass in QC. Also the fact that the auto boosting behaviour push too much voltage on the card. When infact undervolting it made it run cooler + actually had a 1900+ MHz stable boost. VS the abnormal boost behaviour that overheats the zotac's and drop to the advertised 1700mhz edit: the inno3d twin edge 3080's also are notorious for bad QC. Some cards literally have a "hanging" cooler/thermal pads. Where even if fully screwed on, the thermal pad is stuck on the cooler's cold plate not even touching the memory chips.

[@connerstines1578]

I've always been suspicious of the reliability of the HBM GPU's. I guess I wasn't mistaken. It's probably the same interposer failure that plagued a lot of different chips in the mid- late 2000s, which is why you had the red ring of death issues with Xbox 360s, the PS3s yellow light and a lot of cards and laptops from that era are dead from the same issue, the die lifting off of the interposer and disconnecting.

[@Wasmachineman]

interposer != substrate

[@connerstines1578]

@@Wasmachineman Yeah I get them mixed up a lot.

[@saraf01]

vega 56 original adopter and still runs , plays cyberpunk. to be fair I had a golden chip. my memory would oc high as well which is strange.

[@connerstines1578]

@@saraf01 Yeah I just recently upped from my Titan X I got a few weeks after launch in 2015, still played every new game I played on it fine at 1440p almost 7 years later, and still works wonders. 1000 bucks seemed unreal for a single GPU card back then, but it payed off in the long haul. I caved and got a 3080 when the prices became reasonable. I had a Radeon HD 3870x2 back in the day(2008ish), and it lasted about five weeks before it died, most other cards weren't much better then, maybe I just had bad luck with the silicon lottery.

[@tobiwonkanogy2975]

@@connerstines1578 I was silly and bought the tesla version last month I think . still haven't got it to work , I think I need a K series Quadro for the extra support .

[@cracklingice]

My avoid list is very similar. Anything interposer (fury, vega, radeon 7, titan V) Anything polaris x70, x80, x90 Anything RDNA 1 (rx 5000) with more than 4GB of VRAM Anything GDDR6X. Polaris and RDNA 1 cards were popular with BIOS mods that improve mining performance but harm gaming performance. It can be a headache to deal with modded cards, plus they've been in the mines for a very long time at this point.

[@gbitencourt]

This

[@m8x425]

I haven't had any problems with any Polaris cards that were mined with..... aside from the previous owner not reflashing the correct VBIOS back on to the card.

[@cracklingice]

@@m8x425 Yeah, for technical people who have a fallback display output to be able to correct a bad flash they can be a decent buy. The problem is they are still outrageously too expensive. They're selling for like $150 for a 580 8gb and that's more than double what I'd even remotely think of paying for one.

[@dat_21]

Basically don't buy ex-mining GPUs, unless you love gambling.

[@m8x425]

@@cracklingice Yeah, that's a bit high. There are people selling them in my area on CL or Offerup for $120.... but I've seen people selling FTW3 1080ti's for $300. I think it was back in 2019 I bought an 8gb RX 480 Strix for $90, a group of the Sapphire Nitro+ (light blue shroud) 8gb RX 580's for $100 a piece. I'd also seen RX 570's listed for $60 back then.

[@zmmj2024]

Radeon VII's die for 2 reasons. 1) Cooked from high temps 2) The interposer from TSMC is notoriously unstable. The high temps cause the interposer to fail faster, so that's why mining kills them. G6x does age incredibly quickly when held at high temps. Thrashing the G6x with activity pushes it's temps up really really fast.

[@MiningTips4u]

the question is what are high temps? are they closer to the throttling temp (106 or 108) or is 100+ bad.

[@surewhynot6259]

@@Nighterlev above 70 seems incredibly low for that to happen

[@andersjjensen]

I'd like to see some actual hard fact on the claim about the interposer itself. Like, the bottom chip in the first iterations of HBM got stupid hot, so who's to say (without taking shit apart and putting it under a microscope) that it isn't the micro solder bumps under the HBM that gives up the ghost?

[@lm_dccxl4078]

sadly is almost imposible to know if a card was on mining, so good luck to everyone who is planning to buy used this season. i got a xfx 5700xt raw II for $300 a month ago and im suspecting it was used for mining cuz the fans now have an awful vibration at 1600 rpm, a replacent will be good but theres no og or a decent review ones on amazon. edit: im planing to do exactly what buildzoid said, even theres decent 4pin gpu fan to dual pwm case converters on amazon.

[@patrikjankovics2113]

Ghetto modding even makes it way quieter, so...

[@Viennas_Sausages]

bought a used 2060 and it has that weird noise as well but i think it was due to me ghetto cleaning it without airduster but otherwise it works fine

[@sgredsch]

ghetto modding gpus is generally preferable to stock coolers because the 120mm fast are SO SO SO SO MUCH better at everything than those small flat gpu fans.

[@TKIvanov]

Im hoping rx5700xts and or 6700XTs get that cheap around here soon

[@SviatoslavDamaschin]

Just dismount fans and oil them lol

[@TKIvanov]

I had the mining bios thing happen to me. Bought an ex-miner RX570 that had a mining bios on in that the seller didn't tell me about. Spent a couple of hours trying to figure out wtf was going on (dont know why I didnt realize what the problem was earlier). After realizing why the fckin drivers don't want to install I quickly found the correct bios on techpowerup and flashed it and everything was fine (except the fact that the Strix RX570 is no strix at all. Asus just called it that and put a small ROG logo on it, but neither temps nor noise were good. Changing the thermal paste didnt achieve much (if anything). After a few months I got too annoyed and slapped a cheap AIO on it)

[@iguanac6466]

Agreed...those ROG Strix 570's were garbage. Final cash grab/GPU dump at the end of the first big mining boom.

