Ashtefere

Alienware M18x GPU Cooling Design

37 posts in this topic

Documentation

Take a look at those heatpipes! *phew!*

3 thick heapipes per GPU. This thing can easily take a couple of 485m's.

The questions remain:

  • Power Circuitry? is it going to be enough? (probably)
  • Bios whitelisting? Will the bios block non-approved cards? (also, probably)

I am moving in August, and my job contract ends in July, so I will be buying my m18x in sept-oct this year. If by then they still dont have the 485m sli as a selection, my plan is to:

  1. Buy the base config with 460m sli from USA
  2. Buy the 485m sli pair from logicalblueone (local supplier) and MAKE IT WORK
  3. ???
  4. Profit!

Im sure someone will want to buy an SLI pair of 460m's somewhere on ebay if I can get it to work, but overall, I'm excited and hopeful!

I was in doubt about the 485m's until I saw the cooling. It looks monstrous!

As soon as I know everything is all working, I am going to proceed to see if I can do any cooling improvements to this behemoth. It will be interesting to see if they learned from all our work on the M17x-R2.

-Ash

2 people like this

Share this post


Link to post
Share on other sites

Yeah I told ya, the cooling is much improved this generation. The R3 has a similar 3 heatpipe design except it seems with the added ducting and plate over the fans, the M18x will have better airflow. One thing I wish they hadn't done is put the GPU's in the middle again like they did with the R2, that is extremely annoying when you need to repaste. Since it can handle a 330W PSU, I'd surmise that the power circuitry is more than adequate for 485M SLi.

Share this post


Link to post
Share on other sites

The CPU heatsink is placed close to the exhaust and back edge of computer unlike the M17xR2 having this in the center of the system. That's an improvement. The extra copper pipes are going to be excellent at keeping the systems cooler.

Share this post


Link to post
Share on other sites
Two threads, same topic and title?

One is CPU and the other is GPU :)

Share this post


Link to post
Share on other sites

To be honest I didn't notice it until you mentioned it...

Share this post


Link to post
Share on other sites

GPU cooling on this model seems great, I think the problem is the CPU cooling.

Share this post


Link to post
Share on other sites

After pasting my memory chips I noticed they weren't quite touching the heatsink (dell 6970 in m17x, but essentially same as 18x for area over memory).

That said, temps dropped massively vs. pads....even without proper contact.

I was wondering how bad of an idea it'd be if I picked up a thin sheet of copper and affixed it to the heatsink over the memory chips.

1. to close the gap

2. I'd use silver, but it's super expensive. Copper's a decent conductor.

Would really like some input before I do something potentially (more) retarded (than usual).

Edited by kune

Share this post


Link to post
Share on other sites

The problem is that you can't properly attach the copper shin to the heatsink, there is always some air between the copper and the heatsink, which is really insulating.

Why did you remove the thermal pads? And which temps dropped? Usually the gap between the memory and the heatsink is about 0.5-1.0mm, and this too big for a proper use of thermal paste. Pads are the best way to go imo, you only have to be careful that the pads are compressible and not too thick, because this will increase the gap between the die and the heatsink.

Share this post


Link to post
Share on other sites

Because I had an extra GPU to mess around with.

Memory temps are what dropped, and they dropped by a good 20C ++ vs. the terribad Dell stock pad job. I saw memory temps around 100C before.

10minxtremeburn.th.png

Thanks for adding some logic to my bad heatsink idea lol

Share this post


Link to post
Share on other sites

Looking at pictures I would also say that the new types of fans they are using in the R3 and M18x R1 are great too. For quieter and better air flow.

I know I would sure like to have those fans in my R2 let alone the new cooling design. :Banane52:

Cheers. :)

Share this post


Link to post
Share on other sites

Check this out guys... (these are words copied directly from M18x manual - so I don't own these words they are property of Dell+Alienware)

AC Adapter

Output current

240 W

12.30 A (continuous)

330 W

16.92 A

Output power

240 W

240 W

330 W

330 W

Rated output voltage

19.50 VDC

Temperature range

Operating

240 W

0 °C to 40 °C (32 °F to 104 °F)

330 W

10 °C to 35 °C (50 °F to 95 °F)

Storage

–40 °C to 70 °C (–40 °F to 158 °F)

Connector types

DC connector

3-pin, 7.4 mm plug

AC connector

3-pin — C13 (150 W)

3-pin — C13 (240 W)

Now if anyone remembers my overclocked system that I used a custom air cooled design on my power-supply? I had temps atleast up to 120 degrees Fahrenheit. So that means the temps were well above the safe rating. If we overclock on that 330W I think it is important for the life of the power-supply to do the same and add a cooler to it if overclocking to maintain lower than the rated 95 degrees Fahrenheit. This may not be news but to me that is.

http://forum.techinferno.com/general-notebook-discussions/215-laptop-ac-dc-power-adapter-cooler.html

1 person likes this

Share this post


Link to post
Share on other sites

So did anyone end up making custom silver or copper shims to displace the extre heat? Did you go back to using the pads? Curious to hear what worked best.

Good Luck! StevenX

Share this post


Link to post
Share on other sites
The problem is that you can't properly attach the copper shin to the heatsink, there is always some air between the copper and the heatsink, which is really insulating.

Why did you remove the thermal pads? And which temps dropped? Usually the gap between the memory and the heatsink is about 0.5-1.0mm, and this too big for a proper use of thermal paste. Pads are the best way to go imo, you only have to be careful that the pads are compressible and not too thick, because this will increase the gap between the die and the heatsink.

I did the same, - pasted the memory chips with TX-4. Had to apply a very thick layer on the chips but it was well worth it - 10-20C drop across the board. Pads are no longer a good solution for GDDR5 the way I see it.

