W860CU HD5870 Overheating Problem

No modding at all. Yes, it is a 8760. Don't know how much difference that makes.

My heatsink rides up on the can over by the CPU. Given the temps, I've seen no reason to make changes.

 

Wow those are good temps. When did you buy your laptop? Would it be possible for you to take some pics of the GPU setup with various angles?

 

Those temps are beautiful.
+rep

 

How long was the run of Furmark and wPrime?
I guess it was the length of a 1024M run, but that's not entirely fair with the i7-920XM because it does it so fast

One other thing to note is that you should probably run the CPU stress test on 7 threads (or maybe 6) to leave FurMark the CPU power it needs.
15 minutes of Prime95 + Furmark Xtreme Burning would be nice to see. That, I'd be scared to do on my MSI GX640.

 

Bought it from GenTech in March. Sorry, I'm not going to open it up again unless there's something I need to do inside. All you could see of the GPU is the heatsink and pipes.

Furmark would have been probably 15 minutes, could have been 20, but there's really no difference in my experience. Once it levels out, it stays there pretty much indefinitely. On WPrime, it probably was 1024. It does go fast.

Thanks, Ben. I suppose i got lucky with the factory paste.

 

Yeah, I think past 10 minutes it's only going to go up by 1 or 2 degrees at most. On my GX640 (running only FurMark) the temperature curve was basically flat after 10 minutes, although the temps on the GX640 are much worse than on the Sager (see here).

 

Honestly, all in all, thats not bad AT ALL for the gx640...

 

That's my experience too.

Whoa Cheese. That memory temp will make you sit up and notice.

 

MemIO is the memory controller on the GPU, not the memory.
105ºC is definitely on the high side - supposedly the Mobility 58xx does a thermal shutdown at 110ºC on MemIO based on this guy's G73.

Nonetheless, I'm not a fan of furries, so I doubt FurMark is going to be representative of any of the games I'm going to play on my GX640.

 

Couldn't agree more about the furries. Make that GX640 sing in the real world.

 

When I last pulled the heatsink and pads off, I recorded the info off the memory chips and it turns out they are the Samsung K4G10325FE-HC04.

These chips are rated to run at 1250MHz (.4ns) with 1.5 volts. I don't think they get that much however and that is why overclocking (in addition to the heat) is tough beyond ~1100.

The whitepapers indicate that these chips' designed temps are rated by Samsung as "Commercial Normal" which is 0°C ~ 95°C.
Whether you actually want to see 95C is a whole other story.

Either way, 95 or above and you are living dangerously. (And due to unaccounted variables and senors that are not always the most accurate; you really don't even want to be close.)

 

As I said before, MemIO is not the memory, it's the memory controller on the GPU.

 

True, but the single copper heat plate will spread the heat to any other components that are in contact with it.

 

Considering that in my test other parts of the very same GPU (Shader and DispIO) were running at 94°C and 88°C respectively when the MemIO hit 105°C, I wouldn't think the memory would be running above 95°C at the time. In any case, I do agree that I wouldn't want to be running FurMark on my laptop for long periods of time (although who would?), but as I said before FurMark is quite far from normal usage.

 

It would be very close, correct? I see what you mean, but it's an unknown variable and I (personally) have to err on the side of keeping all the components away from that temp.

It's your perference, and really since ATi doesn't release any MTBF stats we'll probably have to a wait a while to know how well these cards work out for people after a year or so of stock temps vs. those who modified it to be cooler.

 

Not running FurMark is a great way of erring on the side of caution, however
Temperatures during gaming are far lower than in FurMark, and if the card does fail, that's what having a warranty is for.

 

Furmark is with extreme burn is a completely unrealistic test. No game to my knowledge ever runs the GPU that hard. It's a good way to bench your system's cooling capability vs others but that's really about it.

 

Oh, I agree. But I've seen my 5870 at 85*C (in 20*C Ambient) while playing games (Stalker: Clear Sky, Enhanced HDR, DX10.1, 1080p). Which is concerning as that game isn't the most intensive out there.

Even at fairly excessive temps your card can last a year to be out of warranty. Extended warranties should be an optional "peace of mind" feature, not required if you want to keep your laptop more than a year. I'd rather just skip the whole breaking part in the first place.

Main thing here is the OP had 71*C idle. That is just something you don't want for the long term and is indicative of an assembly issue.

 

With my 860 on my lap in bed I top out in furmark at 75C. Anyone experiencing poor temps on the 5870 just do not have good contact with the heatsink.

 

I have a 3 year warranty on my laptop and I didn't pay any extra for it, though unfortunately the global coverage is only for 1 year and I'll be going back to Australia in July. In any case, I do agree that extended warranties are often rather overpriced, and I'm not in favour of paying hundreds of dollars extra for them.

