5870 idle temp getting higher every day.

[viivo]

My 5870 used to idle around 38C with the fan at 21%. Then three days ago I noticed it idling at 45C. The room temperature was roughly the same and nothing in my case had changed nor was anything dusty. For the past couple days now it's been around 51-53C. I have to increase the fan speed to 35% for it to get back to 45C, which is incredibly noisy. There is a 140MM fan 5 inches below it feeding it fresh air.

The card is an MSI R5870, reference design. The fan and heatsink are clean. Any idea what could be causing this? Would it be worth it to try changing the thermal compound?

 

[tincart]

Putting new thermal compound on GPU's has never improved my idle temps by more than a couple degrees.

 

[Zap]

Thermal compounds do not degrade that fast. Do you live in a dusty or smokey environment?

 

[mizzou]

Before you assume something is wrong, have your load temps changed drastically as well?

 

[viivo]

Before you assume something is wrong, have your load temps changed drastically as well?

When gaming I usually set the fan to 45% which keeps it around 65C. It hasn't changed that I have noticed. Without manually setting the fan it ramps up really slowly and I start to crash around 79C.

Thermal compounds do not degrade that fast. Do you live in a dusty or smokey environment?

Not at all. I'm very OCD when it comes to keeping my PC clean and the room is usually ~21C.

 

[Termie]

When gaming I usually set the fan to 45% which keeps it around 65C. It hasn't changed that I have noticed. Without manually setting the fan it ramps up really slowly and I start to crash around 79C.

Any chance you've changed something in your system, like adding a second monitor? That would force higher temps.

You should run MSI afterburner to monitor not just temps, but also clocks and voltage. Unless the fan isn't spinning at its reported speed, the only thing that could lead to such a huge increase in idle temps is a card running at above-idle speeds (for instance 400 for bluray or 850 for games), or the card being stuck at load voltage levels.

Also, on a separate note, crashing around 79C sounds like a problem to me. If the above monitoring doesn't reveal out-of-spec clocks or voltages, then the 79C crash might point to a physical problem with the card.

 

[digitaldurandal]

Is the card new? Perhaps part of the heatsink has become loose and is slowly separating from the gpu? Have you physically examined the card? Even a small gap between the thermal and the HS could cause a large rise in temperature.

 

[soccerballtux]

I've had this with nvidia cards but not with ATI ones. I reseated too. It was weird.

 

[Termie]

Is the card new? Perhaps part of the heatsink has become loose and is slowly separating from the gpu? Have you physically examined the card? Even a small gap between the thermal and the HS could cause a large rise in temperature.

I doubt it's a heatsink issue because this would cause sky-high load temps, which he apparently isn't seeing, since it slowly climbs to 79C (and then crashes). Now, if there were a rapid climb to 79C and then a crash, I'd be more inclined to believe the HSF isn't making good contact with the GPU.

 

[viivo]

Could the connectors from the PSU be the cause? A little over a week before I noticed it rising I rerouted my cables. My PSU (not modular) has 6 unmarked 6-pin connectors with different colored wires on the two top right pins - some are black and blue, some black and red, and I think one is red and white and another black and white. Some are also longer and a little thicker than others.

I know little of electrical things like that. Does the coloring matter? Would certain ones be carrying more voltage?

 

[soccerballtux]

Could the connectors from the PSU be the cause? A little over a week before I noticed it rising I rerouted my cables. My PSU (not modular) has 6 unmarked 6-pin connectors with different colored wires on the two top right pins - some are black and blue, some black and red, and I think one is red and white and another black and white.

I know little of electrical things like that. Does the coloring matter? Would certain ones be carrying more voltage?

Hm. I don't think so...

 

[Termie]

My PSU (not modular) has 6 unmarked 6-pin connectors with different colored wires on the two top right pins - some are black and blue, some black and red, and I think one is red and white and another black and white. Some are also longer and a little thicker than others.

I know little of electrical things like that. Does the coloring matter? Would certain ones be carrying more voltage?

Whoa, time out!!! You have six 6-pin connectors? Unless you have a 1200w power supply, I can promise you those aren't all PCIe connectors. At most, you have four 6-pin PCIe connectors, and if your PSU is under 600w, you probably only have two 6-pin PCIe connectors. Now these connectors shouldn't let you plug them into the wrong socket, but if you mixed these up somehow and jammed them into the video card (or into your motherboard), it's possible you have entirely the wrong power line feeding each component. You better go back and make sure that two PCIe connectors are connected to the two PCIe inputs on your HD5870. Look up what a PCIe connector and cable looks like to make sure.

 

[viivo]

Whoa, time out!!! You have six 6-pin connectors? Unless you have a 1200w power supply, I can promise you those aren't all PCIe connectors. At most, you have four 6-pin PCIe connectors, and if your PSU is under 600w, you probably only have two 6-pin PCIe connectors. Now these connectors shouldn't let you plug them into the wrong socket, but if you mixed these up somehow and jammed them into the video card (or into your motherboard), it's possible you have entirely the wrong power line feeding each component. You better go back and make sure that two PCIe connectors are connected to the two PCIe inputs on your HD5870. Look up what a PCIe connector and cable looks like to make sure.

The PSU is a Coolermaster "Real Power Pro" 1000 watt. I know it's overkill for a single graphics card but I got it in a trade and saw no reason to look for another one.

You're right, I just checked - it's 4 6-pin and 2 8-pin that I was mixing up, however all the 6-pins do have the differently colored wires and two are shorter and thinner than the other two. Apologies for the bad information. I've been away from the hardware scene for a while and feel like a neophyte again.



edit: for reference:

 

[Termie]

The PSU is a Coolermaster "Real Power Pro" 1000 watt. I know it's overkill for a single graphics card but I got it in a trade and saw no reason to look for another one.

You're right, I just checked - it's 4 6-pin and 2 8-pin that I was mixing up, however all the 6-pins do have the differently colored wires and two are shorter and thinner than the other two. Apologies for the bad information. I've been away from the hardware scene for a while and feel like a neophyte again.



edit: for reference:

Ok, looks like you really do have a huge power supply. I assumed you were running something much smaller with only 2 or 4 PCIe connectors. From your picture, it seems that the connectors are the proper ones to use. You could try switching some of your PCIe connectors around - perhaps one of them is bad.

You should still run Afterburner to see what your temps, clocks, and voltage are. You'll need to go into the options to enable voltage monitoring.

 

[tweakboy]

Could be the vga fan its slowing down thus you get high temps. You may need to buy a after market cooler for your card

 

[Phynaz]

Could be the vga fan its slowing down thus you get high temps. You may need to buy a after market cooler for your card

Ummmm......No.

If the fan is slowing down you buy a new fan, not a new cooler.