HELP! Can't keep HD5830 cool

[MWink]

Ok, I'm really starting to go crazy trying to solve this problem so any help is greatly appreciated! I have a Sapphire Radeon 5830 ( http://www.newegg.com/Product/Produc...82E16814102946 ) and I just can't get the temps down where I think they should be. I guess I should start at the beginning.

At first the card worked fine with its stock cooler. I could easily OC the core an extra 100MHz and the temps were fine (mid 70s I think). Then I started to hear the all too familiar sound of a dying fan. I took it apart hoping I could just oil the fan, just to find it has a sealed bearing with nowhere to even put any oil. Since I had taken the HS off, I cleaned the original thermal compound off the core and HS and put on a fresh coating of AS5.

As soon as I put the card back in I noticed my core temp and fan speeds increase substantially. I know they say AS5 takes some time to cure so I gave it a few days but things did not get any better. I took it apart again and tried some AS Ceramique with no better results.

Since I knew I'd have to replace the fan soon anyway, I ended up buying an Arctic Cooling Accelero Twin Turbo Pro ( http://www.newegg.com/Product/Produc...82E16835186046 ) Since I'm dirt poor right now I got the "open box" one for $21 but that's beside the point.

I installed the new cooler and now my temps are even worse. I now have to underclock 100MHz to get the temps I used to get when I overclocked 100MHz. I'm afraid it will burn up even at stock speeds.

The one weird thing I noticed with the new HS is that only 2 of the 4 heat pipes seem to be heating up. The 2 close to the output side get burning hot while the 2 close to the power connector side are barely warm. In my mind this would mean that the HS is not quite mounted flat but I see no other evidence of that, and I can't figure out why.

I've reinstalled the HS half a dozen times at least and the temps just stay the same. I'm just not sure what to do. I'm fairly sure the problem must be with the card or chip and not the HS since they both seem to have the problem now.

Right now I'm running the chip underclocked 100MHz so it doesn't fry. Otherwise it can reach over 85C with the fans at 100%. Does anyone have any ideas or suggestions? I just don't know what to do and I can't afford to replace the card. Thanks in advance for any help!

BTW, mods if this is in the wrong category, feel free to move it.

 

[upsdriver]

My 5850 runs at 86C at load (no OC), fans on auto and they're only at 35% or so. Maybe check if the heatsink has good contact with the GPU but I don't think you need to worry about 85C on a video card.

 

[d3fu5i0n]

Ah. Darn, dude I feel sorry for you here.

Hmmh, I haven't had any modding experience with graphics cards, but perhaps you should've seen if they would've accepted it to be returned/replaced even though you OC'd it. Now that you've removed the HS and replaced the compound, that'll be difficult.

I agree with upsdriver, check the contact. Although, they seem to have missed your point about it reaching 85C at 100% fan speed. Now that is a worry...

I'd re-do everything with the stock heatsink. Remove ALL the compound, and clean it up, re-apply the correct amount, resit the heatsink and make sure it has the same tightness as the card originally had [if there's not a way to make it tighter, I am unsure. I haven't modded, so I wouldn't know ].

I'm most probably not of much help, but I'm just thinking as best I can off the top of my head!

Hope all goes well!

 

[MWink]

My 5850 runs at 86C at load (no OC), fans on auto and they're only at 35% or so. Maybe check if the heatsink has good contact with the GPU but I don't think you need to worry about 85C on a video card.

Basically I am trying to figure out HOW to determine if the HS is making good contact. I suspect it is not but I can't think of a way to confirm it. I've tried shining a flashlight through to see if there is a visible gap but I really can't tell for sure.

I'd re-do everything with the stock heatsink. Remove ALL the compound, and clean it up, re-apply the correct amount, resit the heatsink and make sure it has the same tightness as the card originally had [if there's not a way to make it tighter, I am unsure. I haven't modded, so I wouldn't know ].

I've already tried this a number of times to no avail.

Thanks for the suggestions. Keep them coming.

 

[d3fu5i0n]

I've already tried this a number of times to no avail.
Thanks for the suggestions. Keep them coming.

Yeah, I wasn't sure how clean you made it before doing it perfectly from scratch.

