Sciences proves mining kills cards

[Strategic]

This template is a guide for the Tech News Posting Guidelines. Please read the guidelines in the pinned topic before posting, otherwise your post may be removed without warning. If you prefer, you can clear the editor to use your own layout by clicking the trash can icon above, but make sure you incorporate all of these sections and follow the Posting Guidelines.

 

Remove all of the italicised text before posting.

 

Summary

Higher VRAM temperatures which is very normal during mining (seperate from gpu temperature) causes earlier memory degradation within 3 months instead of years. 

 

Quotes

Quote

 

 

My thoughts

There is a common misconception that gpu miners do not kill gpu's faster than people that use the gpu for gaming, because the common argumentation is that they lower the gpu powerdraw but they generally overclock their memory. The real problem is vram temperature being high that will eventually leave you with a gpu that dies faster than it should. Meaning that you shouldn't buy GPU from miners or atleast with an extra discount.

 

 Sources

https://www.cs.toronto.edu/~bianca/papers/sigmetrics09.pdf

https://www.pdl.cmu.edu/PDL-FTP/NVM/dram-retention_isca13.pdf

https://repository.tudelft.nl/islandora/object/uuid:e36c2de7-a8d3-4dfa-9da1-ac5b7e18614b 

https://ieeexplore.ieee.org/document/7056044

[Kanna]

[IkeaGnome]

@MoonzyThought you'd like this.

 

@Strategicbad miners overclock. Good ones don't(I mean as in people good at mining and bad at mining). If we're being honest here, my 2070 saw higher VRM temperatures gaming than mining..

[Strategic]

1 minute ago, IkeaGnome said:

@MoonzyThought you'd like this.

 

@Strategicbad miners overclock. Good ones don't(I mean as in people good at mining and bad at mining). If we're being honest here, my 2070 saw higher VRM temperatures gaming than mining..

here have a common overclocked that has 51 degrees on GPU but 100 degrees on VRAM. This is considered your average good miner, other miners hit higher with their 3090. (I'm in partalert discord, there is a sub-channel talking about mining that I've been following how people optimize their gpu's)

 

[WhitetailAni]

16 minutes ago, Strategic said:

This template is a guide for the Tech News Posting Guidelines. Please read the guidelines in the pinned topic before posting, otherwise your post may be removed without warning. If you prefer, you can clear the editor to use your own layout by clicking the trash can icon above, but make sure you incorporate all of these sections and follow the Posting Guidelines.

 

Remove all of the italicised text before posting.

LOL

But yes 
is cryptocurrency related so @Moonzy

Just now, Strategic said:

here have a common overclocked that has 51 degrees on GPU but 100 degrees on VRAM. This is considered your average good miner, other miners hit higher with their 3090.

-snip-

Yes, GDDR6X gets hot as Moonzy can attest to.
VERY hot.

[IkeaGnome]

5 minutes ago, Strategic said:

here have a common overclocked that has 51 degrees on GPU but 100 degrees on VRAM. This is considered your average good miner, other miners hit higher with their 3090. (I'm in partalert discord, there is a sub-channel talking about mining that I've been following how people optimize their gpu's)

And this is limited to only mining?

https://www.nvidia.com/en-us/geforce/forums/geforce-graphics-cards/5/432491/rtx-3090-fe-inadequate-vram-cooling/?topicPage=4

 

[Strategic]

Just now, IkeaGnome said:

And this is limited to only mining?

https://www.nvidia.com/en-us/geforce/forums/geforce-graphics-cards/5/432491/rtx-3090-fe-inadequate-vram-cooling/?topicPage=4

 

There is a difference between a gamer playing games 5 hours a day, and a miner that runs their gpu 24/7

[Moonzy]

16 minutes ago, FakeKGB said:

GDDR6X gets hot as Moonzy can attest to.

