GX640 Cooling Mods?

laptop modding complete. drilled 5 holes over the fan, and removed the extra plastic on the cover at the gpu area (and the copper shim upgrade ofc). still hitting some bad temps tho, 3 minutes @ furmark running on 760/1050 got it up to 94 degree Celsius (idle @ 58). i think i might have added too much paste, or maybe left a fingerprint on the copper shim. ocz freeze is one of the best cooling paste, so i don't think that is the issue

 

Ah, the smell of freshly drilled plastic and lapped copper.

 

yeap... its prob the combination of the fingerprint and a bad paste job. id clean the shim and redo it. Ive drilled holes and taken out the thermal pads and in gaming i dont reach 85oC often, in reality my average temps in bc2 or sc2 is 80-81oC. 10 mins of furmark gives me an average of 89oC while maxing out at 91oC.

 

I dont use the shim even tho I bought one. Its the GDDR5 RAM thats the problem. Replace the pads for past.

Also I used MX-3 and it is tested to be even better than ocz!

I also replaced the thermal pads for thermal past and get about the same temps lol! Exept I can reach 82c in bc2 now and 74c in assasins creed 2 after like 3-4hrs.(@860/1080)
Both games were maxed all the way to.

I also added a 3rd fan for the laptop cooler and some heat sinks for the gpu.

 

860/1080? how do you get the core up so much?

 

I'll just leave this right here. http://forum.notebookreview.com/sager-clevo/416414-clevo-m570tu-cooling-overhaul.html

 

My temps can handle it and if I want to risk overvolting. I would push higher on the GDDR5 RAM. I read that someone here got there GX640 @ 860/1100! I get the "gray with vertical lines screen" if I go any higher than 1080 for the GDDR5 RAM.
Im fine with that, so its not worth the risk lol! (small difference)

 

Does tampering with the bottom cover void warranty?

 

It shudnt do. MSI are a bit more lenient compared to dell or hp.

 

Soviet Sunrise this is one bloody sick mod! +rep added.

 

argh. its obvious that i suck at modding have had 5 attempts, two last attempts was with removing the vram pads and the copper shim. i now use a crappy cooling paste i think (cooler master) but that should not push me up in 96 degree on stock speeds. could it be that the default vbios and bios has this much to say on temps?

 

I'll just leave this here...

 

That cant be from the GX640, but great temps?

What was the room temp?
How long was it on or did you just turn it on?
Whats the GPU in there?

 

Hehe yeah I just came home, turned it on. Room temp was 21C

 

haha!
I thought so...
If it wasnt, I would think you used liquid nitrogen or water cooling mod!
lol

Was this your GX740 or a different one?

 

Yep the E7405

 

What do you guys have under your fan?

I have removed some plastic under mine because my keyboard is directly over it. Now it intakes a lot more from the keyboard and my temps are actually OK with bottom vents totally covered.

 

I know it's a GX660R, but there is never enough coppah.

Or holes.

 

forge, with those extra coppers, you are still able to close the lid without any problem on it?

am interested with those small copper, where did you get it from?

 

I'll just leave this here...

GX640:

 

I dont use them, but here:
DDR DDR2 RAM Cooler Heat Sink Heatsink f RAM VGA RHS-03 - eBay (item 250691673062 end time Sep-07-10 21:19:54 PDT)

 

Hey Gamer, for OC'ing the memory does it go unstable if it's at 1.05? For example my core can be 850 but my memory can only go up to 1060 before it crashes in Dirt 2. I'm running at 1.05V and I heard you're running at 1.15V?

Right now it's about 78C in games and if I go to 1.15 again I'm thinking it's just gonna heat up again.

 

1.15v is correct! I did not change it my self. It came that way with the updated new vbios.(from msi website)

You can see the 1.15v in here:
YouTube - BF:BC2 MAXED-OUT all the settings - MSI GX640
This was before I changed it to 860/1080 @ 1.15v and get 81c after 3hrs.(still stable in games)

I think you should get the new vbios or change your volts to 1.15v.
But you have the GX740, so Im not sure about your vbios...