[@BeHappyTo]

had same thing with my 580 nitro+. Performance was different than what I saw on youtube comparisons. Changed bios and everything worked well after. I bought it years ago for like 125usd. They still go for twice that price (used) in my central europe country. I was happy when I heard the gpu prices fell, but I don't really see that here on used or new. My rx580 still works fine, thought one of the fans been getting louder. Right now I'm thinking about upgrading my display to 1440p along with beefier gpu. But I might also just wait for ryzen6000 laptops to fall in price or even build a 6000series SFFPC and sell my desktop.

[@TKIvanov]

@@BeHappyTo im in eastern EU. Got my Strix 580 for about €130. Prices skyrocketed, sold it for about €350 and got the 4gb 570 from my earlier post for about €90. Now I have my fingers crossed that second hand 6700XTs would hit ~€300-350. But as you said they still have a lot to drop. Cheapest I've seen is €400 for a low-end model.

[@funderburke43]

I got a 580 from a guy who thought he bricked it by trying to flash a mining bios on it. He failed to put the mining bios on it entirely and never tested it as a primary card again.

[@DeadNoob451]

@@BeHappyTo it looks like GPU prices have stagnated In EUropa since April. You can see this in the price history of many current gen GPUs. Its like the inflation is increasing the price at the same speed as the crypto crash is decreasing it. WTF.

[@OfficialJamesNewberry]

All of your content is very informative, but the stuff I care about most is Watching you overclock current/last gen CPU/RAM/GPU, especially with all the random oc videos out there. You are usually the guy that actually knows the correct way to do things. Appreciate what you do, thanks.

[@certaindeath7776]

EVGA has the best warranty for 2nd hand market. so keep an eye out for evga cards which are still on warranty.

[@Battleneter]

Depends which country your in, outside North American EVGA warranty is nothing special

[@kekeke8988]

Are you saying EVGA warranty still applies even if you are not the original buyer?

[@LastSecBloomer]

I bought a Strix 1080Ti from a miner some 2 years ago. Fans were fine, I replaced the thermal paste (which dropped the temps significantly) and it still runs perfectly.

[@South_0f_Heaven_]

Bought a ROG STRIX card as well but it was a Liquid Metal fail, was artifacting. Cleaned that junk off which took a long time but same thing mines never skipped a beat in the 2-1/2 years I’ve owned it.

[@marcogenovesi8570]

As for the thermal pad argument, that's one of the main selling point of the "thermal putty" compounds. Depending on how pliable they are they can make a horrible mess but should be fine for thermals and won't be too "high" so putting mechanical stress on the board/solderballs/whatever. Also afaik they are not that cheap. But they are easier to get than specific size/height/consistency of original pads

[@Magnulus76]

How hard is thermal putty? It seems to me it could still put stress on the card if it doesn't compress alot.

[@Megalomaniakaal]

@@Magnulus76 it'll deform, but not in a temporary fashion like foam.

[@gaunterodimm3606]

The problem with VRAM on mining cards isn't just heavily reading and writing, but heavily reading and writing while overclocked and doing it for a long time, non stop. Just electron migration alone could potentially cause degradation and that's not to mention SMD wear.

[@bariole]

No miner is overclockig cards. Miners strive for maximum power efficency which is usually accomplished with undervolting and undeclocking.

[@gaunterodimm3606]

@@bariole Bullshit, the core may be undervolted and underclocked but the VRAM is likely overclocked and gets thrashed. It's obvious you've never mined ethe.

[@opinali]

Second the advice about thermal pads. I had a hard time when I bought my 6800XT Midnight Black. It started crashing a couple months after buying (new but from stockx so no warranty). I saw that the fans were never ramping up with load, they were stuck at minimum rpm. So I disassembled the card, it was probably just loose header cable, but to be sure I also replaced all original pads by new ones using the best stuff available yes the high-K harder pads. That was a nightmare, now my fans worked but the card would shut down immediately if I launched any game, eventually it wouldn't even POST. I had to redo everything a couple times more until I got everything at perfect thickness including the GPU die (combining graphite pad and paste just like the factory build). In the end I got the card running cooler than day one thanks to better pads including lots in the back but not a recommended experience.

[@Heinz76Harald]

this sounds like you had a lot of fun with the card, feel ya man.

[@opinali]

​@@Heinz76Harald Guilty as charged a have fun taking things apart and tweaking them. But let me tell you the part that wasn't fun, the weeks between the first crashes and the final fix, when I thought that I had burned $1,3K at the time on a lemon card (bought from stockx so there's no return after just a week, and no warranty since you're not the original buyer)...

[@noanyobiseniss7462]

I've bought used mining cards from fleabay since4 2013 and resold them during mining peaks and have had no regrets whatsoever. I never had a failed one. I look for those that have had been bios flashed for undervolting.

[@konchy]

How do u look for those that have been undervolted? Do u just ask the seller?

[@tag206]

I bought an rx470 that was mined on for 3 years, and I have daily used it as an htpc/gaming card for the last 3. Minus having to reinstall drivers every other update, it just needed repasted. And the fans needed replaced, but it was an xfx card with easy swap fans. Hindsight says I should've bought one of those 2x80mm fan bracket things off Amazon for 15$ instead of 15$ a fan for replacements.

