880M heatsink on the 780M

FYI, I just cross checked the service manuals for P370EM, P370SM, and P377SM-A, and it turns out the GPU X bracket and heatsink screws are almost identical* across all 3 models. This means that P370EM owners with the 680M could probably use the 880M heatsink as well and enjoy cooler temps as the heatpipes are beefier (6mm vs 4mm). Also, the CPU heatsink for the P370EM fits the P370SM, so the reverse might also be true. If so, you could potentially use the 780M or 880M slave GPU heatsink and keep those temps down as well. However the CPU will probably run hotter due to loss of the third heatpipe.

*the screws for the vRAM heatsink has a different part number in the P370EM (6-35-B6120-6R B instead of 6R 0 in the P370SM and P377SM-A), however this may be a non-issue as this screw goes into the bolt screws that hold the GPU to the motherboard, and those bolt screws are the exact same across all 3 models

======================================================================================================================================

Inspired by HTWingNut's review of P377SM-A; photo credits due for shots of the 880M heatsink

Summer in SF Bay Area + no AC in house = ambient temperature of at least 25°C (77°F) when gaming = massive strain on cooling system. Doesn't help that the games I play (Crysis series, Far Cry 3, Watch Dogs) enjoy turning my P370SM with 4900MQ+780M SLI into a micro furnace.

When I was going through HTWingNut's review, I noticed that the 880M heatsink for the master GPU in the P377SM-A seems to have slightly improved compared to the 780M heatsink: it's now a one-piece construction instead of separate core+vRAM heatsinks, so I figured this might give better airflow, and as Meaker pointed out, more fin area. After cross checking the service manuals for P370SM and P377SM-A to be sure that the X-bracket and screws are the exact same between the Clevo 780M and 880M, I went ahead and ordered the 880M master GPU heatsink. Took a gamble but hankfully the 880M heatsink fits the 780M just fine, and the heatsink itself fits snugly into the P370SM chassis. Of course I still had to verify the 880M heatsink would actually bring some thermal benefits, so I fired up a quick n' dirty Kombustor stress test.

Test parameters: Kombustor 3.0, 1920x1080, 8X MSAA, Fullscreen, 15-23 minutes (will explain)
Prior to testing, heatsinks and fans were thoroughly cleaned, and fresh MX-4 paste was applied. The test began as soon as iRST finished loading.

RESULTS:



As you can see, after running Kombustor for 15 minutes, the 880M heatsink cooled 780M is 3°C cooler than using the default 780M heatsink. I decided to let Kombustor go longer just to see if it will reach the 87°C recorded with the 780M heatsink. I stopped after 23 minutes because the temperature seemed to plateau at 85°C. What I'm not showing in the tables above is that between minutes 19-23, the temperature constantly fluctuated between 84 and 85°C, so it seems reasonable to conclude the temperature maxed out at 85°C. I will admit that a temperature difference of 2-3°C could easily be attributed to paste job quality or better die-heatsink contact, but temperature certainly creeps up more slowly with the 880M heatsink. Also, it's debatable whether the 87°C seen with the 780M heatsink would've increased further had I let Kombustor run for another 8 minutes.

All in all, I'm happy with the improvement, considering all I did was swap one heatsink for another. As usual YMMV but I'll be using this 880M heatsink from now on.

COMPARISON OF HEATSINKS:



Forgot to say: the thermal pad for the VRMs on the 780M heatsink is actually one long strip along the bottom instead of 2 short strips as you see in the above picture. I swapped it out because the long strip seemed to offer better contact.

One small gripe: There is a strip of empty void in the 880M heatsink for no apparent reason at all. The fins above the vRAM heatpipe are missing. Thankfully Clevo had the wisdom to add more foam so the fan is flush with the heatsink. I'm going to assume this is for cost cutting purposes, but seriously Clevo? Hopefully you can see what I'm talking about:

Location of void: (photo credit to HTWingNut)


The void itself:

 

Is that the master GPU? Thought that was the slave GPU.

 

Very interesting indeed, I noticed that there is a strip of tape that wraps around the fin array of the VRM heatsink on the 780m version. Have you noticed? I wonder if It would make any difference if that was removed.

 

is the left or right GPU the master GPU?

 

Left GPU when the notebook is turned over lying on the lid is the primary, right is the secondary. It is marked on the SLI cable this way as well.

 

so the thread starter have marked the pictures wrong right?

 

No. That is the heatsink that is on left side of the machine when it is lying on its lid. When using the laptop, the heatsink is on the right and that's where all the heat comes out when using a single GPU.

 

With the laptop vents facing upwards or away from you when you take the cover off, the master GPU is on the left.

 

So, right one on this picture is slave?

 

That's an Alienware machine, isn't it? Alienware machines with the graphics cards under the keyboard, left side is master (so if it was lying on the lid like the Clevo, the now inaccessible video card would be on the right side).

 

ahhhh, now I understand! Sorry, my mistake

 

You mean for the black tape on the 780M slave GPU heatsink? It decreases air leakage I believe, but otherwise the heatsink is right up against the fan so the difference should be minimal.

In the case of the 880M heatsink here, that tape is absolutely essential because it covers a void in the fin array (why on earth Clevo chose to do that I have no clue).

 

Its confusing because Alienware is one of the few machines that does this. You think its bad when you're trying to figure out which card is which, imagine when trying to find replacement fans lol.

Rule of thumb - when you are looking at the GPU heatsink, its going to be the one on the left. Alienware puts the fans beneath the cards, Clevo puts fans away from cards and uses heatpipes... Alienware is much better cooling for this reason (usually - their design really sucks when the fan gets clogged, taking apart those machines is highly unpleasant compared to a Clevo)

 

yeah, looking forward to get my P570WM for this reason.

