Looking for advice to prolong life of my video card

[thestain]

Hi,

I am looking for advice on what can be done to help my old video card.

Right now I cannot game with it. Current best Idle temp is in mid 50's in 2D, but if stressed.. spikes to 100C plus in less than a minute or two and generates nvidia warning message.

I was getting nvidia messages card was too hot from just surfing or watching tv, with tv tuner card in my pc, and heard fan seemed to be going on card, so swapped it out for thermalright HR-03 and had to tweak mounting little by little to get to really transfer heat to cooler.

Initially would idle at 64C, now after working on for a few hours last night and letting cool down and heat up a couple times, idle is 55C

Even though heatsink appears to be doing its job, the core temp will spike to 100C in Ati tool core test within 2-3 minutes as core clock speed goes from 275 default start to 300. I have to abort test or risk ruining card by running test any longer than a few minutes as temps keep rising and don't level off. Thermalright heatsink is toasty to the touch, so appears it is connected well to gpu core, but that rise in temps has me concerned.

I was wondering if card was somehow getting too much juice. Is there any way to know if card is just dying? Can it be saved by undervolting or something else?

Looking for ideas.

Thanks

 

[MarcVenice]

I'm just going to ask the obvious, but it's all I can think of. Are you running the HR-03 passive ? How's airflow in your case ? Did this suddenly happen/start btw ?

 

[thestain]

I was using stock cooler and was getting messages of card running too hot over past two weeks. It sounded like fan was cutting on and off at times.. maybe fan was starting to go.. I had a Thermalright HR-03 laying around.. bought two about a year ago.

Idle temps with nvidia stock cooler for 7900 GTX were in mid to upper 60C's and if I watched TV or surfed net, etc.. would get into mid 70's and then for some unknown reason.. would spike to 100C plus at times and nvidia heat warning message would appear.

So.. I installed the Thermalright. I am using with 92 mm nexus fan at stock at around 1200 rpm.

If I test using Ati tool, temps spike very quickly.. at present, it was worse last night.

say first temp reading is 55, then 70 then 80 then 84 then 88 then 91 then slows a little but jumps to 94, 95, 97, 98, 99, 100 102 and I abort and close 3d view at this point all within 2-3 minutes of starting test.

Case temps are ok, but do rise a couple degrees as card begins to toast with CPU going from 36-41, and case air temp going from 31 to 33. The heatsink, while not hot enough to burn your skin is pretty close, it is doing its job.

Outside of heat messages from nVidia.. no visual problems.. no artifacting, etc..

I was wondering if editing card bios and flashing to lower voltage a .01 or .02 to make up for what might be it getting too much voltage would help.

Room temp is 24-25C

 

[humanure]

try running with the door off and a w/ a house fan pointed at the case, see if that makes any difference.

 

[error8]

It's very weird. The first thing that came in my mind is that the new heatsink is not installed properly, but if you say it's hot to the touch then that can't be the problem. Even so, did you use thermal grease when you've installed it?

Another thing that I don't understand in your first post, in ATI tool you say that the core goes from stock 275 to 300? What does that mean? Your GTX should have a stock core clock of 650 mhz if I'm not mistaken, and it can't fluctuate in any way.

 

[Bateluer]

100 degrees C is waaay too hot, even for a video card. You need to address that issue before you do any gaming, or anything that will heat the card up. Reseat the HR-03, make certain you cleaned up all the old thermal glue off the GPU die, increase air flow, etc. There's a lot of things you can do. But those temps will drastically reduce the life of your card, even it runs fine now, its only a matter of time.

 

[thestain]

The card is overheating well before it gets to stock speeds.

When it goes from 275 to 300 it already is hitting 100C within 2-3 minutes and normally test will take all the way up to 650 and beyond.. My other 7900 GTX goes to 700 and never hits 100C with stock cooler.

I did point a fan at it.. large one with door off.. made very little difference.

Yes I know there is a temp issue here.. that is why I have posted to begin with... I don't like it running so hot ever.

I did check and double check and mount and double mount with new thermal paste from thermalright each time.. I am sharing and hoping for ideas, because this is weird.. maybe some kind of resistor or regulator or something on card is not working right and card is getting more voltage was my guess.

Are there any programs that will let you know what voltage your gpu is actually getting?

I suspect it is getting too much.. I cleaned areas on back of video card that might have gotten dirty, but that did not change things.

If card was somehow getting extra voltage, would that accont for inability of properly mounted and functioning heatsink to get rid of heat fast enough to avoid spike upwards in temps?

If so, could lowering voltage with bios editor be a good idea?

 

[KBTuning]

its possible your temp probe is hosed... do you get any artifacts?


if not i wouldnt worry about ti.

 

[zephyrprime]

reseat it.

 

[taltamir]

is the BACK of the card (the PCB) burning hot to the touch? as in, scalding? if it is really 100c you wouldn't be able to touch it for more then a second or two.
That being said, it probably is, is the headsink rated for this kind of card? take off the heatsink, clean the thermal paste, apply arctic silver 5 as per the instruction, and then reseat it. Don't put too much thermal grease.

 

[firewolfsm]

The probe is off. Your card isn't really that hot. 7900s draw very little power, and this is impossible if you seated it properly. Use something else to measure it.

 

[humanure]

you can use GPUZ to see the voltages and temps of your card.

 

[thestain]

Hi,

Update.. reseated gpu.. thought it had fried.. frozen screen immediate nvidia over heat message, etc.. artifacts, etc.. 97C at idle.. immediately turned off and slightly loosened and retightened and gambled and played with nvidia bios editor.. and flashed bios, lowering voltage just little.. and.. it worked.

Now idles around 50 and if I run Ati tool does not go over 77C after a half hour of testing.. stopped trying for max core after I hit 650. No artifacts or problems for last hour or so.. knocking on wood.

Voltage only adjusted from 1.4 to 1.36 and from 1.3 to 1.24 for 3D and 2D.. so far so good. I never flashed bios before.. but figured.. not risking much after last crash and artifacts on reseating.. think I will now leave well enough alone.

I did try gpu-z, but could not find voltage info.. so still wondering how lowering voltage ever so slightly corrected heat problem.. strange.. but.. after spending whole weekend on this.. I quit.. Thanks to all for your help.

Mike

 

[taltamir]

it is VERY possible for a 7900GTX to reach that kind of temperature.

 

[humanure]

look under the sensors tab for voltages in gpuz

 

[thestain]

humanure,

for my system all that I found under sensor tab were clocks, temps and fan speeds. No voltages.