[@alexusman]

Be wary of 3060ti (both used and new), as they almost always use Hynix 1GB GDDR6 chips as of late and those die from a slight overheat. And the only way to fix them afterwards is to swap all 8 chips of VRAM for different vendor like Samsung. Becasue while you use the hot air gun to swap 1 chip, all chips around it die from that.

[@Lobonova]

my 3060ti's do much better than me. i cant sit in the room for long.

[@nickpetriashvili2863]

All the models of 3060 ti?

[@alfredgamerzzgamers3970]

Is this all model of 3060ti

[@80sBaby925]

@@alfredgamerzzgamers3970 no not all models.

[@DylRicho]

Happy with my ex-mining RX 570. The seller said it wasn't used for mining for very long. No idea how true that is, but I've had it since April 2021 and it's been brilliant. She seemed genuine enough. I did have to flash the BIOS, but that was easier than I thought it would be (hadn't done that before because I've never needed to).

[@VanishingPoint96]

My previous RX 570 and current RX 470 are both working great and were ex-mining cards. BIOS flash and a deep cleaning and they were as good as new

[@ronakparikh]

I bought a used 1070 fe when miners were dumping them after 20 series launched and BTC fell from its peak of $20k (I think it was at least 4 years ago now). It still runs great and even looks brand new, no signs of wear on the fan even. Buying it used worked really well for me since the previous owner took great care of it, but it is definitely very ymmv.

[@CasualGamers]

A GPU fan modding video would be interesting

[@supersaiyan17online]

Is it though? It is literally just taking off the old fans and zip-tying new ones on.

[@maicondouglas8047]

salve

[@TheLateral18]

X2

[@Dandan-tg6tj]

Yes, of course. This way there will be so many people trying the mod and fail. Then I will have more clients and of course I will ask for double the price I am asking from someone who is coming to me to do it professionally and not trying to do it themselves. I barely can't wait.

[@maicondouglas8047]

​@@Dandan-tg6tj Easily i can find a lot of videos of random dudes killing her gpus aplling tthermalpaste using dot method. In fact users cant be underrated.

[@ThatKoukiZ31]

I bought an exmining RX580 in jan 2020, and its been solid. I even mined on it last winter since I was paying for heat I might as well profit from it. It was a wash with crypto crashing, but thats absolutely fine with me since now GPUs are becoming normal. Still this old card marches on.

[@bmoney2324]

After mining on most aib 3080/3090 models, i would recommend Strix models as they had the coolest vram temps stock cooling in my mind experience. Avoid gigabyte models as they are terrible at stock and need thermal pad replacement. If you buy one of those from a miner, make sure the thermal pads were replaced or the cards likely ran at 110 deg C vram for long periods. I would also guess that Zotac/PNY models have this issue as well. The FTW3, XC3, MSI Suprim, Trio and Asus Tuf 3080 models would generally be around 100 deg C vram while mining with stock cooling. I would feel ok buying any of these 3080/3090 models second hand.

[@eurobeatbandit]

What about memory degradation of chips other than you mentioned? I personally have exmining RX580 from previous mining craze and for a while everything was ok. Later memory started to throw artifacts from being worn out, and now only way to keep the drivers stable is no boost whatsoever on memory clocks and +30mV voltage.

[@yspegel]

absolutely agree with replacing fans with case fans. Did that on my 1070 in the past and it gave a good performance boost besides being more silent. Just used cable ties to strap them on, perfection!

[@SylvanFeanturi]

How did you deal with powering and controlling the speed of those case fans?

[@yspegel]

@@SylvanFeanturi had the case temp sensor attached near the card, the fans attached to the motherboard and set a fan curve for case temp. Works like a charm.

[@SylvanFeanturi]

@@yspegel Okay, thanks for the tip! May do for something like that if I don't manage to get mini pwm -> pwm plug on a cheap :D

[@prep0wer]

AMD stopped making the VII when it was the most powerful deep learning GPU and still is. It's the only high end video GPU that's still usable on the Apple platform or other Linux based workstations.

[@prep0wer]

white blue and brown dust

[@N0N0111]

I want to say these thumbnails are lit dude, they go so well with all you do keep them coming!

[@corsairsloop3234]

I replaced the thermal pads on my gigabyte card because the stock pads leak oil from normal use and have trash cooling peformance as well. I used thermalright they are a bit harder than stock. But what I did was heat up the gpu before reassembling it to make the new pads softer. Memory Temps dropped massively and no negative peformace on the core where I used noctua nt-n1.

[@iguanac6466]

EVGA pads on my old 1070 did that as well. Scared the crap out of me.

[@SpuriousECG]

So far I'm 0/4 for working 2080 Tis from EVGA B-Stock, all have had green or purple artifacting, Code 43 errors, and low resolution only output.

[@K31TH3R]

I've mined for over a year on my 5700 XT. It's on an EK Quantum Vector, it runs at 794mv @ 1396MHz. It's on a custom BIOS with tightened GDDR6 timings that is suitable for both mining and gaming, and the timings even boosted FPS by about 2.8%, the VRAM runs at 1800MHz. Max core edge temp while mining is 40c, TJ at 46C. GDDR6 temp maxes at 55-58C depending on ambient. VRM runs at 42C. Eth hash is 56.12MH/s at 95 watts. If someone offered me that kind of information about the card I was buying, I'd buy it without a single second thought. I'd also have zero questions about reliability about buying my own card. In fact, I'd be even more willing to buy a card which was mined, especially if temperature and power metrics were given, because chances are higher that the card was mined on than not, and if they're not upfront about it, then it was probably ran hot and carelessly.