 

I can't tell you how many times I thought I wasn't going to be able to reconnect the button pad when taking out the keyboard to get access to the video cards. It's very nice not having to worry about pulling tiny ribbon cables with my Clevo and their heatsink design makes it easier to repaste since you can leave the card in the machine - with my M17x R1 you had to remove the video card and heatsink and that was tedious at best.

I won't deny that the Alienware excelled at keeping those GTX 260s cool though. I don't think I broke 80 under load with them clocked to 280 speeds until the fans started to clog... But I had to clean the fans every 60 days... Definitely a trade off.

Sent from my HTC One_M8 using Tapatalk

 

This is why the P570WM has the best of both worlds: fans beneth the cards, AND they are easily accessible by popping the bottom cover off. The tradeoff of course is that "overkill" desktop CPU, and dual PSU setup. Honestly the only thing that turned me away from it was the dual PSU setup, and the thought that I would lose every ounce of mobility (which turned out to not be important anyway).

 

Yeah I'm kicking myself pretty damn hard for not going for the P570WM. Seriously wish I could redo my decision and get that one instead.

Regardless of what they say, you *NEED* two power adapters if you overclock. Found out the hard way today that my machine will throttle the graphics if I overclock the CPU... good thing that I don't like my CPU running at 89C.

 

Yeah single PSU is only good for CPU overclocking or moderate overclocking on the GPUs (aka stock vBIOS). For serious overclocking you definitely need a dual PS setup.

Just for reference I've seen my 4900MQ draw 120W during XTU benches, so you can imagine what kind of power draw a heavily overclocked CPU+GPUs will be like.

 

I mean in your first pic of the 780m heatsink right on the top left there is a bit of tape that seems to wrap around some fins, I wonder if it is beneficial to remove it...?

 

Good question actually, I never bothered but might be something worth trying. In the meantime I'm going to stick to the 880M heatsink and do some "real-world" testing aka gaming when I get off work and have some time. If the 880M heatsink proves its worth I'll be sticking with it.

 

Dual PSU wouldn't do me any good though, the temps are too high. But yeah I don't see why they don't just jump to a 400 or 500 Watt custom brick. They certainly have more than enough machines that would benefit.

Sent from my HTC One_M8 using Tapatalk

 

The short strips make fine contact, it's only for 4 chips on the top edge of the card. Replacing it with a long one did nothing.

 

^But not detrimental either right?

For some reason the heatsink doesn't have very good contact with those 4 VRM chips on the top edge. I figured a long strip offers more opportunity for contact compared to short strips, assuming they're of the same thickness and the heatsink has the same contact pressure.

 

That sounds like you need to adjust the heights, but JUST having complete contact with all the VRMs is ideal.

 

Do you mean that having the pads just touching the VRM is ideal as opposed to pressing down hard on it?

The master GPU heatink pads show a very faint impression of the VRMs on the top edge, whereas the slave GPU pad shows a much deeper impression.

 

Because they are not balanced for even loading across the card or though the screw mounting positions do bias it a bit.

 

I think what Meaker is saying is that maybe some pads need to be 0.5mm thinner or thicker in order to equalize the height of the heatsink on each part. Uneven pressure and heights can result in some components not making contact at all and the heatsink being slight raised at one end affecting contact on the die.

 

I get that, I guess it's just the way it was worded and the emphasis on "just" made me think less pressure = better (unless that is actually what he meant).

But I may finally have an excuse to splurge on some Fujipoly pads lol

 

I think I read somewhere that the Fujipoly Premium ones were better than the Fujipoly Extreme builder ones because the premium ones are much softer allowing easier equalization of heights. That's the one I would opt for

 

Yes that's exactly what I read as well -- that the one with the second highest thermal conductivity is the one to get. Actually I'll have to settle for 3rd best since they the top 2 don't come in the thickness I need.

So unfortunately it seems the problem is much more serious than I anticipated. It turns out the root issue is that the 780M itself is deformed.

As you can see in the picture below, there is a noticeable downward deformity on the top right corner of the 780M. (top right if you view the card top down with the pins facing downward) The end result is that the gap between the VRM chips and the pad is larger than what it's supposed to be.



Another view along the top edge of the card:



The top left is less affected, but a small gap still exists:



No wonder contact between the VRMs and the heatsink is poor. Doesn't help that the VRM is the hottest running component on the card. Will get some 2.5mm pads and see if contact improves.

 

Increased pressure increases heat transfer.

A smaller gap however is much more important.

 

I bought the Extreme builder version in .5 mm and 1 mm. I haven't opened them yet.

Just ordered some of the Premium version as well. Should be here the end of next week.

 

So I managed to fix the poor VRM contact by using thicker (2.5mm) Fujipoly Premium pads. For some reason Fujipoly pads behave as if they're thicker than the stock ones even though the thickness is supposedly the same. The VRMs leave a very deep impression in the 2.5mm pads, and I'm starting to wonder whether Fujipoly 2mm pads would've worked.

The way I know it worked is by checking the temperature with my finger. No seriously I popped off the back, turned the laptop upside down and ran Unigine Valley for 10 minutes. Now the master 780M heatsink feels just as hot as the slave heatsink. Previously with the poor contact, the area over the VRMs felt much cooler to the touch, to the extent I could leave my finger on there indefinitely and not get burned. Now after a second or two the only thing that registers is "OW!"

 

It's important to get it just right, on my heatsinks the VRMs need 1.5mm pads for instance.

 

That's true, I've noticed if I use the 2.5mm pad on the VRM along the long edge, one of the inductors (the smallest one) now barely get contact with the pad.

 

So long as it contact you are fine though.