This is annoying though. Another reason why not OC'ing and keeping a warranty pays off - but I suppose this is a rare occurance.

You know you said that the fan appeared to be 'dying'?
Perhaps the GPU was damaged during that situation.
I don't have enough knowledge on how GPUs can retain their integrity or receive any possible damage from overheating, so that's just one guess of many

Still, I wish you the best of luck.

 

[MWink]

I cleaned the old stuff off using Arcticlean so it was about as clean as it gets. Also the old fan never failed to the point where it was not cooling the GPU. It just started to develop a case of the clicks.

 

[TakeNoPrisoners]

My 5850 runs at 86C at load (no OC), fans on auto and they're only at 35% or so. Maybe check if the heatsink has good contact with the GPU but I don't think you need to worry about 85C on a video card.

Wow, mine will hit 70C max and it only does that when I use furmark.

 

[BD231]

Did you remove the thermal pad that came with the new sink? And does the heat spreader on the GPU look level?

 

[KingFatty]

Your description is consistent with the heatsink not sitting flush on top of the chip.

I wonder if there is a speck of debris on top of the chip, like on its edge or a corner, that causes the heatsink to tilt slightly and open a gap. The reason I say this is because you mentioned two heat pipes are hot, and two are warm - that could be caused by the heatsink being tilted so that it doesn't fully engage the top of the chip.

Another thing that could cause the heatsink to tilt is an issue with the bracket/backplate/support used to mount the heatsink. Are you sure there wasn't something that interfered with the backplate installation? Besides the obvious like missing a spacer/nut or a bent bolt, what about something harder to detect, like a speck of debris that got stuck between the video card's circuit board and the backplate/bracket on the back-side of the circuit board opposite the chip? That could slightly tilt the mounting hardware, so that when you bolt down the heatsink, it causes a gap on one side, leading to warm heatpipes where the gap is, and hot heatpipes where no gap exists.

I just can't see how anything but a physical issue could be causing your symptoms. Are you able to take out the card and photograph it from the side, or otherwise measure it to verify that the heatsink is perfectly parallel with the chip?

You mention re-mounting the heatsink multiple times, and maybe that would let you catch the issue if it were debris between the chip and the heatsink. However, I bet you would still miss the issue if it's something with the mounting hardware or debris between the mounting bracket and circuit board on the backside of the video card.

So take everything apart and examine it closely for anything being bent or anything out of place such as something blocking flush-mounting of the bracket.

 

[Arkadrel]

Other users on newegg are saying stuff like their 5870 dropped 15 degrees after putting on this :

"ARCTIC COOLING Accelero TWIN TURBO Pro Fluid Dynamic VGA Cooler"


Yours actually got worse.... it has to be a fitting/mounting issue.

Are you useing artic silver paste? sometimes, their will be "corn"'s of it that ll harden, one of those can make the gpu have gaps without contact.
Im not a big fan of artic silver, bought a tube, so im still useing it, but I hate always getting silvered fingers that cant wash off (ontop of the hardning issues, and it being a conductor of electrisity.... im not buying that stuff again).

Typical error too, is useing to much thermal grease, it needs to be very little, next to nothing you put on, giveing your cpu/gpu a bath in it doesnt help in anyway (its actually something that ll hurt performance more).


**edit:

it seems this cooler (artic cooling accelero twin turbo pro), is hit or miss.
On newegg (comments), either people mount it and it fits, or people *fight* to figour out how to mount it, and then get appalling results with it.

 

[RavenSEAL]

This sounds to me more like the chip rather the fan...

 

[Arkadrel]

@RavenSEAL

stock cooler fan is dieing... ofc it gets hotter then.
Not chips fault.

look at newegg, see how many have success with the non stock cooler, and see how many fail to figour out how to mount it properly and run into issues.

again not chips fault.


Im not saying it cant be the chips fault, just that it doesnt sound like it is.

 

[Schmide]

As someone who changes cooling a lot on my cards; including this exact model, a few things can go wrong with mounting.

I've had poorly spaced screws that actually did not produce the proper mounting pressure. This ended up burning up a 8800gs.