[@Dennzer1]

Keep doing video's that are just you sharing a thought you had, and so it will allow the channel to have videos that we wouldn't always expect.

[@sznikers]

Remember during 2017 cryptocraze 1080Ti was 1000$ (vs ~ 600msrp) and at the bottom of despair after crypto crashed they were 200-250$ on craigslist. The crash was Dec 2017 - Jan 2018 and the bottom of despair for miners came 2019. In 2021 GPUs went skyhigh again due to next crypto peak amplified by supply crunch. I sold my used 4yo 1080Ti for 500$ thanks to that. Truly absurd price for chip that old. Looking at time scale (crash was nov-dec 2021) and general sentiment in that market I dont think we have reached the bottom for GPU's yet. This may be bottom for crypto but we need it to stay flat line for few months before true capitulation starts for miners. People still hold shitcoins and believe they will be millionaires next year... They need few more months for reality to check in ; )

[@kaisyaya8492]

the difference between 2018 and 2022 is that the crypto deluded so many people that even with 7 month of being down, they think it will gain value with some weird ass twett from Elon or something

[@AAjax]

Lots of youtube videos on "gpu price crash" when now-old-and-used cards are just a hair under MSRP. While profits are way down, for some regions electricity costs mean mining is still just-barely profitable. Others have shut off rigs and are waiting to see what happens, not selling their gear. They know what happened in the previous crash too. I don't think we'll see full capitulation until a month or two after the Eth merge, when every miner witnesses the extreme loss they'll be mining at.

[@fontyyy]

Profits are low, but even rubbish Polaris cards are still profitable if you've got cheap electric. We're no where near the bottom yet, you're going to see 30 series cards way under $300. 10 and 16 series cards will be hard to sell at all as there are litterally millions of 30 series cards going to be dumped. Small miners who can pay their electric out of pocket and do it partly as they love it won't stop, but all those massive warehouses with employees and rent to pay are doomed when ETH goes POS.

[@dazley8021]

I bought two ASUS Strix 3090 recently on ebay for much less than they usually go used. (800€ each) The seller didnt mention any mining on them, but he was selling three of them of which one was sold already, so i got the remaining two. The only thing he mentioned was he was using it for CAD. He is the owner of a shop for doll supplies and toys, oddly enough. He also mentioned that he changed the thermal paste and pads. He also was selling a couple of powerful PSUs, a rather weak CPU and a metal shelf on his Ebay shop. The cards were in quite a pristine condition tho, very little scratches, very little dust. The fans are not rattling. Heck they still even smell new! BIOS seems to report correct informations, tho the BIOS switch was flipped to Performance. The only thing really weird about those cards is they get hot quite quickly. I run Furmark on them for like 40 minutes and they got to a toasty 80+C but remained there. HWM reported hotspot temps of about 100c tho.... But no artifacting or anything and 3DMark score is around 19k for GPU score in time spy and around 13k in Port Royal. I'm just wondering if this is normal behaviour? IDK how to tell if they were used for mining tho, are there definite tell-tale signs that are easy to spot?

[@SerpentXTech]

Agree, original 3090, and cards with HBM are trashed in the process. Some run within or below spec, but 9/10 most just copy someone else's clocks and don't consider their environmental conditions / thermal conditions might not be similar to the users whose clocks they just copied. Running their hardware well beyond spec, and as you said some are not as good from the jump, i.e. 3090 Ref. 3090's, 3080's, Radeon Vii, and Vega can be bad buys, leading to an expensive paperweight for the buyer. However, not all are bad apples and some users do take care of their hardware. The problem is the seller can just lie (as you mentioned in our discussion) and trying to find that good user who took care of their hardware and honest, is rare these days... Thank you for content as always. Take care

[@KillaBitz]

the price of Vegas is higher than the 1080 because they do 50-55mh on eth instead of like 38mh same idea with the VI, it does insane hashrate on eth (but will up and die if you fart near it)

[@SwordQuake2]

16:30 no, no, you're right. People who smoke next to their computer are disgusting. I'd go as far as to say that all people who smoke are disgusting.

[@tourmaline07]

I was thinking exactly that :)

[@DrakkarCalethiel]

Had to clean a smokers PC once. What a disgusting thing that was! Smelled like an ashtray. Hardware was all sticky, dust wouldn't come off the heatsinks unless bathed in white spirits for at least an hour. Took two days to clean and fix because of that shitty tar and smoke residue. The fans were the worst.

[@applicablerobot]

Gpurisers has 20W/mK pads that are soft and work super well. Kritical pads has kits with precut pads for a lot 30xx cards. Haven't tried them myself but I suspect it's the same material as gpurisers

[@ActuallyHardcoreOverclocking]

intresting I might check those out.

[@applicablerobot]

@@ActuallyHardcoreOverclocking I can't vouch for kritical pads, but iirc gpurisers has a UK warehouse. And gpurisers pads come in 0.25mm increments.

[@kajurn791]

The suggestion of replacing dead fans with case fans or deshrouding is a great idea as well. Not only you'd get better performance, one would likely have to clean/replace thermal paste on the card anyway, so that's a good excuse to mod it for better thermals and noise.

