Anyone just given up on looking for a new GPU?

[MeatLoaf]

Well today I got called 'entitled' for saying I would absolutely not buy a 2nd hand GPU this year unless my current one outright dies because I'm not taking a ragged out, year old, ex-mining card for more than RRP.

So I guess miners are starting to get edgy and thinking about unloading their hardware? Miners getting out can't be bad news for anyone who wants to use their PC for games, as the good lord intended

Let us pray for real stock soon...

Ive never really understood why people wont buy mining cards. A lot of the time there not 'ragged out'. They typically have a decent life undervolted and looked after. But saying that it would depend on the price

 

[FoxEye]

Ive never really understood why people wont buy mining cards. A lot of the time there not 'ragged out'. They typically have a decent life undervolted and looked after. But saying that it would depend on the price

I bought an ex-mining R9 280X and the memory died within two months.

Never again.

The cores might be fine/unstressed but the memory on mining GPUs is "ragged out" to the max.

 

[RipGroove]

I bought an ex-mining R9 280X and the memory died within two months.

Never again.

The cores might be fine/unstressed but the memory on mining GPUs is "ragged out" to the max.

This.

 

[potatolord]

Ive never really understood why people wont buy mining cards. A lot of the time there not 'ragged out'. They typically have a decent life undervolted and looked after. But saying that it would depend on the price

I won't buy second hand for the same reason.

I would have to assume the card had been used for both gaming and mining, and that it may not have been undervolted.

Given likely warrantee issues, that's not a risk I could afford to take.

If it was half price I might reconsider, I guess.

 

[ASMB]

Ive never really understood why people wont buy mining cards. A lot of the time there not 'ragged out'. They typically have a decent life undervolted and looked after. But saying that it would depend on the price

I have always bought new when end of life for a good discount so never felt the need to buy second hand. There are always going to be those who feel the need to update asap even if it is by buying from scalpers, this generation it includes a greater than usual risk of buying ex-mining cards. I am quite happy to wait for current new prices to continue to decline rather than buy second hand from scalpers/miners

 

[Dirk Diggler]

One anecdote from someone who bought a second hand card that went on to fail does not constitute proof that mining leaves memory "ragged out". All GPUs are guaranteed for 3 years (with a couple of 2 year exceptions) and anything you can tweak in Afterburner is what the card is rated to handle. It's only when you get into BIOS mods or hardware mods (thermal pads) that warranty becomes a risk.

 

[Teabreaker]

I was really hoping for a drop today. Been 2 weeks now without a drop. Maybe next week is the one!

 

[FlyingFish]

My last GPU (EVGA 1080Ti FTW) was bought from a miner. I stuck the hybrid AIO cooler on it, gave it a mild overclock, and it ran flat out for gaming with zero issues.

 

[zx128k]

Never buy a mining card. Some cards the cooling on the memory is not the best. I have heard that the 3090 has no cooling on the memory chips on the back of the PCB. They can overheat when mining, causing the card to fail. I watched a mining video, were they added thermal pads to the memory chips at the back of the card.
Basically the perfessional cards have lower clocks, so the card will run stable doing mining like work loads. The clocks are dialed up for gaming as the work loads are not that hard. Also with mining, the videos show the gpu core frequency being reduced but the vram is overclocked. Heavy workloads, maximum overclocks to vram and no cares given to extra cooling needed. Leads to 100-110c vram reported on 3090 cards. Maximum safe temps 95c.

https://linustechtips.com/topic/1325035-sciences-proves-mining-kills-cards/
https://www.nvidia.com/en-us/geforc...-3090-fe-inadequate-vram-cooling/?topicPage=4
https://www.nvidia.com/en-us/geforc...-gddr6x-temps-while-gaming-a-concern/2993895/

 

[Dirk Diggler]

@zx128k utter pish. My 3080 hits the same temperature on Vram when gaming as it does when mining and 3060ti is currently sitting at 58C on the Vram. As for all the 3090s failing, we wouldn't need your insight to tell us that, because if it were the case then the forum would be full of posts regarding the failures and RMAs. As it stands, I haven't seen a single post from anyone regarding this.