I've used an included sticker shim in addition to the metal shim that is pre-mounted on many graphics cards, this caused poor contact but not enough to have a total lack of cooling. Card survives. I've decided regardless of whether there is a metal pre-mounted shim, do not use the sticker shim in any mounting.

It is possible to slightly dislodge this metal shim and cause poor contact with the chip. It is actually quite easy to remove it. If it does come loose, a small amount of rubber cement is good to keep it in place.

I recommend using standard white zinc-oxide HS grease. Its advantages of non-conductivity and ease of spread over naked chips, far far outweigh any minor lesser heat transfer you may occur.

When mounting for the first time, it is best to mount twice. Spread your HS compound over the whole core, mount, unmount, check spread, then remount. This does not degrade the HS compounds effectiveness and insures good coverage.

Although I watercool most of my graphics cards, when I return the original HSF, temps often go down with the use of the the white grease and proper mounting. The thick hard pads they use in the original mounting have much to be desired.

 

[MWink]

You guys have made some great points, unfortunately I've already looked into many of them. I can't find any imperfections, specs of debris, etc. that could be causing this. I remounted it a few more times today with no changes.

I've mounted plenty of aftermarket GPU coolers in the past with no issues. I'm also pretty good at using the proper amount of thermal compound. Just enough to cover the core, without much extra getting squished out.

Neither of the coolers use a backplate so that isn't the problem. The chip has a metal shim and it seems to be stuck on there pretty well. I'm pretty sure the shim is not making contact with the HS. Also all the mounting hardware is there and appears to be functioning properly.

The only thing I've noticed is the thermal compound looks a little different in two spots on the core. Almost like those two spots are where the core and HS are making the best contact. The thing is, I can't see anything different about those two spots when I clean the thermal compound off both parts. I've got some pics of this. Maybe I'll post them if I can find somewhere to upload them.

I'm almost tempted to try lapping the HS but that's something I've never done and I'm afraid I'll screw things up worse. For the moment I've got an extra case fan blowing down on the back of the card which has dropped the temp a few degrees.

 

[KingFatty]

Hmm, do you have a voltmeter? Also, do you have a Kill-A-Watt?

If your card is making too much heat, that heat must be generated by electricity. If your card is pulling extra wattage or voltage, that would at least help you figure it out and exclude the heatsink as a problem.

Your board is a blue Sapphire I believe, and lacks software voltage control? There should be a pencil mod for it, to increase the voltage beyond stock, as there is a feedback pin where you can adjust the voltage by affecting that pin (e.g., drawing graphite around the pin, or putting something conductive on it, to change the resistance between that pin and ground - maybe there is a drop of thermal grease that happens to be conductive touching the pin and raising the voltage?).

Long shot, but perhaps there is something, a smudge or even grease from a fingerprint, that is throwing off the voltage and feeding more to the GPU to drive up its heat.

You can verify this by a voltmeter (will precisely/accurately tell you exactly what the voltage is so you know for sure) or also by a Kill-a-watt meter (but won't be as good of a trouble-shooting tool because even if it's too high, that could be from something else besides the video card.

I think your blue sapphire circuit board won't allow software to read the voltage, so you'd have to use a voltmeter if you want to check this, and can't rely on something like Afterburner or other software (which reads the voltage regulators on cards that come with software-controlled voltages, and I'm assuming your card lacks that function and instead relies on a feedback pin where you need to externally find the voltage).

But I'm still very puzzled why some heatpipes are hot and some are just warm. Perhaps that is the intended way it would work as the non-hot heatpipes are the ones getting the most cooling from the fan? Still, a drastic heat difference seems like it can't happen in a normally installed heatsink.

If you don't have a voltmeter, it's a handy tool to have around and you can find them for $9.99 at Sears when they go on sale (look for the kind that has a function to beep when it detects continuity - comes in handy when you are troubleshooting compared to cheaper voltmeters that lack the beeper and you have to keep looking back at the meter to check for zero resistance as you probe around for the broken circuit).

 

[monkeydelmagico]

Apply load to GPU until temps come up to near max. Stop load. Temps should drop at least 10 celcius within like two seconds. Then another 10 ish in 10. If you get slow temp declines you ain't mounted.