W860CU HD5870 Overheating Problem

After some readjusting and new thermal job, I got these temps (with the paperclip method):

Idle:


After 2 Vantage runs:


I also found that after cleaning the paste that sticks to the heatsink of the GPU, underneath it is purple in colour.

Here are the pics:




Any ideas why?

 

Was it purple before you paper clipped it?

 

Yeah. It was. I snapped that picture before I did any mods.

 

How were you able to see the complete details of your GPU when using GPUZ0.4.2? I'm using the same but it seems that it doesn't detect even the clocks.

What am I doing wrong here?

 

I got it from [GPU-Z Test Build] Fix for cold bug on GTX 480 - techPowerUp! Forums

 

Thanks +Rep!

 

I've just trimmed my Aluminum can and applied the Paperclip mod butmy temps are still high. It still reaches 100C but now it takes longer, around 5 mins compared to 3 mins before modding.

I've also tried to peel the doubled thermal pads (the three that I encircled in my previous pics) but I've noticed that they should really be thicker. First off, those three have thicker thermal pads than the remaining thermal pad in the recessed area because the 3 square chokes that go into this three doubled thermal pads are shorter than the square choke that goes to the last flatter thermal pad in the recessed area. I've confirmed this by looking at their relative heights and one more thing when I try to make those three thermal pads flatter and then I put them in their respective components sticky side up and then close the heatsink, thy did not stick to the heatsink which simply means that they don't have contact with the heatsink if they are made flatter. So I guess the default thickness of having double thermal pads in those recessed areas is correct.

I'm not sure why sp-1, LaptopNut, and 4649 have all cut into half those 3 thermal pads in question here. I maybe wrong, please correct me.

I really don't understand why I don't have any improvement in my temps, I'm getting frustrated just by thinking of it.

 

Well, higher temps increase the oxidation rate, what kind of paste do you use? Silver nitrate-based pastes have been known to do that, especially at higher temps. The color is harmless and will not reduce the thermal conductivity of your heatsink.



@Kevin: That's five minutes of normal gaming use, or 5 min of furmark or some other stress test software?

 

If you believe you are getting good gpu contact the my only other question would be is the fan running full speed. Check to make sure it is not plugged into the extra fan header instead of the gpu header. You could also pull out the questionable thermal pads completely and just boot into windows to see if your idle temps are lower then you know if they are interfering and can make them thinner. Just don't stress test without the pads in place.

 

The correct fan connection is surrounded by a white outline and says VGA on mine.
I couldn't get a fan to even spin up when I plugged it into that extra connection. I was thinking about using it to power a third fan.

 

That paste, was the IC7 (or so I was told by my reseller). It was part of a "package" when I decided to upgrade my GTX 280M to the MR5870. Since you say that it's harmless, I guess I don't have to worry much. Thanks!

 

That's 5 min. of Furmark/OCCT.

Yes it is running at full speed, I can hear the speed go up while temperature go up. When you made those three thermal pads thinner and installed them back in the heatsink and closed the heatsink, were they flushed properly to their respective square chokes and can you see the imprint on the properly?

 

Yep, that copper could turn green like the statue of liberty and will still conduct heat pretty well.. (Nonetheless, stay away from salt water.)

Hm. And your ambient temp is still in the 32-35*C (90-95*F) range?

The rapid rise you used to have was defintely indicative of an assembly issue, but 5 min in furmark/occt in 35*C room temps might be something that you will struggle to overcome.

A more realistic stress test might be a benchmark of some kind, maybe run 3DMark06 or Vantage and see what you get.

 

I was on an airconditioned room at about 30C when I tested it with furmark/occt. So these temps seem to be normal already? I'll try 3dmarks06 now and see what I get.

 

To get a good idea of real GPU temps I just play GTA IV for a while and even after only a few minutes, I could see the temperature differences between the GTX 285M and the 5870 as well as the differences in fan speed.

Furmark is a worst case scenario and is not really indicative of how good your GPU temps are since any real life use will not place as much stress on it.

