[ADVICE] Watchout for high temps of your GPUs

I found that modern heavy cards tend to bend down on standard ATX tower setups because of their increased over standard weight.

I have ASUS Poseidon 1080Ti - it's very heavy card (waterblock + air heatsink) but it applies to any modern GPU if it's water-block or not.

After some time it bends little loosing good contact with PCB/GPU.

What you see on screenshot (GPU temperature on first screenshot) are temps fluctuation 42-57°C while moving right side of card little up. Of course it's a problem with GPUs installed horizontally.

Remember air cooled cards are less efficient in cooling so you have to wait much longer to see temps drop after lifting the card. With waterblock it's almost instantaneous.

Do NOT lift PCB - always lift heatsink/waterblock and be delicate. I lifted plastic with led next to power plugs. So Check it and look after it, It can be even 15-20°C of difference! Fly safe!:pilotfly:

(Cooling is maxed for testing purposes and this max2228 was instantaneous benchmark crash after my curve mistake)

EVEN BETTER:

I have 2 Poseidons, 980GTX and 1080Ti, and luckily do not suffer from this.

Do I understand this correctly ? You assume that through bending due to gravity the cooler plate separates from the surface to be cooled ( GPU surface ) and thus the temp rises significantly ?

If that is correct, I think the card hasnt been assembled correctly. Those cards are firm as a brick, heavy as a tank and the backplate looks very promising to me in terms of "I dont bend a mm".

If you dare, take the card apart and make sure it had been correctly assembled.

OR, do a RMA. Just did that myself, its gonna be a few weeks door-2-door.

and btw...WOW...yours overclocks like there is no tomorrow !!!

2088 ? is that stable, along with 12.5GHz on the Micron GDDR5x ? NICE !

I think my RAM could clock higher but MSI AFterburner wont let me add more than "+1000", hmmm.

My GPU is pretty much tilt at 2025, any higher needs too much voltage and heat.

This card is an amazing undervolter, 1820MHz at 880mV, RAM at 12G.

I have 2 Poseidons, 980GTX and 1080Ti, and luckily do not suffer from this.

 

Do I understand this correctly ? You assume that through bending due to gravity the cooler plate separates from the surface to be cooled ( GPU surface ) and thus the temp rises significantly ?

 

If that is correct, I think the card hasnt been assembled correctly. Those cards are firm as a brick, heavy as a tank and the backplate looks very promising to me in terms of "I dont bend a mm".

 

If you dare, take the card apart and make sure it had been correctly assembled.

OR, do a RMA. Just did that myself, its gonna be a few weeks door-2-door.

I'm not assuming, i just proved it and made screenshot. when i started using a card it wasn't a problem but over time temps were worse. I started investigating my loop, radiator everything. And this was eyeopening discovery. When i meseured if my card is bended i noticeced it was. I noticed other cards bending like that so i wouldn't assume your's isn't suffering.

I used a support, a plastic stick and temps dropped from 52-57° to 40's.

and btw...WOW...yours overclocks like there is no tomorrow !!!

 

2088 ? is that stable, along with 12.5GHz on the Micron GDDR5x ? NICE !

 

I think my RAM could clock higher but MSI AFterburner wont let me add more than "+1000", hmmm.

 

My GPU is pretty much tilt at 2025, any higher needs too much voltage and heat.

 

This card is an amazing undervolter, 1820MHz at 880mV, RAM at 12G.

After lifting card i easily OCed it to 2126. It hanged at 2088 before.

2126 is perfectly stable at 43°C on two hour benchmark loop. What temps do you have at 2025?

OK..taking the door off now while I type and mine....

OHHHHH MFG you are right !!!!!!!!!!!!!!!!!!!!!!!!!!!! immedeate drop DOWN...WOW WOW WOW

BIG REP inbound !

more to investigate, maybe this should be posted on Asus forum !?

OK..taking the door off now while I type and mine....

 

 

OHHHHH MFG you are right !!!!!!!!!!!!!!!!!!!!!!!!!!!! immedeate drop DOWN...WOW WOW WOW

 

BIG REP inbound !

 

more to investigate, maybe this should be poisted on Asus forum !?

you have no idea how happy I am it worked for you:thumbup: I tried posting on PCMASTERRACE here but they downwoted it like i was some noob with magic trics ehh... salty reddit...

from 51°C steady in mining ETH at 1847/1200 @ 943mV to a STEADY FORTYONE °C

TEN degrees less !!!! TEN !!

THANK YOU..will post a screenshot with my Balsa stick...LoL but it works

cant rep you now, think I did it before already ? Did I ??

Trying to get some attention on HardOCP for this, Kyle might have still have the reviewed 1080Ti Poseidon....