 

[Calin Banc]

Ive never really understood why people wont buy mining cards. A lot of the time there not 'ragged out'. They typically have a decent life undervolted and looked after. But saying that it would depend on the price

I bought an ex-mining R9 280X and the memory died within two months.

Never again.

The cores might be fine/unstressed but the memory on mining GPUs is "ragged out" to the max.

I would also look at the fan since is running 24/7 compared to gaming card.

 

[FoxEye]

One anecdote from someone who bought a second hand card that went on to fail does not constitute proof that mining leaves memory "ragged out". All GPUs are guaranteed for 3 years (with a couple of 2 year exceptions) and anything you can tweak in Afterburner is what the card is rated to handle. It's only when you get into BIOS mods or hardware mods (thermal pads) that warranty becomes a risk.

Most warranties aren't transferable, so whether it's 3 years or 5 years, it doesn't matter if you're buying 2nd hand.

There's also the fact that AIBs have actually come out and said that mining is not the intended use of their cards, and they might not honour warranty if you've mined on them.

And then the anecdotes from people like me who bought ex-mining cards which failed...

All in all, nah, I won't be doing that again, thanks very much.

Surprise, surprise that the miners here - who might want to shift on their ex-mining cards at some point - would disagree vehemently.

 

[Dirk Diggler]

Most warranties aren't transferable, so whether it's 3 years or 5 years, it doesn't matter if you're buying 2nd hand.

The point about warranties is that the RMA option is available for that length of time, so how many posts have you seen here were warranty has been refused? In fact, are there even that many cases of warranty being claimed due to failure through mining regardless of whether it was refused?

There's also the fact that AIBs have actually come out and said that mining is not the intended use of their cards, and they might not honour warranty if you've mined on them

Again, is there a single case in point to share? Have the AIBs explained how this will be enforced?

Surprise, surprise that the miners here - who might want to shift on their ex-mining cards at some point - would disagree vehemently.

I'm not disagreeing on the basis that I want to sell you my cards, I'm disagreeing because you're spreading baseless nonsense with absolutely nothing to back it up.

 

[Sarge2255]

I bought an ex-mining R9 280X and the memory died within two months.

Never again.

The cores might be fine/unstressed but the memory on mining GPUs is "ragged out" to the max.

Miners Fry the memory chips on these cards, 24hours, 7 days a week. Knowing they plan to sell the card off to some innocent mug who believed they were only playing games with it.

They even go and swap out the thermal pads voiding any warranty.

I'll advise only buy a used card if the seller checks out to be genuine.

Surprise, surprise that the miners here - who might want to shift on their ex-mining cards at some point - would disagree vehemently.

 

[zx128k]

@zx128k utter pish. My 3080 hits the same temperature on Vram when gaming as it does when mining and 3060ti is currently sitting at 58C on the Vram. As for all the 3090s failing, we wouldn't need your insight to tell us that, because if it were the case then the forum would be full of posts regarding the failures and RMAs. As it stands, I haven't seen a single post from anyone regarding this.

Just the 3090 the VRAM on the back needs to be cooled but there are no heat pads between them and the back plate. VRAM on the back of the PCB will overheat when mining.

 

[Dirk Diggler]

Just the 3090 the VRAM on the back needs to be cooled but there are no heat pads between them and the back plate. VRAM on the back of the PCB will overheat when mining.

That's not correct. You need to watch some more YouTube vids.

 

[zx128k]

That's not correct. You need to watch some more YouTube vids.

Just watched a miner add the thermal pads on the back of the PCB to help cool the VRAM also see the sources I posted were they confirm vRAM is overheating. You would have found the youtube video of a miner installing the pads to cool the vram by 20c in the sources I provided.

In the video at 4:49 you can see the 110c vram temps while hashing before thermal pad replacement. 15:43 installs new thermal pads. Both front and back for the vRAM. 24:42 vRAM temps drop to 84c. This is 84-86C maximum temps with a massive vRAM overclock of +1056. 100-110 hash rate.

Possible to never run into this issue playing games. 3:42 onwards. Both 3080 and 3090 have the issue. Thermal pads are the problem. The ones on the card are not good enough.

