AMD 5870 Stuck at 100% fan, temp sensors are wrong - it thinks it is melting. :(

[set12]

I upgraded to 12.3 the other day and today after a long gaming session my fan became stuck on.

I checked GPU-Z and my card thinks it is at over 1000 degrees!

I've already power cycled and re-installed the drivers but the temp sensors are still stuck.

Any ideas?

 

[ThE_MaD_ShOt]

Have you tried a different driver?

 

[set12]

Yeah, I've tried several different ones.

I even went back all the way to 11.12, no luck.

 

[Kreij]

Is CCC reporting crazy temps too?

 

[set12]

Yeah.

And of course the manual fan control does nothing.

 

[Norton]

What card is this (brand, part #, etc...) a screenshot of the 1st page of GPUZ will also be helpful.

What are the rest of your system specs?

 

[Kreij]

I wonder if the board's BIOS flaked out. You could try reflashing the card.
Would seem either that or a multiple sensor failure in the card.

 

[FreedomEclipse]

Yeah.

And of course the manual fan control does nothing.

http://s17.postimage.org/6uglacmq7/ccc.jpg

because the enable manual fan control box isnt ticked??

 

[Kreij]

The fan speed makes sense if the card thinks its hotter than the Fukishima power plant during meltdown.

 

[LDNL]

The fan speed makes sense if the card thinks its hotter than the Fukishima power plant during meltdown.

Jokes aside flashing your bios might brick the card totally and should be only your last option.

 

[TRWOV]

would seem either that or a multiple sensor failure in the card.

+1

 

[Kreij]

It would seem to me to be either bad sensors or erroneous sensor reporting.

Maybe W1zz will come along and offer some insight.

 

[set12]

It's an Alienware i7-860 OC'd to 3.5ghz, it's been perfect until yesterday.

The card is a 100% generic ATI 5870:

Still working fine... but the noise is really annoying.

I've tried five different versions of the driver, uninstalling and running driver sweeper in between. I even unplugged my computer for an hour today to see is that helped.

 

[set12]

I fixed it...

I decided to run furmark for a while to "show" it what a real load was and what a real high temperature was.

So after furmark the sensors magically fixed themselves and the temps became normal.


Yay.

 

[Kenshai]

Try reseating the card. It's also odd that CCC and GPU-z are pulling completely different temperatures. I agree with Kreij thought on bad reporting or a flaky sensor.

 

[Kreij]

So you cooked your card with a long gaming session and got the sensors stuck, and then cooked it again with Furmark to unstick them. lol

I'll have to remember this fix in case it happens to me.
Thanks for posting the fix.

 

[set12]

Try reseating the card. It's also odd that CCC and GPU-z are pulling completely different temperatures. I agree with Kreij thought on bad reporting or a flaky sensor.

I'm guessing that the only reason they showed such different temperatures is that CCC only supports 4 digits for temp. Notice that the first "GPU Temp" in GPU-Z is negative, it must have flipped around because the software didn't support values that high.

 

[Kreij]

Interesting.

Assuming that the card polls the sensor at regular intervals and places the value in a register, if CCC is only reading the low byte of a >8 bit register into a signed byte, then it would cap at 127.
For W1zz's SENSOR_RECORD structure in his shared memory layout, it looks like he reads it into a double value, so GPU-Z should be able to go a lot higher than 1000 before it rolls over unless the register for the GPU temps is 10 bits (max 1024).

Of course, this is probably only interesting to me, so I'll shut up now.

 

[TheoneandonlyMrK]

might be worth giving the card a blowout ,clean up since it might keep it cooler

 

[Protagonist]

I fixed it...

I decided to run furmark for a while to "show" it what a real load was and what a real high temperature was.

So after furmark the sensors magically fixed themselves and the temps became normal.


Yay.

So you cooked your card with a long gaming session and got the sensors stuck, and then cooked it again with Furmark to unstick them. lol

I'll have to remember this fix in case it happens to me.
Thanks for posting the fix.

In Swahili language there is a saying that goes like this: Dawa ya moto ni moto (Translation: The remedy for fire is fire).

 

[FreedomEclipse]

In Swahili language there is a saying that goes like this: Dawa ya moto ni moto (Translation: The remedy for fire is fire).

otherwise known as "fight fire with fire" in the english language (but thank you for your translation!)

 

[95Viper]

Of course, this is probably only interesting to me, so I'll shut up now.

You mellowing out in your old age?
You made me .
You inject some fine points into the flow of conversation.

Back on topic:
I was going to say freeze it and let it come back to room temp and try it, but I like the set12's resolution.
New meaning to burn-in!
set12, you ought to let the people over at furmark know they saved your card.

 

[Kreij]

otherwise known as "fight fire with fire" in the english language (but thank you for your translation!)

Oh ... good one FE. I thought it refered to the old idiom, "If you sit your fat ass in a pan of hot grease, it's not as bad the second time." But yours is more appropriate.

You mellowing out in your old age?

Tatty asked me the same thing in a thread in the mod lounge not long ago.
So yeah, I'm probably getting more mellow these days.
This is probably good for you guys as I'm more lenient as a mod, but now it seems there is not the linear progression. I go from mellow Uncle Kreij to maniacal banstick weilder of doom.

On topic : I guess this is not too far removed from tossing a card in the oven when you get it so hot you create an open connection from loosening solder.

 

[FreedomEclipse]

"If you sit your fat ass in a pan of hot grease, it's not as bad the second time."

Doesnt actually make sense or apply unless you are a zombie. you are always going to feel the pain, but the smell of bacon might be somewhat comforting.

 

[Kreij]

@FE : I can't believe you made a logical response to that idoim that I obviously made up.
Anyway .... MMMmmmm. .... bacon.

I wonder if freezing the board would have worked to unstick the sensors? We'll never know not that he fixed it by cooking it again.