 

lol the new vBios changed it to 1.05V

 

I didnt change the volts myself, so it must have been with the vbios update.
Is changing the volts easy? How do you do it?

I might consider raising the volts...

 

It has to be done vBIOS I think.

but changing volts cause it to heat up more (Power = Voltage x Current)

I was up to 90C with 1.15 and 80C with 1.05V

 

Yea same here buddy, laptops cant cool components too well so cranking up voltage too much isnt a good idea. Unless you have done some SERIOUS mods it should be avoided. Also there isnt a huge difference between 850 and 750 core anyways.

 

My best mod was to take the thermal pads off the GDDR5 RAM and use MX-3 thermal paste instead. MX-3 on the gpu die aswell. I did not skimp on the thermal paste, because it is non conductive.

That, a few heatsinks on the gpu, and with my 3 usb powered fans, it rarely gets past 81c(74c on most other games) after hrs of BF:BC2(maxed). (@ 860/1080 @ 1.15v)

What mods did you use?
Did you have a laptop cooler and heatsinks on when you reach 90c with 1.15v?

 

I do have a bit of room in my temps for more, but maybe its not worth the risk.
Your right that its a small difference... 750-850 (I think its only about 5fps difference in bf:bc2)

 

Right now it's at 840/1080 @ 1.05V, reaches 79C with bottom cover off and everything else stock.

I think that's my sweet spot I'll just keep it there.

 

yea avg fps is already 40 neways for bc2 so not worth it for the heat output. good for benching tho

 

now i have aquired mx-3, and aplyed it on both gpu ram, gpu and cpu. i cleaned off the old cooling paste with condence cleaner, and added a paper-thin layer of mx-3 on gpu/cpu. now i want to cry, as i rice to 100 celcius on gpu with 5 minutes of furmark exreme burning mode(gpu @760/1050) is there some set time on this mx-3? what the f*** did i do wrong now? ah truly sadface here. is there some set time on the paste? did i f*** upp with drilling the holes? WWHHHHAAAAT? oh god i hate this computer by now i stil have not upgraded default vbios ec or bios tho, and ati powerplay is disabled. anyone got an idea of what might do the damage here?

 

Which part of the GPU (the core or vRAM) is showing the high temps?

As I understand you replaced the thermal pads on the vRAM chips with paste. Correct? Did you keep the pads on VRMs (the square metal thingies next to vRAM chips)?

There's a good chance that the heat sink is not making (good) contact with either the vRAM, the core or both. You should remove the heat sink to make sure whether there is contact or not. If not you need to apply a thicker layer of paste.

The heatsink is designed in such a way that even if you remove all the thermal pads (from the GPU side) and it touches the core, there remains quite a sizable gap between it and the vRAMs and VRMs. OTOH the stock thermal pads are so thick that they push the heat sink away from the vcore. The solution would be to either
- do the copper shim mod
- replace the thermal pads with thinner ones that allow the vcore to make contact with the heat sink
- or remove the thermal pads altogether and replace them with a rather generous amount of thermal paste

Hope this helps

 

yea, i kept the yellow pads that covers the VRMs, are those supposed to go away also? the vram pads is ruined now, but i have plenty mx3 left. so should i remove the yellow pads and add mx-3 there as well?

core temp is 100c, but how can i see vram temps?

 

Not necessarily. But removing them should help reduce the gap between the heat sink and vRAMs a little which would help reduce the amount of thermal paste required to fill the gap. I kept them at first but removed them since and this helped me reduce temperatures further by 1-2 degrees. I'm now idling at a cool 52-53 deg C on vcore and vRAM (~22 deg ambient) without any cooling pads or modifications to the bottom panel. I can't tell you the max temps as the last time I ran FurMark the memory appeared not to clock all the way up to 1 GHz and therefore the temperatures I got are probably quite a bit lower than they would be at full clocks...

As for your temp values. HWMonitor only reports the highest of the three sensors' (vcore, MemIO and shader) readings, which is likely the vRAM. To get all three values I would recommend downloading GPU-Z which shows the values for each of these sensors (with #1, #2 and #3 being vcore, MemIO and shader respectively) individually.