In this video you can see no thermal pads on the vRAM at the back of the card. At 0:20 you can see the vRAM chips via a hole in the back plate. Even if there wa a thermal pad, its mostly connecting with thin air as most of the back plate has been removed. So you glue a massive heat sink onto the back of the card to fix the temp issue.

This is the other method of attaching heat sinks to the card to cool the vRAM when hashing.

If you look here there is nothing connecting the vRAM chips to the heatsink from the front. Often there are thermal pads to the metal plate below the heatsink and no thermal paste or thermal pad to the copper heatsink. 3:10 One of the fixes it thermal paste there. 3:58 no love or thermal pads for the vRAM on the reverse side of the PCB. Backplate looks plastic.

Why do all this? This is to stop the GPU from dying while mining. Never buy a mining gpu. They vRAM is overclocked to the edge and they have to mod the card to stop it from killing the card with the high temps.

 

[Joxeon]

Why do all this? This is to stop the GPU from dying while mining.

It's more likey so they can push up the hash rate as once VRAM hits 110% you lose about 15% income from mining as the card throttles back.

 

[zx128k]

It's more likey so they can push up the hash rate as once VRAM hits 110% you lose about 15% income from mining as the card throttles back.

The temps are high with gaming. Some people have cards that hit 110c ingames. So the high stock temps become worse as you overclock. Miners look to overclock the vRAM, so they have to go to exteme lenghts to cool the vRAM chips. The memory just burns and then once it wont overclock any more they sell the card to gamers.

Here Tjunction hits 104c in games https://www.igorslab.de/en/gddr6x-a...i-der-geforce-rtx-3080-fe-im-chip-gemessen-2/ 3080 FE.

In response to questions among colleagues, for example from the R&D departments, it was unanimously agreed that the maximum temperature Ttot before the start of a possible destruction of the chip should be 120 °C and that Tjunction should probably be set at 105 °C or for the GDDR6X even at 110 °C is specified as the maximum value.

On Mirco's website the op. temps should be 0c to +95c/ +105c. So getting a temp of 104c stock is at the edge. https://youtu.be/ph4_Sr8YNXI 1:31 The white paper warning is 2:36 The maximum rating is 105c.

Manufacture warns that, "Exposure to absolute maximum rating consditions for extended periods may affect reliability of the part for various device and package reasons." Basically if you keep below 95c then the vRAM will work as intended for a long time. So at 105c even the best vRAM chips starts not to function correctly and will degrade. Maximum temps are 95c. Some people mine all day every day at 110c. You need to water cool the cards for mining.

So ex-mining cards are worthless.

This is a video of a mining rig with heatsinks stuck on the back of every card and airflow from a fan. Note how not all cards have a hintsink. I guess you have to be careful overclocking the vRAM.

So hat happens when you water cool the cards. You hit vRAM temps of 94c when mining. 2:19 At 2:55 note how one card has vRAM temps of 104c. 104c is a really bad vRAM temp. Guess which card will die or you dont want to buy on ebay.

The datasheet for GDDR6x https://media-www.micron.com/-/medi...rief.pdf?rev=974854f170e4405a8b6dee54366eb308

Page 3 https://media-www.micron.com/-/medi...apps.pdf?rev=71bbcae453c14b29a736fbb246e1aedb Tj Max 100c

Minh Thai Le
1 week ago (edited)
Hello, is 24 7 mining at 96C memory temp ok on a RTX 3080? I underclocked the memory to -500 and the TMem went from 100 to 94. I would prefer to mine at 80 MH/s but to preserve my card than mining at 100 MH/s with a higher temperature, but Im not sure if 96C is fine for 24 7 mining over the next 6 months.Thank you, instant subscribe

 

[Joxeon]

The temps are high with gaming. Some people have cards that hit 110c ingames.

If that's the case then it's a design fault from Nvidia or whichever AIBs are effected. Should be able to RMA if the card it's hitting 110 in games and throttling performance.

Let's face it though Nvidia do have a lot of experience making Gpus and if they set the throttle point at 110C then they obviously believe the cards will last if VRAM is run high else they would have either set the throttle point lower or spent a bit more and used decent pads.