Gpu overheating to over 90C in seconds

[eidairaman1]

Problem is you are using furmark, stop using it.

You are abusing the part.

 

[Fouquin]

it should have proper pads on the following locations

The image you shared is of a reference R9 390, not a 6950. (Though why it shows Crossfire connectors befuddles me...) It's not that big of a deal but the 6950 is significantly simpler in design as it has 1/3 the VRAM components.

Also @diegander it's not impossible that the vapor chamber has failed. If the chamber was compromised the liquid inside would eventually escape. If that chamber is empty you essentially have a copper sleeve with a big pocket of air separating the card and the heatsink assembly. There's not really much to be done but get another cooler for the card at that point.

Another possibility is that the thermal monitoring hardware baked in has suffered a malfunction. Have you put your hand up to the card to feel if it is actually getting as hot as it's reporting? I've personally had cards report they are well into the hundreds of degrees, but feel rather lukewarm to the touch after hovering over the back of the card to check the temperature radiating from it.

 

[Derek12]

Problem is you are using furmark, stop using it.

You are abusing the part.

+1
It streeses the GPU way beyond realistic usage

 

[SnakeDoctor]

You overclocking your gpu? Seem highly possible
Reset to default ,turn off -Apply overclock on startup

 

[trog100]

i use furmark.. it heats up my gpu which is why i use it.. but not quickly.. i can watch the gpu temps go slowly up on the furmark graph.. it takes a good three or four minutes to get up to max tempts and then the graph levels off..

a cpu temp will go up and down quickly a gpu should take a while to both go up and go down..

the reason for this is because the gpu heats up the entire card the cpu just heats up itself..

if your gpu temps are going up and down quickly it suggests a problem with the interface between chip and the heat spreader.. basically your cooling system isnt working properly..

trog

ps.. furmark is one of the best tools there is for checking gpu tempts.. you can see exactly what the temps are doing and see how much the card is throttling down plus other information.. all on the one nice screen.. most cards will throttle down once 80 C is reached.. furmark shows all this happening perfectly..

 

[Ruru]

I'd say that Furmark is only good for testing the ultimate maximum for GPU temps. I rather play games to see my temps, since Furmark is faaaaar away from anything real use.

 

[DeathtoGnomes]

You have to babysit Furmark, it can physically damage your card too.

 

[basco]

gputool is best

 

[trog100]

You have to babysit Furmark, it can physically damage your card too.

maybe in the far distant past.. modern cards self protect in the bios.. its almost impossible for a user to damage them they just throttle down when running furmark..

it dosnt represent normal usage but nether does running prime95.. but if there is a cooling problem furmark would show it up..

whats unusual (not normal) is the OP comment that his gpu hits 90 C in seconds.. this should not happen it should take a fair while longer than this for the gpu to hits its max temps..

ether way the card will throttle right down to stay as close to 80 C or so as it can.. this is all done in the bios.. the boost speed would show very low and gpu utilization would also be very low..

furmark would instantly show all this up..

to me the only way any gpu will hit 90 C in seconds is if the gpu chip isnt connected to the cooling array properly or the heat pipes in the array have failed..

the bottom line is.. a modern card is pretty much idiot proof.. it has to be.. the makers know that some users will try and overclock the sh-t out of them.. safe guards are all built in..

trog

 

[P4-630]

any gpu will hit 90 C in seconds is if the gpu chip isnt connected to the cooling array properly or the heat pipes in the array have failed

^This

 

[RejZoR]

For testing, use 3DMark Stress Test or run Valley, Heaven or Superposition in loops for a while. I never use Furmark because it's entirely unrealistic and doesn't help evaluate stability if it goes beyond anything any game will ever do to a GPU. Killing Floor 2 game also seems to be a good indicator because it tortures GPU quite a bit with hordes of enemies.

 

[jboydgolfer]

The image you shared is of a reference R9 390, not a 6950.

My point was still just as valid, the placing was almost identical(&my point was to give an example or guide to what placing would be...there are several differing PCB layouts for the 6950.OP was asked if it was ref design, he didnt respond ,i did what i could.) . Does this meet Your approval?

