Graphic Cards HD6950 Overheating on Two monitors.. that too without load

max_demon

@maxdemon
Adept
Hi, since few weeks my MSI HD 6950 Twin Frost edition is giving overheating problems.. I have not been playing any games on it recently. the problem happnes only when I'm on Multi Screen card activity is 5-10% but the temps reach upto 105 degree celsius and computer shuts down.

even on single monitor the temps are hovering near 60-70 degrees (this has not been case few months back..)
please advise on what to do.. These temperatures are on windows only.. that too without any Load. even uptime is about 9-10 hours max.

The Fans don't seem to be running.. although they do some action if I give it a little push. what to do? should I replace the fans? from where could I find these type of Fans?
 
Last edited:
RE-TIM the core. And clean out the shroud. What speeds is the fan spinning at. CCC would tell that.
 
The Fans don't seem to be running.. although they do some action if I give it a little push. what to do? should I replace the fans? from where could I find these type of Fans?

Probably the fans are about to die. I had an HIS 4870 1 GB GPU with a fan that died a premature death (within 4-5 months IIRC). Removed the fan shroud and with some zip-ties, attached a couple of 80 x 25 mm medium-speed fans which worked better than the stock fan. The fans were powered directly from the power supply and not the motherboard fan headers. If you have one 120 mm fan, that would do as well.
 
I re-applied the Thermal paste after cleaning the previous one and cleaned off the dust. the temperatures are about 80 degrees after a min of FurMark and idle at 60 degrees on both monitors. CCC shows fan speed in % and I had set it to 100% it shows that even if the fans are not in motion. I was thinking of replacing the fan as well. I had checked the type of fan it read some type of Brushless fan. they are smaller than 80mm fans. I can get a 120mm fan but not sure how to fix it over them as there doesn't seem to be any place to mount it.

My card is : http://www.msi.com/product/vga/R6950-Twin-Frozr-III-Power-Edition-OC.html
 
This might not just be the problem with your cooling. Although by your description your GPU fan seem to be culprit, but there is another issue with dual monitor setup that is causing these cards to hit high temps. I've had a same issue and this seems to be the problem with either drivers or the card's BIOS/Firmware. If you monitor you'll notice that when two displays are connected the GPU core voltage spikes up. Disable a display and the core voltage will drop down. I've tried multiple combinations and even going back to old drivers, but it seems once started the issue persists (even with complete clean install and new install of windows). Also, I suspect that not just drivers, but Windows as well might be responsible for this, as I noticed the same behavior with nVidia card (670GTX) as well as different ATi cards (HD6950, HD7970 and HD7850).

My current GPU's core voltage (HD7850) stays at 0.900v idle with two displays active, but drops down to 0.825v when only single display is active. Earlier, it used to idle at 0.973v, but after tweaking some settings in Catalyst display options, I was able to get it down to 0.900v at idle.

Unfortunately, I'm not sure which option worked as I was trying many things at once, but you can try different things in display options (like DVI frequency, Alternate DVI operation mode, disabling GPU scaling etc.). I didn't underclock the card as that kinda defeats the purpose. By paper specifications, my card's idle temp should be under 40, but currently it idles at 43-47 C (depending on when the AC is running in my room). I had noticed my card's overheating some time back and I suspected that my card might be reaching it's end. But Then I noticed the same behavior while writing GPGPU/OCL article on different GPUs and had planned to do proper investigation further. But couldn't pursue this further due to working schedule.

Do try playing around in Catalyst and do report your findings.
 
My core voltage (i think VDDC offset) seem to be 1.2V as reported by GPU-z and it is constant regardless of the monitor setup.. Furmark Burn In test now reports max. temps of 88 degrees in 1 min of testing.. this is significantly less than the readings before..(earlier used to cross 105 degrees and shut off)

also when fans are in motion the RPM is about 1800-2000 RPM even when set 100% in CCC
 
Ouch... how old is your card? Seems to be an issue with card itself. If the core voltage is staying 1.2v idle regardless the number of monitors, that seems high, even taking into consideration that this is last generation's card. If it's bought locally and has warranty left, I suggest you get it RMA'ed. I'm not sure what specs the Twin Frozr's fan has, but considering the fact that it's dual fan card, 1800-2000 seems okay. With two fans, I doubt there should be a need to go past 2k RPM. My Sapphire HD6950 had single fan and it would go above 2.5k when set to 80% in CCC.
 
Make sure your displays are connected to the same type of output from the card - i.e. both are connected to HDMI or both are connected to DVI or both are connected to Display port. If there's a mix like 1 display port and 1 hdmi or 1 dvi and 1 hdmi etc, the clocks on the card will shoot up - in my 7970 GHz, it goes up from 300/150 idle on the GPU/VRAM to 500/1500 idle. It is something to do with sync'ing the timing across multiple displays - not exactly sure.

If you are not gaming on the second display, connect it to the integrated graphics on your cpu. Solves the above issue.
 