[@LetsPlayBojangles]

Some buying tips to protect your money. Buy on places that have buyer garuntees like eBay. You have no recourse if you get sold a dud card on Craigslist but ebay you can basically do a charge back if they don't refund you. Also you can get some level of manufacturer warranty If you play your cards right. EVGA and MSI will do second hand buyer RMAs but it's from the manufacture date. You could get 12-18 months on current gen cards or even longer on 3080 tis and 3090 ti. Just look up the brands policy before you make the purchase. I myself bought an 1080 ti that was mined in, it ended up dying 8 months later but because it was EVGA it still had like 6 months of warranty left and they sent me a 2080 as a replacement. If the original buyer never registered the product. You could ask for the original proof of purchase and get the full warranty. You might just need to make a second account if you know what I mean.

[@Apothecarii]

I mined with 3 Vega 64 cards. Two were the LE models with the aluminum air coolers, and one was the binned LC model. All mined for over a year at first. After mining, the two air cooled cards would not be able to hit anywhere near they rated frequencies even with thermals under control. It is possible these were defective from the get go. It was just never noticed since they were heavily downclocked and expierenced no 3D load until I tried to game on them. Or some part of the power delivery had degraded (shurg). The HBM seemed fine. The LC card has now mined for 18 months and is now gaming just fine in my son's computer. Just my experience.

[@PocketDrummer]

Honestly, even if it's a great deal, we shouldn't buy used mining cards. We should let them eat the inflated costs of those cards so that their ROI goes down and they hopefully give up on mining.

[@splindizzle5413]

couldnt agree more

[@8lec_R]

I would love if people just stopped mining. But I don't think they will. They'll probably wait for the price to go up again and mine again, the next time more cautiously

[@seeibe]

In general right now we shouldn't buy any GPUs unless we absolutely need them. Let scalpers and miners sit on their cards until prices go down a lot more.

[@mylittlepwny3447]

Who cares? If the price is right I'll buy anything!

[@InsertValidName]

@@mylittlepwny3447 i bet you start caring if you save 20% at buying but the card is dead in weeks

[@Premosilva]

Don't buy any. Let the prices fall even further 😎

[@InternetEntity]

*cradles Radeon VII* Please don't die! Mine has been gaming hard for the past couple of years and overclocks really well. I'd better hit HWBot sooner rather than later. Hope there's a really cold spell this summer (UK). Or it lives at peak performance until Winter.

[@myagi20000]

I've had my Radeon VII since launch and it still works great primarily gaming on it

[@jerrieonly5599]

Aw sick man what kind of clocks are u getting on your Radeon VII

[@jerrieonly5599]

Thats sick I can only imagine the FPS

[@myagi20000]

@@jerrieonly5599 i've limited the FPS on the card because otherwise it gets really loud

[@MoraFermi]

Regarding Radeon VII, I'm pretty sure that the problem can be traced to just the cooler quality. The reference design has a large vapour chamber with mounting posts staked directly to it with just paper-thin graphite "solid" TIM on top of it. That combo is super sensitive to manufacturing tolerances; on top of that the default BIOS likes to run that card ULTRA hot, regularly hitting 110C.

[@Heinz76Harald]

wrong, this happens to watercooled cards too. its the internal parts which fail, not the type/quality of the cooler. sure they die on heat, but you just cant tell its the coolers fault.

[@richwhilecooper]

I just bought an MSI 3080ti gaming x trio from ebay. Not in original box. So far so good. Runs very cool. No issues following a lengthy stress test.

[@REOGURU]

If the fan on your GPU is giving out, it's a good time to refresh the thermal paste and pads. There are plenty of videos on KZbin showing you how to do this, including the size of the pads specific to your GPU. I've had the same 6 x 1070 ti's (18 fans total) mining consistently since 2018 and performed this task successfully a multitude of times.

[@righteousone8454]

I bought an ex-mining card (GTX 1080), it didn't even last a year. Don't.

[@falconn121]

Ive been using 10 series for mining purposes Ive been running them for the last 4 years

[@gubercc411]

I have a Radeon VII. Bought it for CAD drafting when they were cheap for the 16gb. You nailed it. It is serving me great. But i modded the heck out of it, using thermal putty instead of pads, liquid metal, and a full waterblock to keep the hot spot and edge temps from differing to much. Since the HBM2 memory is on the die, this is bad. Stock, can differ by like 30-40°c, which can't be healthy. But mine stay within 10c. And i have never had issues.

[@zmmj2024]

Mining on it will degrade the interposer, so if you care about the card, retiring it sooner rather than later, is best. I had a full cooled VII damage itself. It now lives with a heavy power limit and downclocked memory in a display driver role. It's just a when, not if situation.

[@gubercc411]

@@zmmj2024 I never used mine for mining. I bought it because it had the memory for the buck at the time. I mainly use it for rendering drawings, a bit of video games, and folding@home every once in a blue moon. Between f@h and rendering, i don't know what hits the interposer harder. I have it overclocked to 1920mhz core @1120mV (3mV increase from stock) and 1150mhz vram, with a 10% increase in max wattage. Do you think those settings will it wear out faster?

[@DimkaTsv]

Btw for some people information. At least there are some news about that from Russia. It seems that miners created burn-tool (basically shorted PCI-E riser, or one with swapped lines). It is to attempt to RMA GPU to either seller or manufacturer. So markets started to decline more GPU from warranty repair than usual, and you may find one of these in wild (aka ebay, aliexpress or local markeplace).

[@GrimK77]

All semiconductors can degrade in heavy usage at higher temperatures. It's just like burning out of LED sources, but much slower.