For example, after a while of furmark my GTX 285M can reach 88C but during gaming from GTA IV, SR2 to BFBC2, it rarely goes beyond 78C. In some games my GPU temp stays around 60C.

 

Here are my temps after 3dMark06:

 

Ah ha! Yes, that is much more normal, especially since your room is 30*C. The instant jump to 100*C was a problem, but now I think you are right where it is expected to be.

 

Sweet !! Congrats...

 

Maxed out at 78*C too, and my ambient is much lower. I still haven't done any mods (like shims). At a certian point you are limited by the fans that don't ramp up until a specific temp threshold is reached.

 

I never run furmark for very long because it has the potential to damage GPU RAM (The GTX 285M has no RAM temp sensors).

However, the few minutes that I did run it, my temps didn't go beyond 88C. I also get maximum GPU temps of 71C after 3DMark 06 and my ambient is 23C.

Since I no longer have the 5870 I can't double check to see what you meant, to see whether it was correct or not. Maybe wait until the other guys who did the same thing comment. However, it makes no sense to design a heatsink with a recessed area only to fill it up again with a thick thermal pad, think about it.

I also remember the heatsink of the Nvidia GTX 260M, both of the recessed areas had an accompanying flat thermal pad, none of them were thick and if the pads were thick it would have affected the contact of the heatsink to the DIE.

 

Yes I do still have imprints from the chokes. One option you have instead of thinning them down would be to just make them a little smaller so they squish down more easily. The thermal pads they use are pretty soft and move fairly easily but I noticed that they were so close to the edges of the holes they are in that they just hit the edge and stop squishing down at that point.

 

Ok. Around how many mins. Of furmark did you run yours?

Ok. What I don't understand is why do we have the same thickness problem but when I peel of the first layer off of the thermal pads on mine it doesn't have the same effect as yours where it has the choke imprints on them?

 

I ran Furmark just now for 10 minutes and my GTX 285M GPU temps maxed out at 87C.

 

ahh. What are your ambient temps again?

Regarding the recessed areas, yes I got your point if you put a thick thermal pad in a recessed area. What I just don't understand is thay when I peel the first layer of the doubled thermal pads to make them a single thermal pad,
they don't contact with the heatsink which means they are too thin. Why is this not happening to you guys when you do make those 3 thinner?

 

My ambient temps are 23.5C.

 

I had to cut mine they would not peel and i only cut about 1/3 from mine.

 

That's weird, why don't they have the same default thickness (that can be peeled off) as mine and 4649's thermal pads.

With the default thickness (without cutting), were they flashed with the surface of the heatsink itself? I mean are they level with the surface of the heatsink when put in the recessed area?

 

Kevin, just so you know.
I'm choosing between 2 laptops,
Kosti was responsible @the MSI board for me to regain faith in the GX series... just by removing the pads, he instantly went down like 10C...

If you get this done, at your ambient temps...
you'll be responsible for me putting down $1500+ for a laptop.

 

Which pads? Can you give a link to his post?

 

http://forum.notebookreview.com/msi/482836-thermal-past-upgrade-worth.html

have a good read!

 

Post number 17

Having read through, it sounds exactly the same as the issue we were discussing and the same reasons why myself and others reduced the thickness of those thermal pads. I don't think it is a good idea replacing the thermal pads with thermal paste though.

Edit: Be careful with that shim Mod because it looks like it was the cause of both dead GPU's here

 

Yeah I just removed the paperclips after I read that thread. Thanks for the heads up though,

Regarding the thickness of those thermal pads, I still am confused about those three in particular. When I make them thinner (by peeling the first layer of thermal pads off of them), put them on top of their respective components (sticky side face up), close in the heatsink, then removed the heatsink again, I would noticed that they won't stick to the heatsink.

When I do it the other way around (putting them on the heatsink instead of on their respective components), put in and removed the heatsink, I don't see any imprint on them.

This simply means that when you make them thinner, at least when I did, they don't have proper contact with their respective components.

Note: We shouldn't be comparing the thickness of those three recessed area thermal pads to the thickness of the 4th recessed area thermal pad because the component in the 4th recessed area is taller than those of the three (which has the same heights) and thus need a thinner thermal pad indeed and thicker ones for the first three recessed area thermal pads.