10°C is a BIG thing. I am still more than puzzled what to do with this finding O_o

https://hardforum.com/threads/asus-poseidon-1080ti-coldplate-issue-multiple-users.1954766/#post-1043492517

I did sent it to gamersnexus and JayzTwoCents and making some post on Asus/ OC forums

My QuickFix:smilewink:

My QuickFix:smilewink:

:D

btw, what are you using to clock those Microns to 12500MHz ?

I could really use 500MHz more for ETH :)

Asus GPUTweak ??

btw, what are you using to clock those Microns to 12500MHz ?

 

I could really use 500MHz more for ETH :)

 

Asus GPUTweak ??

Yes:thumbup: DCS World likes bandwidth in VR so i made it 12650

Hmmm, I have the EVGA hybrid. Maybe I should start testing. Thanks for the info!

Hmmm, I have the EVGA hybrid. Maybe I should start testing. Thanks for the info!

Do NOT lift PCB - always lift heatsink/waterblock and be delicate. I lifted plastic with led next to power plugs.

This is extremely interesting. You should probably contact ASUS about it and whomever else it might concern, since it appears to be poor design.

This is extremely interesting. You should probably contact ASUS about it and whomever else it might concern, since it appears to be poor design.

It's not only ASUS. It's ATX standard fault. My last MSI gamers X 1070 had the same problem. When i uninstalled it after a 1,5 year use it was heavily bended. I didn't noticed it before i took it off the case.

Modern heavy cards tend to bend down on standard ATX tower setups because of their increased over standard weight.

Hmmm, I have the EVGA hybrid. Maybe I should start testing. Thanks for the info!

Your heatsink is modular and much smaller. It's not like poseidon's big waterblock + heatsink. I don't think it will help but you can try

down to 37-38°C now, wow.

it took a while to cool the card down and find the new equilibrium

Did some quick OC with Asus GPU TweakII

2101 GPU and 12528 VRAM :megalol:thats 5% more with the Balsa Stick:megalol:

It not only runs cooler as you say, it also overclocks WAY HIGHER than ever before amd quite frankly, I havent even tested any higher values yet. i dialed those in and let it run for an hour, it runs. My MH/sec went a up a bit in ETH, VRAM speed is everything

Did some quick OC with Asus GPU TweakII

 

2101 GPU and 12528 VRAM :megalol:thats 5% more with the Balsa Stick:megalol:

 

It not only runs cooler as you say, it also overclocks WAY HIGHER than ever before amd quite frankly, I havent even tested any higher values yet. i dialed those in and let it run for an hour, it runs. My MH/sec went a up a bit in ETH, VRAM speed is everything

:thumbup: My stable score for now is GPU@2126 MEM@12650 @1075v

As of September 2019, supporting the card with a stick to prevent sagging is not enough anymore. My card's temp is 45-47°C in idle, it used to be 49°C in DCS, I got 65-67°C there now.

According to a guy on ROG forum the springs that press the GPU cooler down onto the GPU itself are too soft and over time, this leads to a separation between the two, coupled with a thick thermal padding it has this bad effect at a certain age of the card.

What helps is either mounting a full cover waterblock for 150-200€ for an Asus 1080ti-Strix ( same PCB )..or...put two washers onto the springs so they tighten up all the way and thus press the cooler firmly back onto the chip.

I am doing the washer method asap and report back...

next time,,,FE and after market block !

As of September 2019, supporting the card with a stick to prevent sagging is not enough anymore. My card's temp is 45-47°C in idle, it used to be 49°C in DCS, I got 65-67°C there now.

 

According to a guy on ROG forum the springs that press the GPU cooler down onto the GPU itself are too soft and over time, this leads to a separation between the two, coupled with a thick thermal padding it has this bad effect at a certain age of the card.

What helps is either mounting a full cover waterblock for 150-200€ for an Asus 1080ti-Strix ( same PCB )..or...put two washers onto the springs so they tighten up all the way and thus press the cooler firmly back onto the chip.

 

I am doing the washer method asap and report back...

 

next time,,,FE and after market block !

No problems here after all this time - no temp change:thumbup::music_whistling:

Maybe someone overdone it... Good luck!

Thank you for having shared this.

Do NOT lift PCB - always lift heatsink/waterblock and be delicate. I lifted plastic with led next to power plugs.

I've googled gpu brace and I've noticed that many braces support the plastic of the heatsink. However, I've attached an image in which you can notice that my OEM graphics card was supported on the PCB.

Maybe I'll do that the next time I open my PC to remove the dust (I'll do that more often than I used to with the old GPU, don't worry :) ) and I think I'll implement the following procedure:

1) place a bubble level onto a surface of the case which is reasonably supposed to be parallel to the GPU,

2) add shims under the PC until the bubble is centered,

3) carefully place the level onto the PCB of the GPU,

4) create a brace which makes the bubble centered and anyhow stop if I feel that too much force is being applied to the GPU.