@Chaos... I did try that once, but for some reason newer catalyst drivers don't play well with IGP drivers (Intel HD4000), I think because of OpenCL. Even when I managed (once) to get both working properly, Radeon's OCL performance goes drastically down. So, as of now... there is not clear way to get everything the way you want. :(
 
Ouch... how old is your card? Seems to be an issue with card itself. If the core voltage is staying 1.2v idle regardless the number of monitors, that seems high, even taking into consideration that this is last generation's card. If it's bought locally and has warranty left, I suggest you get it RMA'ed. I'm not sure what specs the Twin Frozr's fan has, but considering the fact that it's dual fan card, 1800-2000 seems okay. With two fans, I doubt there should be a need to go past 2k RPM. My Sapphire HD6950 had single fan and it would go above 2.5k when set to 80% in CCC.

I got the card as used one more than 20 months back the purchase date is of early 2011 or late 2010. I didn't had any problems whatsoever till this month, I had even Overclocked it and Mined bitcoins some months back. but as I rarely use for gaming I prefer to underclock it (previously used to do it through CCC) but it didnt had many options to tweak VDC or even display it. but as I searched more I found good utility from MSI (Afterburner) and had set the core voltage to maximum. of 0.900 mV but still it jumps to 1.150V I'm not sure why this is happning

w41z.png


note that the GPU usage is no more than 18%
 
@Chaos... I did try that once, but for some reason newer catalyst drivers don't play well with IGP drivers (Intel HD4000), I think because of OpenCL. Even when I managed (once) to get both working properly, Radeon's OCL performance goes drastically down. So, as of now... there is not clear way to get everything the way you want. :(
I do OpenCL code writing all the time and didn't notice any drop at all. Make sure you install catalyst drivers after Intel drivers.
 
Probably the fans are about to die. I had an HIS 4870 1 GB GPU with a fan that died a premature death (within 4-5 months IIRC). Removed the fan shroud and with some zip-ties, attached a couple of 80 x 25 mm medium-speed fans which worked better than the stock fan. The fans were powered directly from the power supply and not the motherboard fan headers. If you have one 120 mm fan, that would do as well.

Could you please share the method you used to attach the fan onto the graphic card? Flexquick/Feviquick should work?

I had tried oiling the old fans a few days back, but as there was no hole in the existing fans I had drilled a bit to make room for one.. It didn't turned out that well one of the fans is now no longer in working order although it wasn't before anyways and the second fan somehow worked after oiling... even then it didn't last long and today my GPU temps almost doubled again :( and reached levels where it automatically restarted..

Replacement fans of same model seems expensive, so I guess I should try with 120mm or 80mm fans.. I guess it should work well as you said.

Still I'm confused what method to use to attach it onto the GPU, please help
 
My goodness! You drilled a hole on the fan?! :dead:

I used zip-ties to attach the fans to the heatsink. The zip-tie would pass through one of the holes of the fan and then you pass the tie through one of the fins of the heatsink. Route it back and tie it. Like this:

95AuJob.jpg


photo3fu.jpg


Attach the fan's header to the power supply's molex connector or on the motherboard.
 
Thank you very much. Will try tomorrow and see the results.
and drilling was the only logical thing to do that time. as it was like trying to Frankenstein a dead body.
 
To everyone else reading this thread looking for advice.
If there is no hole in the fan for adding lubricant, then the fan is not meant to have external lubricant, means please do not drill any holes.

@max_demon,
I am guessing the fan was a sleeveless one. My GPU fan is, and it also does not have any holes. Drilling to Frankenstein the dead body was nice thinking, unfortunately it did not work. Good luck for the next time you try such "jugaad".
Also remember that zip-ties would be the best option for Gannu's solution as the card's backside will also get hot and whatever you use for tying the fans will have to withstand the heat for a prolonged period.
 
My core voltage (i think VDDC offset) seem to be 1.2V as reported by GPU-z and it is constant regardless of the monitor setup.. Furmark Burn In test now reports max. temps of 88 degrees in 1 min of testing.. this is significantly less than the readings before..(earlier used to cross 105 degrees and shut off)

also when fans are in motion the RPM is about 1800-2000 RPM even when set 100% in CCC

Why are you using FurMark. It will burn out your card. It is a non-real-world scenario. Just check it on games. If you are hitting 88C on games, it is manageable. Those cores can handle that.
 
I was trying on furmark to check max temps possible, and also the fact that I don't even have any game installed. just using the card for GPGPU purposes and dual monitor set-up.
 
My goodness! You drilled a hole on the fan?! :dead:

I used zip-ties to attach the fans to the heatsink. The zip-tie would pass through one of the holes of the fan and then you pass the tie through one of the fins of the heatsink. Route it back and tie it. Like this:

95AuJob.jpg


photo3fu.jpg


Attach the fan's header to the power supply's molex connector or on the motherboard.
Today I tried to do it.. there wasn't much space for 2x80mm fans with a bit of improvising managed to attach it onto the Heatsink, but unfortunately I had forgotten that I'm out of thermal paste, due to many times removing applying thermal paste it got depleted. had used up some existing paste on the core and switched on the card.

now even with fans on the Idle temps are 60 degrees. so I'm wondering will applying thermal paste will solve the issue or I will need to change the fans too.
 
Back
Top