[@TheDude50447]

Recently heard from Steveon Gamers Nexus that apparently RTX 3xxx cards especially the high end can have transient spikes of iirc around 700 watts. Which is a good indication on why the vrms are exploding on 3090s.

[@frozenskeptic345]

Wow some hot takes in this one! Case fans instead of replacing the stock ones. Interesting take. Not sure if I would agree in all cases. The thermal pad section in this is very overblown imo. I may not have replaced as many as you have but its been at least 20. I have never killed a card with thermal pads. The worst thing thats happened is some heat pump on the thermal paste on the die but its pretty easy to catch in time if you monitor temps. I have used thermal grizzly to the cheap soft ones to thermal putty. I have even messed up thickness by multiple mm with thermal grizzlys on 1080s and it just pumps the paste out, its never bent the pcb or de balled any cores in my experience. Even the hard thermal pads dont exert enough resistance to break anything even if you are a few mm off. It wont run for long but its not like immediate death. And getting the correct size pads takes 3 seconds with a good measurement tool.

[@syncmonism]

This was one of the videos I watched to get info on used cards while I was shopping around for used cards over the last month or two. I eventually bought a used EVGA FTW3 RTX 3080. The transferable warranty coverage which EVGA offers gave me more piece of mind, as does the fact that the card has two bioses, and, of course, it also has a better than normal cooler. I also watched Buildzoid's teardown video on the FTW3, and I realize that the components used and board design was disappointing to a lot of people, but for someone who doesn't want to do anything more than occasional light to moderate overclocking at most, it seems like a great card, especially because I didn't have to pay a price premium for it over other used RTX 3080 listings (I paid about 580 USD including shipping). I'm still not sure if it was the best choice, and I'm not all that impressed by Ray Tracing, but at least I'm not too worried about the card actually dying on me any time soon. I probably should have just forgotten about Ray Tracing and gotten an RX 6700 XT.

[@andrey7268]

The problem with case fans is that they usually have a different connector. You either connect them to the motherboard and lose the fan curve defined by the GPU, or have to get the right connectors and solder the fans to them.

[@totalbullion5882]

As with most things, it depends on who you're buying from. I mine, but I I keep my gpu temps low and do regular cleaning on them. My advice for buyers would to be cautious of cards with gddr6X memory as I've seen way to many idiot miners running dangerously high temps or just letting their cards run while thermal throttling, I would not want one of those gpu's lol.

[@waifuhunter9709]

Oh thank you for information. I will never buy your gpus even if you sell them for 90% bellow msrp 😀 Yes its depends who i am buying a gpu, but miners are scumbags in general. So it does not depend on which miner i am buying from. Scumbag is scumbag. I am not that fool to take a proved scumbag words as face value and buy anything from them

[@johnezekielluna8040]

I've used my 2nd hand 1070 for mining for 6 months and before I can oc it +100mhz core and +200mhz. Now I can't even +50mhz both but it still runs stock and I still use it for gaming. Though I have put heatsinks on the backplate and thermal pads on vrms. It definitely degraded I would say but hey I think the "mods" worked.

[@giglioflex]

To be fair, that card is old enough to the point where some degradation from age is expected.

[@Nero9937]

As a guy who bought the red Radeon VII, and had it die on me, only to buy another one (but grey), I know all too well that the RVII is a very, very delicate card.

[@IIARROWS]

I haven't noticed any degradation on 3090 (it's watercooled though, on a single system). It's overclocked to +1000 "MHz" on memory (although these days I reduce it during the day to avoid reaching temperatures too high). I would trust mining GPU anyway, miners don't want to risk the GPU to break in any way, it would be unsellable and it would reduce your hashrate. Miners care about that, money is all that matters after all. If the card degrades or breaks it doesn't make money. They usually have extra fans blowing on the rig, and for 3090 they change thermal pads to improve performance and durability. Having a working card is an investment, remember that if they sell the card and the buyer notice it crashes, eBay will stay on the buyer side. Besides, constant heat is better on electronic components, those card keep 80°C or even 90 °C TJMax constantly is good, there is no thermal expansion/contraction and both the chips, pads, and board.

[@shiningthetruth]

thanks for this video! there are literally piles of ex-mining Vega 56 being sold for 170-220 usd in my area, I almost tempted to get one, but I guess I'll look for something else

[@ihurtyourbrain3685]

The cards would run fine. The problem is that Vega56 performance ain't worth ~$200 anymore imo More like $100...

[@applicablerobot]

Keeping original thermal pads won't be an option for 30 series gpus that have been mined on. Perhaps some 3070s and lower that were kept cool, but the 3080 and up all need pad replacements

[@DryadMachine]

I've found the tar from smoker's computer parts is far more of a problem than the smell. The tar forms a sticky surface to which the dust adheres. The Datavac won't touch it so you have to resort to lots of IPA and cotton buds.

[@mariotosic768]

Any electronics degrades over time, so does memory, id even say that degrades the most after capacitors. I know from personal experience on my hd7850, after 7 years it doesn't run stock memory clocks anymore and just crashes to black screen. I ran my msi single fan hd7850 OC at 60c on 2400rpm most of its life, with an overclock and undervolt. The last year of its life I had to lower the clocks every 4 months by about 50mhz until it just black screened one day. It was well used for gaming. Its using elpida memory and did +150mhz over stock when knew and for 5 years.

