M17x R2 full internal cooling mod - huge improvement!

No. The immediate problem is power. The MXM 3.0b standard says that the maximum power draw for a single card is 75W. The 480M requires 100W of power to run, so the laptop just can't provide enough juice. The power requirement could be dropped somewhat by only using 1GB of RAM instead of 2, but I don't think the drop will be quite enough.

After that comes the heat problem, with the card running hotter than most. A big chunk of this is the memory (DDR5), and dropping the memory may help. But it'll probably still run hot enough to cause headaches.

 

that's not true, we don't have power consumption figures, the 100W is a TDP number, which is the amount of heat that must be dissapated.

 

ANNNND... the graphics card does not dissipate ALL the energy it receives as heat, meaning the graphics card could consume more (I think thats right, isnt it?) than 100 watt.

Plus the bigges issue is the 480m has ramchips on both sides of the card, where as the mxm 3.0b standard requires only one side of the card.

If there was some kind of gtx 470m (1gb ram, ram on one side of the card only, lower clockspeed) it would almost certainly fit.

-Ash

 

it obvisully can work in a MXM3.0B slot considering that it's being sold in a MXM3.0b laptop with no visable additional power connector. That would be a collosial error on nvidia's part if they were to design a card that requires a special power source. It's a 100W thermal design power card, Watt can be used to measure heat and energy but the units are not interchangeable. It needs a HSF that can dissapate 100W worth of energy, but that doesn't mean it needs 100W worth of power.

 

It is hard to say how much power the card needs. We don't know whether the +25% TDP (over the 5870M) means +10%, +25% or +50% power draw.

Whatever it is, I think it's safe to agree that it's more than a 5870M. Given that people with CFX 5870Ms and the i7-9x0 CPUs are already running into power-based throttling, a 480M SLI config would probably be even worse.

 

Is this a motherboard limitation, or a power brick limitation? If it's the power brick, I would just string two together in parallel. You've also got to remember that the current 240W power brick trips under extreme overclock conditions, not during any kind of typical use.

 

If there's any limitation it would be in the mainboard, but only in the extremely unlikely event that the 480M's need an additional power source (although there's no information available that suggests this). I can't see it being hard for dell to get their hands on a bigger power brick for their mainboard connector however. It's a 3 moffset mainboard, so it can probally handle 300W, but without the datasheets for the moffets there's no certainty in that.

 

Yeah, we would need 2x 480m's with sli cable and another powerbrick for any kind of testing.

Good luck finding a 480m by itself though.

-Ash

 

Agreed. I think this will be pure speculation for quite some time to come. It might be a cool upgrade a year or so down the road when the 5870s are starting to show their age. And it will be a spendy project for sure. Lozz, do you mean mosfet?

 

Hahah... we are about to see 6xx0 cards be released by ati.

If they manage to sort their drivers out... I think we will be laughing.

I would even hazard a guess that by the time nvidia brings out 480m gpu's to the mainstream, or even 460m to the mainstream, ati will have working silicon of the desktop 6870.

And then a mobility product will only be 2-3 months away.

-Ash

 

Yeah, the desktop world will always be ahead of us because the sky is the limit there. I really like being able to transport my laptop around though. Keeps me from getting too terribly bored at work and stuff. It's worth the trade-off.

 

can i ask if there's anybody with 4870s that did this mod and can share his idle and load temps ?

 

I have not read the whole thread yet but am going to look into this for the 4870s in my R1

 

the 4870's use the exact same HSF, I can't see why it wouldn't give a big difference in temps too. Obvisully you should be expecting 40-50C load, but low 60C's or high 50s sounds resonable.

 

naah, i get low 70s under load.. even with this mod. would be interesting to see a 4870 with low 60s with furmark running so i know i have room for improvement.

 

the serie 4 runs naturaly hotter no mather what you try

 

If anything the 5xxx series should run hotter due to DDR5 memory.

 

Agreed. The GDDR5 is a huge heat source for these cards.

 

So what is the better thermal pad for the memory ? The 0.5MM or the 1.0MM ? Thanks

 

1.0 mm, the one from Dell looks even thicker then 1.

 

I think I'm going to keep the pad on mine rather than use ACS and just do the retention mod + repaste on the core only. Should be enough to drop temps significantly.

 

it should provide some improvement, but using .5mm pads would help alot even. I've looked at the thing and the tension is low, but the main reason is that the pads are just too thick to begin with.

nah, it's 1.0MM, I verified it via a feeler gauge, and the 1.0mm stuff I have is the same thickness.

 

Have you guys tried stretching the dell pads to make them thinner? I may go that route.

 

I would streach them, they're really not that durable and you'll probally end up breaking or tearing them before you get them to desirable thickness. If you want some thinner gap pad material I might be persuaded to send w/e i have left over to someone <_<
but serisously, I managed to ripp the stuff just prying the HSF from the 5870 I got the Bios rom from for you.

 

Yeah the overall problem seems to come from (in order of worst to least worst)

1: Waaaaaay too thick pads.
2: Way too loose tension.

Increasing tension while using the stock pads will just put more pressure on the ram, the gpu die will still be starving for attention.

-Ash

 

So for some reason the MX-3 I have at home seems kinda runny like AS 5. The first batch I had when I lived in Miami was incredibly hard to apply since it was more putty. I wonder if this batch I have now is defective in some way. If I use it on my ram, I'm afraid it'll run. How well has ASC held up? I can probably go grab some from radioshack today and do the mod tonight for the GPU's.

 

Witcher had that problem with MX-3 and found by squeezing out enough to get rid of the runny stuff in the front of the tube helped ... So the back end of the tube should be more paste like...

I stretched the pads on my M15x and ended up with the same results as using the .5mm pads but it takes some patience to stretch them.... So if you dont feel like waiting for new ones it cant hurt to try ...