Yeap, for 3090, they hit 110c (throttling temp) easily

For 3080 it depends on GPU design, I only have one particular model so not much comments, but mine hovers around 100c under mining load

Productive tasks share a similar story, or so I heard

 

For gaming load, I've heard the VRAM temps gets up to 100c+ on 3090, and low 90s for 3080, water cooling a 3080 will bring it down to 50-60 (from what I saw in a thread) if your block adequately cools the chips

 

But in the end I don't really worry too much about G6x, because they have temp sensors and they would throttle themselves to prevent damage

 

But GPU that uses other VRAM are what worries me, because there's no temp sensor to detect and throttle from what I understand

So I (and many miners that I know of) crank fan speed to near 100%, because we can't tell how hot the temp is, so just crank the fans and that should be the best for it

Normal users don't know VRAM temps too and often lowers fan speed for silence, so I can't comment on how their cards are well maintained

 

As for VRAM degradation, yea I observed some of my cards OC capability dropped over the period of owning them, from +1400 to +1200 for example, but these clock speeds are never stable for gaming loads (almost instant crash, stable around +800-900), so I'm not sure if it's just it being unstable to begin with, or it's degradation, or burn in period, because they only dropped after first few days, then stabilizes, and it's still way above stock speeds

 

All in all, there's quite a lot of factor to consider, whether a card that's used for mining is any worse off than cards that are used "normally", because there's also thermal cycling and what not to consider

 

Buying used have far more risk than just miner cards, imo, that's why I myself avoid used.

Unless I know how the card was treated before, ie from a good friend

[Strategic]

6 minutes ago, Moonzy said:

Yeap, for 3090, they hit 110c (throttling temp) easily

For 3080 it depends on GPU design, I only have one particular model so not much comments, but mine hovers around 100c under mining load

Productive tasks share a similar story, or so I heard

 

For gaming load, I've heard the VRAM temps gets up to 100c on 3090, and low 90s for 3080, water cooling a 3080 will bring it down to 50-60 (from what I saw in a thread) if your block adequately cools the chips

 

But in the end I don't really worry too much about G6x, because they have temp sensors and they would throttle themselves to prevent damage

 

But GPU that uses other VRAM are what worries me, because there's no temp sensor to detect and throttle from what I understand

So I (and many miners that I know of) crank fan speed to near 100%, because we can't tell how hot the temp is, so just crank the fans and that should be the best for it

Normal users don't know VRAM temps too and often lowers fan speed for silence, so I can't comment on how their cards are well maintained

 

As for VRAM degradation, yea I observed some of my cards OC capability dropped over the period of owning them, from +1400 to +1200 for example, but these clock speeds are never stable for gaming loads (almost instant crash, stable around +800-900), so I'm not sure if it's just it being unstable to begin with, or it's degradation, or burn in period, because they only dropped after first few days, then stabilizes, and it's still way above stock speeds

 

All in all, there's quite a lot of factor to consider, whether a card that's used for mining is any worse off than cards that are used "normally", because there's also thermal cycling and what not to consider

 

Buying used have far more risk than just miner cards, imo, that's why I myself avoid used.

Well the 3000 series are new so we won't know how fast they can die with mining. As for reference my friend that mined with 1080 ti, vega 56 and vega 64 liquid cooled that lost about 20% of their gpu just after the 2 year mark (he mined from 2018 till 2020). Luckily some died within warranty I think he had some cards with 3 year warranty, but he lost in total 5 gpu's of the 25 gpu's (3 vega 56 of the 14 and 2 1080 ti's of the 5).

[Moonzy]

1 minute ago, Strategic said:

Well the 3000 series are new so we won't know how fast they can die with mining. As for reference my friend that mined with 1080 ti, vega 56 and vega 64 liquid cooled that lost about 20% of their gpu just after the 2 year mark (he mined from 2018 till 2020). Luckily some died within warranty I think he had some cards with 3 year warranty, but he lost in total 5 gpu's of the 25 gpu's (3 vega 56 of the 14 and 2 1080 ti's of the 5).

Can I know what PSU he's using?

Was it a mining PSU? I myself use corsair AX1600i and HX1000, so I know the ripples are well within spec and won't damage things

 

But yea many miners use cheaper PSU, and that, perhaps, kills GPU more

 

I mined from 2017 till now, so far only one fan died, and it's my old old gaming card and only one month into mining, so idk what to make of that

Though I only had 2-4 cards until January this year when I ramped up to 20

 

Well I'll be sure to know if 30 series GPU are dying left and right by mining, because I own so many of them now, thanks for the heads up

[Helpful Tech Witch]

Yay.