[@wiggetywut32]

I might be an outlier, but I have a heavily overclocked Radeon VII which has been running compute since launch week. Then again, I never experienced the driver issues the media spoke about. Just my 2 cents. Also mine is watercooled. But anyway. Cheers 🍻 edit: watercooled not watercolors lol. Autocorrect smh

[@MegaSegawa]

Great video ! What's your opinion on manufacturer OC cards, like the MSI Suprim 30 series ? Thanks for all of the infos.

[@neur303]

AFAIK any microchip does degrade by using due to electromigration. The question is how much margin did the manufacturer leave for it to not be relevant. If the margins are too small, unexpected heavy use might kill it early.

[@ActuallyHardcoreOverclocking]

well yes but I've never electrically killed memory chips. With the exception of degrading the HBM on an R9 fury with 1.45V.

[@monzitj1]

@@ActuallyHardcoreOverclocking Do you srsly believe that a GPU such as RX 570 with no pads on mem. chips hoping only the airflow will cool them is suppoosed to be healty after 4 years of mining?

[@zacharymorrison4974]

LMAO my gpu fan bearings are clogged with weed resin from ripping my bong all day

[@compwiz00]

According to hwinfo64, th HBM on my Vega64 uses about 5-10w while gaming and uses like 20w mining and the HBM gets much hotter. Samsung HBM gets glitchy and can't hold tight timings past about 80c Miners usually put Vega56 firmware on their Vega64s to allow lower memory voltage and temps.

[@DJSammy69.]

Excellent video! Been looking one of these!

[@BenTan89]

If you're afraid of aftermarket thermal pads killing your GPU, just use thermal putty instead (TG-PP10). It's quite fail and idiot proof because putty is well... putty. They are more expensive than pads for sure, but it brings the risk of frying your GPU core or bending the PCB down to near zero.

[@rakon8496]

Could you include chapter marks, would make it easy to jump to rants/topics of interest! Thanks! 💙

[@siliconalleyelectronics187]

I deal with a lot of used parts, in my experience the best way to get rid of bad smells like cigarettes is with an ozone generator. You can buy a good enough one off of amazon for $60, chuck it in a cardboard box with your stinky stuff for an hour or so and it'll take the smell out. It'll smell like it's been sterilized for a few days but afterwards will be scent free. I see a lot of BIOS failures on Vega cards as well.

[@T.K.9]

That explains why most mining gpu's bleed out all the oil from the thermal pads. Rendering them useless and in need of replacing since the mem chips will just run real hot.

[@SimmiesSchrauberChannel]

Hi Buildzoid! For the thermalpad-replacement you can go with thermal putty (stuff like K5 Pro). Tested it on Laptop and Smartphones and it works like a charm.

[@-eMpTy-]

They're quite messy tho. I'd rather get some decent soft pads from Alphacool.

[@kajurn791]

It's a shame the Radeon VII was so fragile and so good at mining. It had a cool name and the stock cooler looked amazing. Gonna wait a year or two(?) and buy a dead one for cheap to have on my shelf.

[@samgoff5289]

How is 7 a cool name

[@intensifiedhipster]

Well some of them went into old Mac Pros for video editing (because they are natively supported by modern macOS, and gave the best performance in Final Cut and other apps), so you may be able to pick up one of these in perfectly fine condition, after those people will start upgrading to Mac Studios and such.

[@NormanPAquino]

I found it incredibly funny when you were repeatedly warning against dead Radeon VIIs. :)

[@mufc1417]

Hi guys, so getting mining 6900XT for $550 is safe & good deal?

[@EliteRock]

Thermal pads; Gelid Extreme and Ultimate are soft (they're more like pre-formed sheets of putty - you can ball them up and use them as such) and amongst the best thermally (ahem, and the most expensive). I've replaced every pad on every component over the past few years (two MB's and four GPU's) with them. With GPU's I always disassemble again after fitting to check I've got the thicknesses right (having measured everything the best I can, clearances/gaps or the original pads). Temperatures _always_ drop markedly, if there's no sensor a finger test confirms the heatsinks are always warmer.

[@applicablerobot]

Fan failures are very common with 30xx cards. Idk why but at least 10% of my 30 series gpus have died or are dying after just 1.5 years.

[@propeldragon]

Which models? I’ve only had 1 gigabyte aorus 3060 ti die after 5 months.

[@applicablerobot]

@@propeldragon primarily zotac, but asus is a close second and I think I have a gigabyte 3070 with a stopped fan. Been a while since I took count, need to clean and repair all my cards but that's quite a chore and comes after other more important tasks...

[@Baz87100]

I saw a 3090 on ebay for £690 ($815) and immediately sent the link to my friend. After about 5 minutes of deliberation he decided to buy it as I said it looks like a great deal - seller has 100% feedback over 200 transactions and my friend has buyer protection anyway. The item is listed as 'in great working order' and well within transferrable warranty. I know this sounds bad, but we neglected to check the location of the seller and he is in Hong Kong - so to me it seems likely some mining was done on the card. Still, let's say the card lasts for 3 years with a 10% or less performance deficit then it's still a pretty good deal imo.

[@mitar35]

Does it still work? What are your thoughts?

[@marsovac]

I was mining and gaming for years on a Vega56 unlocked to 64 and HBM overlocked to 1100Mhz (+300Mhz) and VCore undervolt. The memory limit was 1106Mhz when I first bought it, and it does exactly the same memory overclock today. I think memory degradation is not a thing if memory is not overheating. And I made sure that I have no overheating (was always

[@jerrieonly5599]

Bro I got V 64 Liquid Cooled the most stable I can go is 1090 MHZ - I been Playing RDR2 at - 1800P Ultra 30FPS definitely a 1440p Card at this point

[@marsovac]

@@jerrieonly5599 That's too bad. Seems silicon lottery was not on your side :(

[@workstationmark4103]

I use less than 80 watts mining and use over 100 watts gaming. Temp Control is monitored and adjusted accordingly when mining.