 

Simple solution, compare the size of the pads to the resulting temperature of the GPU RAM during a benchmark like 3DMark 06. Hwinfo32 will display the temps for the actual RAM chips and you will know what is better. However, without a mod to give more pressure on the heatsink to the DIE, none of that will matter much anyway. So since the shim / paperclip Mod has the potential to kill the GPU, there is not much left to do.

 

I'm pretty sure HWiNFO doesn't show the temps of the GPU RAM chips, actually; at least not for the Mobility HD 58xx cards.

 

What do you nean compare? Does the ram chips need the pressure mod also?

 

I mean that you could compare the temps under load before and after changing the thickness of the pads.

 

Maybe your best bet then would be to compare the difference in height of the pads in the recessed areas of the heatsink to the difference in height of the components. The problem is probably that while 1/2 height is to short full height is to thick.

 

Same thing happened to me. My speculation is that, buy using the shim mod, yes, the contact with the core is better hence the temp drops, however, because the heatsink is pushed up by the shim, it's likely that the memory has bad contact with the heatsink, and the memory could be getting very high temps and you won't know it as there's no temp sensor for the memory, and over long term the mem could just die from high temps.

IMO, the best solution is still to just send it back and ask your seller to adjust the heatsink, it takes precise placement of the thermal pads. And no, don't use thermal paste to replace the thermal pads, I did that and it's a complete failure, the gap between the heatsink and the mem is simply too far.

 

yeah, i@m guessing that is the best reason I thought of too. I think I have to be around 3/4 of the original thickness. How do you handle the therml pads when cutting them/. They are pretty sticky and very hard to cut I presume, and you even have to wear plastic bag to avoid oil from your fingers to goto the pads.

I also noticed that all the time the MemIO temp in HWInfo is always the highest among the three. Is that indicative of poor contact with the memory chips? And is that MemIO sensor even accurate?

 

memIO is not even the memory temp, just the memory controller which is still in the core chip. There's no sensor for memory temperature.

 

does the paperclip mod leave a permanent bent on the metal springy joint? Or it returns to its original state when I remove those paperclips?

 

Indeed. I think the memory controller seems to have to work very hard to keep up with 1000MHz GDDR5, which is why it gets so hot.

Compared to a 256-bit controller using GDDR3, although the bandwidth is the same, interfacing with the 256-bit bus takes a lot more space and so the heat isn't quite so concentrated in that case. I don't know how the power consumption compares - it might be roughly equal, in fact - but I would think that because the memory controller does so much work in such a small space (40nm makes it even smaller, though it should also reduce power consumption) it's no surprise that it gets rather hot.

 

My temps are pretty even for all 3. I have nice defined impressions in all my thermal pads. As for cutting Mine were pretty easy since they were one thick pad. I don't know how best to cut them since yours are thin; you might be better served by just buying some thermal pad material of the correct thickness. If I had to guess the memory controller being hotter probably indicates that the heatsink is not making good contact in that area. Most likely being held up by the memory chips and power circuity thermal pads.

 

If I buy thermal pads, what are the correct thickness for all the thermal pads?

 

Well you could measure them yourself. I would assume since the two sacked are probably 2mm each a 3mm pad would work.

 

I have a W870CU with a MR5870 and this is my laptop at idle in a 22C room.


I've been trying to follow this thread, but it's difficult to follow, what would you guys recommend I do? I'd really rather not RMA it, but I will if I absolutely have to.

 

That's pretty high for idle, did you reapply thermal paste?

I know that the Clevo 5870M doesn't use Powerplay, so you will have to manually down clock via AMD GPU clock tool, and you can set hot-keys for the clock settings.

But it shouldn't be that high idle.

 

Have you cleaned out the dust from the fans/exhaust vents (properly including inside the fan casings)?

 

I haven't yet, I just realized it recently and was hoping it was a driver issue.

I have not, but I live in a very dust-free environment, and there seems to be barely any dust in it. Would opening the fan casing require re-applying thermal paste? It seems a pretty singular unit.