23 minutes ago, IkeaGnome said:

bad miners overclock

the core yess. But good miners overclock the vram as it has a huge perfimpact.

 

 

In reality, this title is wrong. It should be "Science proves heat kills vram".  Any other vram intensive task on a card with little vram cooling (or none at all). I believe moonzy works pretty hard to not have the mem able to boil water.

[Strategic]

1 minute ago, HelpfulTechWizard said:

I believe moonzy works pretty hard to not have the mem able to boil water.

That actually sounds very funny.

[Jtalk4456]

47 minutes ago, Strategic said:

Remove all of the italicised text before posting.

hmmm....

[Moonzy]

7 minutes ago, HelpfulTechWizard said:

. I believe moonzy works pretty hard to not have the mem able to boil water

You'll have to believe otherwise because I can only do so much about badly designed GPU, particularly the 3090 gigabyte, it hits 110c and throttles even after strapping a wraith stealth on the backside pcb, where the vrams are overheating

 

Kinda stupid design, 60W of heat on a flat backplate, I haven't test it under gaming load yet because I truly cbf, probably hits 110c on stock settings as well

 

The MSI card hits 110c but at least it doesn't throttle and occasionally dips to 108c, probably the heatpipe on the backplate doing its job

[bcredeur97]

I mean yes using the gpu will consume more of it's life than not using it. 

I still think mining on a card for years, then giving it a little love, cleaning it, replacing the thermal paste and pads, etc. will leave you with a perfectly usable GPU for gaming 99% of the time.

That said.. i could see the particular issue with the memory getting hot on these new cards just being an issue in general. But like the 10xx series? pfft just clean the card up and maybe replace thermal materials and you're good.

Computer parts generally last a while.

[Strategic]

8 minutes ago, Moonzy said:

You'll have to believe otherwise because I can only do so much about badly designed GPU, particularly the 3090 gigabyte, it hits 110c and throttles even after strapping a wraith stealth on the backside pcb, where the vrams are overheating

 

Kinda stupid design, 60W of heat on a flat backplate, I haven't test it under gaming load yet because I truly cbf, probably hits 110c on stock settings as well

 

The MSI card hits 110c but at least it doesn't throttle and occasionally dips to 108c, probably the heatpipe on the backplate doing its job

How will you manage the summer actually? Or do you use A/C?

 

As your question about PSU I actually can't find it in the discord chat with him. He had a gold psu from corsair, and 2 platinum psu's that server rigs use. You know that ugly, long loud thing (he also had more PSU's but didn't go in detail with me). He had a mining rig at the company of his brother that agreed to let him mine (he had a empty room to mine, and that room would hit around 40 degrees in the summer it). 

 

 

[Frostbobomb]

That's unfortunate. Sometimes we do just get shit hardware. However, not all card are created = 
Here you can see my 3080 doing its normal 100MH/s and the Mem temps are super cool.
So long as you didn't get stuck with a bad card and you treat them right, mining is not going to kill your hardware.

[SansVarnic]

-= Moved to Graphics Cards =-

Not a tech News item.

 

OP: This discussion has been around for some time; it is well known that high usage of electronics will degrade it over time. As to what degree is under argument.

Also you didnt follow the instructions of the template when you posted, please take the time and clean you r post by using the edit button.

"Remove all of the italicised text before posting."

[Moonzy]

2 minutes ago, bcredeur97 said:

That said.. i could see the particular issue with the memory getting hot on these new cards just being an issue in general. But like the 10xx series? pfft just clean the card up and maybe replace thermal materials and you're good.

Iirc the EVGA 1080ti first batch had VRAM overheating issues, that's why they came up with ICX, which was loading up temp sensors all over the board

 

Other than that, I haven't heard much of anything about VRAM killing cards on a large scale

 

1 minute ago, Strategic said:

How will you manage the summer actually? Or do you use A/C?