[@hyperstimmed]

Unless you know the miner personally and know 100% for sure they properly undervolt memory, buying the card is just a bad idea period. A lot of miners don't know how to undervolt Nvidia. Or, they just don't give a shit about heat and want that extra 2mh/s. I wouldn't touch any 30 series that was used for mining

[@mikeymaiku]

Love the mis-information you spew lol

[@hyperstimmed]

@@mikeymaiku nothing I said was inaccurate, miner.

[@elecman748]

@@hyperstimmed b-but mining gud!

[@hyperstimmed]

@@elecman748 depending on your perspective it can be lol I paid off my 5700XT and 2060 by mining with them.. but that said I dislike the farms that create piles of e-waste. And in places without hydroelectric power nobody should be mining, but unfortunately there are huge farms in areas where they burn coal for power and it's creating a massive environmental impact

[@user-eq2fp6jw4g]

Totally agree on this one. Any memory ran close to 95 degrees nonstop is no good thing. Even gaming might end up getting "thermal cycles" its still better than pushing 24/7 @ almost max temp range that chip can handle. Not to mention all the other parts example the memory VRM on the budged midrange-highend models etc. It all narrows the general design of the cards

[@FerralVideo]

Additional comment on R7's. Very large die GPUs are already sensitive to interposer failures. R7s are also not known for running cool, and other comments saying that mining heats up the HBMs. Uneven heating kills substrates, and Eth kills memory. Another commenter said R7 is an absolute weapon at mining, which seems to all stack up to a perfect storm of graphics card murder.

[@devonmoreau]

Great video, I appreciate the thorough discussion of this topic from an expert!

[@JadeIsler]

+rep to all of this advice. In general, I've found miners take way better care of their cards (as they see them as investments) than the average gamer. This is especially true in "green states" like Colorado where I live. I'd take a card I have reflash the bios on over one caked in sticky green leaves. Plus, many of the mining cards I buy have upgraded fans, paste, and thermal pads, and the seller is often happy to add a high-wattage power supply to the bundle for cheap. The one exception that I've practiced (that you've mentioned) are RTX 3090's, where even good cooling and upgraded pads aren't enough to cool all that VRAM. The 3090 in my personal system has nerfed VRAM overclocking headroom thanks to the previous owner being a noob miner running ethhash without optimizing the power limit.

[@Fulmir-]

This is by no means universal especially for larger mining operations where the majority of these cards are coming from. The problem is a lot of these mining ops don't cool their cards sufficiently because that eats into profits and this slowly cooks the cards.

[@Fulmir-]

@@JadeIsler That's still not a great environment for these cards to be running in 24/7. You can't just push air in and out of a building and expect electronics like this to run happily 24/7 like these cards get used. There's a reason that datacenters spend literally as much cooling the data center as they do on the electricity to run the actual physical machines. You need that sort of active cooling setup to keep these things stable and happy over the long term, just pushing hot air out of the building helps but it still means the card is running hotter than it would in even an airconditioned apartment or house, and it puts a LOT more stress on the components over time since these cards are running 24/7. This is why the general advice is to not buy a used mining GPU unless you're okay with a high failure rate or getting something that just flat out doesn't work, because these cards have a lot of years taken off their lifetimes from the environment they're used in. When you say "they take better care of their cards than most users" and then talk about cooling with little to no active cooling of the air I just roll my eyes. The cards may be in better *physical* condition, but they're *way* more likely to fail before they'd otherwise be replaced.

[@mephistophelesgg7963]

miners deserve to lose everything, let those scum to rot with their gpus

[@TempleKa]

Bought a MSI GTX 1080Ti Gaming X back when the crypto bubble burst a couple of years back from a miner for around 220-250$. I cleaned it (even though it was pretty clean), changed paste and thermal pads (did research for the right thickness) and it still works flawless to this day. Hands down the best buy I ever made for my PC. I finally replaced it with a new RTX 3080Ti I got cheap. Maybe I was lucky.

[@MultiKokonutz]

My daily driver is an ex-mining vega56 that I bought in January 2019 for dirt cheap. Best purchase I have ever made

[@EBMproductions1]

Sadly no gpu is dirt cheap anymore people just got used to $400+ and think $250 used is "cheap"

[@fontyyy]

@@EBMproductions1 depends what you call "dirt cheap" 50% of MSRP is surely dirt cheap? Top end GPU's are big money, a used 3090ti for under $1000 would be dirt cheap IMHO, even though a thousand bucks for a GPU is fucking insane.

[@EBMproductions1]

@@fontyyy $30 to $200 for gtx 10 series and earlier is cheaper their very expensivs still.

[@fontyyy]

@@EBMproductions1 when ETH goes POS there will be literally millions of 30 series cards dumped on the second hand market. 10 (& probably 16) series cards will be almost worthless. The real GPU dumping has not even started yet, ETH is still profitable for miners.

[@WladislawNikitjuk]

I have been using Vega Frontier Edition for over 2 years now. Not sure if it was a mining GPU. Bought it as not working, flashed it with Vega 64 LC bios, edited register, but still cannot instal gaming drivers, professional only, but it is still ok for 2k