Yeah, there's definitely a problem there, considering it takes only 5 seconds for the card to reach 100°C in FurMark.

 

My average idle temp is around 65C. If you're getting 85C when you played a game in 20C ambient then I guess my temps with 35C ambient are pretty normal right? Ambient temps have a very big effect in th temps of the components we are monitoring.

With some bit of reseating here, the card doesn't reach 100C now in 5 seconds. It reaches it in about 3 minutes or so. I have to try to put the card without the aluminum can and with the spread method of Shin Etsu and see how it goes.

 

15 degrees higher for the ambient temperature ought to result in 15 degrees higher component temperatures, but only if the fan speeds are the same.
Given your high ambient temperature, reaching 100 degrees Celsius in 3 minutes of FurMark is probably acceptable - my GX640 would almost certainly get that hot faster if I ran Furmark at 35 degrees Celsius ambient temprature - but 5 seconds is too quick.

 

I just posted above your post that it reaches 99C in about 3 minutes now but I have yet to optimize my reseating. I just can't stand it reach 100C so I exit Furmark/OCCT by that time.

Did you personally tryit without the Aluminum can?

 

I updated my post as well. My laptop is quite different, so I can't comment on the specifics of your heatsink setup, but you should see how the temperatures go when running an actual game.

 

Now here are some pics:

Here's the top view of the heatsink:



I've noticed that the thermal pads in the encircled recessed areas are thicker than the rest. Shouldn't they be thinner since they are already in the recessed area and the components that go in there are higher?


Here are the pics of the 4 edges of the heatsink in an angle:










I hope I can understand this better now that I posted these pics.

Thanks.

 

I am very happy to report that I have completely solved all of my ATI 5870 heat issues by using this mod Thanks to witcher4360

I also reduced the thickness of some of the thermal pads that had twice the number applied instead of just one but I only noticed this after physically placing the thermal pads onto the GPU's chips and then noting what size they should be. It is very easy to miss this as simple as it is.

Now my 5870 runs cooler idle and under load and my fans no longer spin up too much causing unnecessary noise like they did before. The fan noise is now quieter than it was when I was using my GTX 285M for some reason.

I am not sure about the temps because HWinfo shows so many different ones that I don't know what one I should be looking at.

@ kevindd992002's pics.

Does everyone agree that those three thermal pads that he has circled should be a lot flatter?

Besides, it makes no logical sense having a recessed area and then pads thick enough to fill it since why have the recessed area in the first place

 

Dear Lord, was it that simple?

 

I would have expected that a completely new heatsink specifically designed to fit the 5870 would have had enough tension to actually make good contact with the DIE to the Heatsink. After all of that fuss, I won't be using it anyway because I get much better performance in terms of smoothness and stutter free game play with the GTX 285M than I do with the 5870. A friend of mine has already taken the 5870 of my hands lol.

 

I'll haven't pulled my heatsink off in a while (waiting on some 3M 5519 pads to arrive), but the pads would only be the correct thickness if the higher components sit roughly 2mm higher. The regular pads are 2mm and the thicker ones are basically flush with the copper plate.

LaptopNut has the right idea though, lay all the regular thickness pads on their respective chips, the taller chips should be about even with the regular chips AND their pads. If they are still taller, then the recessed areas need thinner pads.

I plan to implement those shims on the spring arms as well. This is a big difference in laptops as opposed to desktops. The components just don't have as high of a pressure between them and their heatsinks. A lot of the modern pastes we're using work best under higher pressures.

The only foreseeable problems with the shims are breaking the rivets on the arms or accidentally applying too much pressure and damaging the solder joints of the die (which may not happen right away).

 

So you completely rule out the 5870>>>>>285m situation??

 

Well, even if it's better in some situations, I doubt it's worth the extra cash for the GTX 285M.
The "5870>>>>>285m situation" is definitely a falsehood, though.

 

I've been doing massive research... on this... and yes...
5870 >>> 285m in benchmarks, undeniable..
But in real world scenarios; they are kinda like the same, or 2 fps apart...

 

What you state is certainly true but all other laptops I have used have never had this issue of not having enough pressure between the components to the level where it caused overheating. This was the first time I had that issue and I doubt it is common.

As stated before, the Nvidia GTX 285M GPU and heatsink can be screwed down with normal pressure with absolutely no problems and perfect contact. Another laptop I had that used the GTX 260M in the MXM 2.1 format was also perfect.

I used to think that myself and I was sure the 5870 would absolutely crush the GTX 285M in real world gaming. In practice I just noticed more pop in, more stuttering in some games or exactly the same performance as the Nvidia but in most cases, the Nvidia ran with less issues.