 

all pads removed now, and it seems to have helped a bit. now the core is warmer than the rest, and i get the feeling i did not ann enough paste there. can this be the case? its enough to not see the metal on the core, but not more than that.

now i have the old bios, ec and vbios. would it help the temps to get the updated ones?

if not, then there has to be something wrong with the heatsink or the fan itself. this time i did the operation so delicate and clean there just CAN'T be any flaws!(i hope )

 

Those temps look quite OK to me. Especially considering that you're running at stock clocks. The core is probably fine. You really don't need to put a lot of paste there since it should make quite good contact with the heat sink. The higher core temperature compared to the other parts might just be an effect of the high clocks. Since the core is basically the only part getting any load whatsoever during idling, it would make sense for it to run a little hotter under these circumstances. I think... Of course the only way to make sure is to run a stress test - if you dare to after the temps you were originally reporting

I don't have any experience with updating the BIOS or EC software but they shouldn't have any effect on the temperatures anyway. Getting the updated vBIOS on the other hand would definitely help lower the temps and I would definitely recommend you to do that.
As far as I know the threshold for kicking up the fan speed is typically around 55 deg C (assuming of course that the CPU is also running at a cool temperature). So if your idle temps are lower than that the fan should stay at the lowest rpm setting making the computer much quieter.
Another option would be to create an "idle" profile in CCC with lower frequencies (e.g. 200/300 MHz core/vRAM) for regular use. Especially if you're feeling uneasy about fiddling with the vBIOS.
The two are of course not mutually exclusive...

 

I had a horrible experience with MX3. Did you try removing the heatsink to see how well it spread. Mine would just clump up into little blotches here and there, but it looks like I got a bad batch which I read was a problem others were having too. Maybe you got one of the bad ones too. Biggest waste of 20 bucks imo. Just use OCZ Freeze. Best paste I've tried.

 

Check this out folks. Ghetto works beautifully! There is a small gap when I forced the back shut, but I see almost a 15 C drop in temps with furmark! I am going to buy a few more RAM-sinks, this time a bit low profile and touch up every bit of Cu real-estate. Before = 95 After =81

The thin stock plate is pathetic.

 

A 15C drop just from adding those? Not bad. I have some that I got from Dealextreme for like 3 bucks that I might try on mine.

 

impressive, what clocks?

 

7 fps?
Try w/o multisampling.
What catalyst? ^^

 

Stock 625/1000 clocks. I will try 700/1000 and see what happens. How high can I go?

I turned on multi-sampling as I figured it would make the burn-in more GPU bound than CPU bound. Mobility Catalyst 10.8. (10.11 is garbage).

 

You can go a lot higher for core since it will be less than memory temps anyways.

 

Yeah, I think with the core people have been able to go above 800, 750 should be attainable for most cards I think! The memory is trickier an you'll be quite lucky if you can get anywhere near 1100 MHz. Just go in small steps (5-10 MHz) and check for stability every step of the way.

 

finally i did the bios update, and the vbios update. now i idle @ 48-49 core, and 45 shader 45 vram in 22 degree room temp. have ordered 80 pieces of ram heat sinks also, so ill be god for my next 3-4 laptops will furmark this and see how high i go tnx for all the help, and btw. the MSI Forum HQ USB flashing tool was the easiest tool ever to flash both bios and vbios

edit*
ah, this seems alright. i have removed all the pads under the gpu, and drilled some holes as my current modifications. should the result be better? im looking forward to getting the ram heat sinks at lest


will try som oc now. what temp should i stop at, do you guys think?=)

edit 2*
vram seems a tad hot, but i know i used more paste than i need to. should i worry about this anyway?


i think i'll stay at 760/1000 till i get the ram heat sinks

 

Nice... i exactly the same temps as you but at 750/1000 (should'nt be much of a difference because my voltage is lower).

 

Here at 750/1000:

VGA BIOS has NOT been updated. 750/1000

 

use imgur.com or min.us please...... imageshack is terrible.. have to click twice and it's still not full screen image.

 

nice. try the tool i mentioned earlier, it was very easy you should drop down a bit on the temp with that