Tropical, and I manage the heat by placing them at a location where no one will go and keep a window open, it's been doing ok

 

2 minutes ago, Strategic said:

As your question about PSU I actually can't find it in the discord chat with him. He had a gold psu from corsair, and 2 platinum psu's that server rigs use. You know that ugly, long loud thing (he also had more PSU's but didn't go in detail with me). He had a mining rig at the company of his brother that agreed to let him mine. 

I see, well I can't comment much unless I know more details on exactly how his cards have failed, could be overheated VRAM if he didn't crank fan speed up and leave them at stock, because GPU core temp is low for eth mining the VRAM is hit hard

 

Could be voltage ripples from bad PSU, could be anything really

 

But in the end, I guess mining does degrade a GPU more than if it's sat on a shelf, comparitively

[Moonzy]

3 minutes ago, Frostbobomb said:

That's unfortunate. Sometimes we do just get shit hardware. However, not all card are created = 
Here you can see my 3080 doing its normal 100MH/s and the Mem temps are super cool.
So long as you didn't get stuck with a bad card and you treat them right, mining is not going to kill your hardware.

EVGA icx measures package temp I think, which is the external temp of the VRAM casing(?)

 

Most G6x monitors Tjunction, which is the hottest part of a die and is often higher than other temp measuring methods, from what I understand

[Strategic]

3 minutes ago, SansVarnic said:

-= Moved to Graphics Cards =-

Not a tech News item.

 

OP: This discussion has been around for some time; it is well known that high usage of electronics will degrade it over time. As to what degree is under argument.

Also you didnt follow the instructions of the template when you posted, please take the time and clean you r post by using the edit button.

"Remove all of the italicised text before posting."

"As to what degree is under argument." Not really as shown in youtube video referencing the research, at 350  (1157,407 days) kelvin at already degrades significantly faster than than at 300 kelvin as 400 kelvin compared to 350 kelving ( 100000 seconds = 116 days)

 

[leadeater]

1 hour ago, Strategic said:

There is a common misconception that gpu miners do not kill gpu's faster than people that use the gpu for gaming, because the common argumentation is that they lower the gpu powerdraw but they generally overclock their memory

No there isn't a common misconception, the common misconception is that cards will die in 3 months like this source claims which is SOOOOO clearly false. Like you don't even need to do a scientific study to know this is false. Mining literally wouldn't exist if the hardware was dying within months lol.

 

Increase temperatures in any silicon device, be it the GPU die on the memory chips will decrease the lifespan but it's significantly longer than the useful life of the card. Memory chips have a lifetime rating at TjMax and it's a very, very long time. So unless they are claiming Samsung/Micron etc are lying then the testing itself has been flawed and compromised the data and therefore the the conclusion from it.

 

Seems like yet another hit piece on mining to drum up anti mining sentiment. This is unnecessary, people that dislike it don't need any more convincing to dislike it and making stories about it that are false, fake, or otherwise not correct doing nothing but to discredit the discussions around it.

 

Like I've said in topics before, server GPU which use the exact same GPU dies and memory chips that operate in just as if more more extreme temperature conditions to mining GPUs (certainly more so than your causal gamer miner using 1 or 2 GPUs with actually good coolers) are alive after 10 years.

[Strategic]

1 minute ago, Strategic said:

"As to what degree is under argument." 

Not really as shown in youtube video referencing the research, at 350  (1157,407 days) kelvin at already degrades significantly faster than than at 300 kelvin as 400 kelvin compared to 350 kelvin ( 10000000 seconds = 116 days)

 

[Frostbobomb]

2 minutes ago, Moonzy said:

EVGA icx measures package temp I think, which is the external temp of the VRAM casing(?)

 

Most G6x monitors Tjunction, which is the hottest part of a die and is often higher than other temp measuring methods, from what I understand

Hmmm, anyway for me to check the Tjunction? Or is that not possible on EVGA?

[leadeater]

3 minutes ago, Strategic said:

"As to what degree is under argument." Not really as shown in youtube video referencing the research, at 350  (1157,407 days) kelvin at already degrades significantly faster than than at 300 kelvin as 400 kelvin compared to 350 kelving ( 100000 seconds = 116 days)

Well I'm afraid that their testing and sample data is likely filled with bad data as the literal hundreds of thousands, no probably millions, of server GPUs proves this research data wrong in real life.