 

[DeathtoGnomes]

the bottom line is.. a modern card is pretty much idiot proof.. it has to be.. the makers know that some users will try and overclock the sh-t out of them.. safe guards are all built in..

that may be true, but Furmark users are not.

 

[trog100]

^This

yep..

i will post a pic of furmark running on my system.. it does take several minutes for my pair of room heaters to fully heat up.. and this is running the "evil" furmark.. he he

trog

 

[jboydgolfer]

yep..

i will post a pic of furmark running on my system.. it does take several minutes for my pair of room heaters to fully heat up.. and this is running the "evil" furmark.. he he

trog

 

[trog100]

the furmark pic i promised.. the time period isnt marked but the graph shows about three minutes of running.. note the slow and gradual increase in temps before the max is reached and things level off..

if max temps (90 C) are being reached in seconds something is fundamentally wrong with the OPs cooling system..

mine would never go over about 82 C no matter what.. the cards boost would just drop off and so would the gpu utilization (power usage) to stop them going any higher..

trog

 

[DRDNA]

maybe in the far distant past.. modern cards self protect in the bios.. its almost impossible for a user to damage them they just throttle down when running furmark..
trog

So are you saying that Furmark will no longer deep fry the VRM's on a AMD card? I'm just wondering and wondering when AMD fixed that?

 

[trog100]

So are you saying that Furmark will no longer deep fry the VRM's on a AMD card? I'm just wondering and wondering when AMD fixed that?

running furmark should not deep fry anything it simply lowers the cards power usage to stay withing reasonable limits.. but i must admit to not owning an amd card.. if amd havnt fixed such things they would be big time at fault.. i think its reasonable to assume they have..

trog

 

[DRDNA]

running furmark should not deep fry anything it simply lowers the cards power usage to stay withing reasonable limits.. but i must admit to not owning an amd card.. if amd havnt fixed such things they would be big time at fault.. i think its reasonable to assume they have..

trog

Yeah maybe but I don't assume anything when it comes to AMD....I use to but not anymore.

 

[DeathtoGnomes]

running furmark should not deep fry anything it simply lowers the cards power usage to stay withing reasonable limits.. but i must admit to not owning an amd card.. if amd havnt fixed such things they would be big time at fault.. i think its reasonable to assume they have..

trog

well, there ya go, thats why it works so well for you. Carry on.

 

[eidairaman1]

Since its a 6000 series card, yes it will screw it up.

 

[trog100]

ether way furmark or not i recon the card is faulty.. i first used furmark over 12 years ago (maybe longer) and way back then i did use amd cards..

power usage and control is basically what "boost" is about.. if things (temps and power usage) aint within tolerable limits you get less or no boost..

trog

 

[Bones]

I've ran Furmark before and yes it will make a GPU heat up.

Many already know Furmark is really a burn-in proggy rather than an actual bench, largely used for stress testing and temp checking under max load. You can bench with it of course but in reality it's just a tool as described for seeing if your card is stable and not getting too hot at whatever settings you try it with.
If it can run where you have it in Furmark then the settings and card itself are fine.

If and when I run Furmark that how I use it.
Even at stock settings it can make a GPU work hard and if the card is OC'ed you'd better baby-sit the system while it's running or at least to the point you see temps stabilize within an acceptable range.

 

[Fouquin]

My point was still just as valid, the placing was almost identical(&my point was to give an example or guide to what placing would be...there are several differing PCB layouts for the 6950.OP was asked if it was ref design, he didnt respond ,i did what i could.) . Does this meet Your approval?

Chill dude, I clarified because a lot of people don't know what the card under the shroud should look like. OP could have become confused if he was comparing direct to the original image, wondering why his board looked so different. It's not entirely uncommon for people who are past their comfort zone working on their own hardware to question the little things so they don't end up breaking it.

Not saying OP specifically falls under that demographic, but you have to take everything into consideration when giving people advice to tear their components apart.

 

[diegander]

i sent my overheating gpu to a friend to check and he told my the gpu is working fine, which is wierd , i pluged my monitor to the motherboard then i noticed something the computer still crash , and i also notice if i copy something to a usb flash it is is copying for examble like copy hold hold hold copy hold hold hold copy.... which is also wierd

is the problem that caused my this trouble is the motherboard or power supply or the cpu ??? pls help