The 5870 felt like a downgrade in some respects, more heat unless I modded the sink and having to dial down settings in GTA IV (my favourite game) to get the same smooth performance as the Nvidia.

I noticed better performance with the 5870 in a Just Cause 2 Concrete Jungle benchmark but it was not a lot.

One things for sure, I will be keeping a very close watch on what Nvidia will be offering as an answer to the 5870 but for now, I am very glad to have the 285M.

 

Here is what I did a while ago on mine. Since paperclips are common it is an easy mod.

 

Have you guys seen the brand new W860CU review comparing the 5870m and 285m in Tom's Hardware?

AVADirect's W860CU: Mobility Radeon HD 5870 Vs. GeForce GTX 285M : The 'Build Your Own' Alternative?

 

I thought modern GPU and CPU dies nowadays can withstand and intense amount of pressure before being chipped off?

So if I put some shims on the spring arms, I should not screw down the heatsink all the way?

 

(Snip photos)

Very clever. +Rep for sharing. (When allowed)

 

@LaptopNut: You are quite correct, this shouldn't be a problem. It seems the design or manufacturing of the heatsink is shoddy. I think the desktop allows for less precise mating surfaces because of the pressures available. Which is why these small differences are producing large results on our equipment.


@SP-1: Paper clips is a great idea, I've been rummaging around for a bit looking for something suitable.


@kevindd992002: The actual chip can take the pressure no problem, the solder joints of the gpu and the memory chips are more fragile, especially when you heat cycle them like we do in our laptops.

I'd still tend to think that the heat is your main problem, so as long as you don't go overboard with the pressure, the benefits outweigh the risk.

 

Ahh. But when using paperclips/shims, you still can tighten those 4 screws all the way?

 

Yes I fully tighten the screws.

 

Revelator,

I would edit your post, remove the link and say something like moved. Start a new post for this. It's worth having as a new post so that it's gets some discussion. It got virtually lost in this thread.

 

Read that review, and seems to be an nvidia-fanboyish review, it kept on harping about how 5870MR sucks because it's an underclocked desktop 5770, and only barely winning 285M which is a desktop 9800GTS. What does he expect it to be anyway, a 5870 in mobile form unneutered?
In his review, every bench where 5870MR wins it sounds like it stabbed him in the heart or something, and he will just say it's expected, and it's only winning by so little, and how 285M is just a weak 9800GTS that 5870MR should be completely stomping on but not doing it.
Also, the review is like simply neglected DX11 like it's not there, the only mention of DX11 is in the benchmarks on saying how bad dx11 is because it lowers the framerate, I'm sure he would prefer a dx9 instead if it offers higher framerate. Also, from the benchmark, it's kind of obvious that the 5870MR is using an old catalyst driver as well. Moreover, comparing a mobile card to desktop card is simply retarded.

Then again, I've already lost all respect for tomshardware's reviews long time ago, ever since they had an obviously-apple-bought review justifying on how a mac is not overpriced hardwarewise compared to a diy PC (of course, by using all the overpriced parts for the PC).

 

That's one ****ty review

 

Guy seems to be offended by the whole concept of performance laptops, as well as a NV fanboy. Agreed, lousy review, not because of the conclusion, but because of the attitude and bias.

 

Paperclips... Omg... What an easy solution and here I was using back the part of the can that I trimmed. Thank you. Thank you. Hopefully I can lower my temps.

EDIT:

Paper clip method WORKS. Now I'm idling in the mid 50's as compared to mid 60's at stock clock. Woooohhhhhhhoooooo!

 

Do you actually have a 5870? Because your sig says otherwise, I haven't heard anybody having temp problems with the 280M

 

I think he upgraded it, several people here have. The 280M idles closer to 45*C because it downclocks properly. That or the assembly was not right in the first place.

 

I do. I just forgot to update my sig. XD Sorry for the confusion.

 

Its just as LaptopNut did, he Upgraded to 5870...
seems like the GPU heatsink system was designed exactly for the 280M, saying that means that the can makes exact and perfect pressure on the chip, thus resulting in lower, managed temperatures.
With the 5870 it is different, you have to add paperclip to FORCE the can to make pressure on the chip (as if the 5870 would be thinner than the 280M), resulting in manageable temps due to the "correction" in heat transfer due to gpu heatsink pressure with gpu..

bad english is bad, I know.

 

Agreed, Toms Hardware is a crap site. I never visit the place because of reasons like this. His conclusion just serves to incense anyone with an iota of tech knowledge, especially when he pokes fun at the 5870M for barely edging out an old 8800 GTS core:

I guess he forgot to notice that mobility requires additional R&D and the market is far smaller than desktops, thus the added costs. What a tool.