My machine comes tomorrow and ill try the pads first and see how my temps are and if not ill move over to just using the paste on the Memory... I have a GIANT tube of Arctic Silver Ceramique just sitting here doing nothing so i can put it to use...

 

Think I can get the MX3 to become more putty like if I put it in the refrigerator? If I seal it in a zip lock, condensation shouldn't be a big problem.

 

I used the stuff for a few months on my Mini 10 where .5 mm gap pads were used and it held up just fine. It's definitally good all around stuff.

 

MX-4 isn't as runny.

 

It will just run again when it gets warm.

As quad suggested, you need to squeeze out the separated fluid that has accumulated over time.

What is left will be primarily a ceramic based compound with silicon oil coating each ceramic 'plate'.

Basically, the silicon (or whatever thermal lubricant they have used) has separated from the heat transfer particles and collected in the end of the tube that has been upright.

Just squeeze it out and it will be all good.

-Ash

 

you'll probally want to squish it down a bit after applying, iirc the lubricant is there to make sure it spreads evenly easily.

 

If I squeeze it out, I'll probably lose 3/4th of it. That's how runny it is.

 

Might be time for a new tube then.

-Ash

 

oh wow, yeah, I wouldn't use that..

 

Yeah thats what Witcher/Speedy said about his tube he had to waste a ton of it to get at the good stuff ...

I just stick with OCZ freeze for everything now since i have had great results with it and i have tried them all...

 

I am not so sure 0.5 pads would be a good idea. Those 1mm pads with the mod are barely making contact (I can see how the memory chips are imprinted on the pad when I lift the heat-sink up) so putting something thinner in sounds counterproductive.

 

I think it was either the OP or talon that used .5mm pads and still saw a reduction in temps by about 10C. Haven't done the mod yet and I could be remembering something else though.

edit: yeah, the OP says that removing the gap pads without the washer mod leaves ~ .5mm space between the ram and HSF. With the mod that should give another .1mm reduction and squish .5mm pads just enough.

so you've done it?

 

Yes, see the photoguide thread in my sig.
That's why i am saying it's weird that 0.5mm would be enough as i saw the memory chips barely imprinted on the 1mm pads after the mod was made.

 

I am thinking it is quite possible that the GDDR3 in the 4870s and the GDDR5 in the 5870s are different height chips thus the success with different thickness thermal pads. Can anyone verify this?

 

I think it's the material you're using, that's EK water block gap pad material right? I used the same stuff and it hardly left an imprint at all without the retention mod, even though contact was good. It's just less 'spungy' than the dell stuff. It's probally more dense of a material so it would need more tension to compress it.

they should be the exact same height. They both use samsung modules afaik.

 

yeah, that's the one.
hmm.. you might be right.. will try with the 0.5mm and see if it improved the temps.

Not sure about the memory height, even though they are both samsung ddr3/5 might have different height ?

 

Let me know what you find out about .5mm vs 1mm mfractal. I am looking to order a sheet and want to get what will work best.

Thanks in advance!

 

will do.
probably will try this today.

 

1mm pads were definately too high for me - perhaps its the way the chips are soldered down?

Or... I think I know what it is.

Its more than likely the GPU die area on the 4870m has some kind of shim or is raised higher up than the 5870m.

...aaaaaand there is. Just a quick google search shows that the HD4870m has a metal shim around the GPU die - this should actually lift the whole thing up a little, and will prevent the heatsink from bending down as much to meet the RAM chips with the washer mod.


It looks like the mod for the HD4870m will be indeed different to the HD5870m.

Maybe this is why the HD5870m heatsink doesnt fit right - because it was made for the HD4870m?

-Ash

 

It is quite possible that the sink was designed for the 4870 and retrofitted to the 5870

Although it was really designed for the GTX 280m's

 

I took the plunge and did the full clip and GPU regrease mod.

I used MX-3 on the GPU and AS Ceramique on the RAM chips (hopefully, that is OK stuff). The Ceramique was very thick and I put just a dollop on each chip.

Before the mod, my max temp (RAM?) using GPU-Z and running FURMARK @ 640x480 (Stability) was 89C.

After the mods, my temps running Furmark with same settings for about 8 minutes were 70C/75C/74.5C.

I was hoping for a bit lower. Maybe my retaining clip screws need to be tightened a bit more? I turned the screws only 2-3 times, each. I inspected the 5870 board and there was a slight bend, so I backed one of the screws off and that brought it mostly straight.

Or...is that about right for having used Ceramique?

Anyways, I'd hoped for more examples of people doing the mod and before/after temps, but seems like most here talk the talk, but don't walk the walk!

 

there shouldn't be any difference between ASC and MX3, the ram chipset indivually don't produce enough heat to warrant it. It's probaly more an ambient temp and notebook cooler providing the extra difference.

 

Hmm... those temps are about 10c higher than mine.

Dont tighten any further, its the right tightness.

How much MX3 did you use? I find that MX3 is difficult to use for GPU cores because of the thickness and difficulty spreading it.

Also, what method did you use to apply the compound?

-Ash

 

I've used MX-3 before, typically for CPU's. My MX-3 is not as thick as Ceramique, and is more like AS 5. No trouble at all spreading it (which I typically use a new utility knife blade and spread a very even, thin coat across the face of the die of the CPU or GPU. I did the same on the 5870 GPU...spread a thin coat evenly across the face of the GPU die.

One thing, too, is the ambient temp is higher than normal today. We have our AC on, but the room temp is still around 80F.

Should I redo the GPU junk? I'm sure all that white Ceramique is going to be a pain to clean up and so I don't really want to remove it. I'll probably never push it that hard anyways, but as we all know, sometimes it's all about the numbers!