1 person likes this

Share this post


Link to post
Share on other sites

Yes, if the gap between the die and the heatsink can get minimized by removing the thermal pads, the temps can really improve.

Recently I got to the conclusion, that the vram modules aren't running that hot... the modules on the back of the 470m which I put in the M15x weren't cooled at all by a heatsink or similar, and it worked fine.

I realized that in the G73JW the gap between heatsink and vram modules was pretty huge and Asus filled it with a chewing-gum like paste, and the system worked fine too. I guess the vram modules don't need as much cooling as generally assumed... I should try measuring the temps of the modules under load sometime.

Share this post


Link to post
Share on other sites

It is virtually impossible for a VRAM module to overheat, even when bare naked. A VRM module though, can but this is very rare. Two very different circumstances separated by just 1 letter.

Toms hardware guide get this mixed up way too fucking much. So do a lot of online hardware stores.

-Ash

Edited by Ashtefere
2 people like this

Share this post


Link to post
Share on other sites

I did it with my R2 here i know is on different forum but i hope OPs dont cut out the knowledge.

M17x-R2 CFX 5870 modding for extreme oc

After pasting my memory chips I noticed they weren't quite touching the heatsink (dell 6970 in m17x, but essentially same as 18x for area over memory).

That said, temps dropped massively vs. pads....even without proper contact.

I was wondering how bad of an idea it'd be if I picked up a thin sheet of copper and affixed it to the heatsink over the memory chips.

1. to close the gap

2. I'd use silver, but it's super expensive. Copper's a decent conductor.

Would really like some input before I do something potentially (more) retarded (than usual).

Edited by svl7
link was broken, fixed it for you ;)

Share this post


Link to post
Share on other sites

So not a bad idea at all using aluminum and paste on both sides and not a bad idea on using just paste so how bout just using a thick paste on the vram and one that has long reliability that doesn't bake out like the IC Diamond in the test/giveaway... Its a paste that doesn't exactly melt like some others so you may be able to add extra for a thicker application over the vram chips so it contacts the chips and the normal heatsink so no extra material from a shim would be needed and would prevent a small amount of heat trapped by using the shim since the shim isn't "soldered".

Share this post


Link to post
Share on other sites

Aye! U R right . Cooling the vram is not pressing isue it mostly stand for 1300Mhz EZ without cooling. I have not checked but i sore they had some reference standard up to that XXX it must work stable. The pressing issue is the GPU cooling m17x GPU cooling is relatively not bad but is lack of proper design. Since the highest gpu is 5870 for m17x dell had no interest to do anything. R3 housing a single 6970M performs similar than 2 5870 not really point to go for single.

Share this post


Link to post
Share on other sites
Aye! U R right . Cooling the vram is not pressing isue it mostly stand for 1300Mhz EZ without cooling. I have not checked but i sore they had some reference standard up to that XXX it must work stable. The pressing issue is the GPU cooling m17x GPU cooling is relatively not bad but is lack of proper design. Since the highest gpu is 5870 for m17x dell had no interest to do anything. R3 housing a single 6970M performs similar than 2 5870 not really point to go for single.

there is a point to get a single 6970m the R3 M17x having one gpu you don't have to deal with crossfire problems (there aren't really any but microstutter and that's not a big deal), its lighter than the R2 and supports the 2nd generation I7 chipset, faster and more amounts of ram and sata 3 hard drive support. So getting an R3 is a nice powerful portable package that doesn't have crossfire issues. R2 has the RGB display and fully unlocked tdp on CPU with xm first generation I7 chips.

Edited by mw86
changed data 3 to sata 3

Share this post


Link to post
Share on other sites

I know the tech works behind R3, i mean for me is no point only i find one attractive element 16GB of ram when you dealing with Hyper-V or vmware it could come handy. Personally i much prefer dual grafics (remembe my 1st one 2xVoodoo2 24MB from Diamond) soon we going to hit 4gb on a single card.

Share this post


Link to post
Share on other sites

There are advantaged on both side. Who goes for what is a personal issue (preference)... LOL

Share this post


Link to post
Share on other sites

Merged into a new thread/topic, I encouraged the discussion I apologize... that was my first move of a bunch of posts I see how it helps organize the forum. Back on topic I will be using the IC diamond on my GPUs and CPU on my incoming M18x it will be the second one I repaste with it so I get a little practice before hand. I believe I am going to try a thicker layer of IC Diamond over the various places Dell has use pads on the video cards to see if all temps can be dropped on the GPUs.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now

  • Similar Content

    • m18xR1 New 105 UNLOCKED bios, with v11.2, 11.6, 12.9.0 OpRom with RAID0 TRIM SUPPORT!!!
      By Swick1981
      Hey Everyone,
      Well, I was FINALLY able to MOD a fully UNLOCKED A05 Bios for the m18x R1. 
      I created several BIOS's that have the following Intel Raid OpRom v10.5, 11.2, 11.6, 12.9, 13.5.  I'm told that 12.9.0 is the BEST version to use with our m18xR1.  Also, these ROMS are modded to give TRIM on a RAID0 system!  I've tested it and can confirm it does work!
      I also have an intel ME Firmware upgrade that you can do, as well, but I only sugest people that know what they are doing use it.  Only because, for some reason, if you flash the ME firmware, then the Bios, there is a problem and you can brick your ME Firmware.  (You'd need a new motherboard).  It locks your firmware. 
      Anyway, if you flash the bios that you want, and not even deal with upgrading the ME, then you're fine.  But, if you do the ME firmware upgrade like I did, if you flash and bios's, you will need to downgrade the FW, before any BIOS flashing...  (I THINK).
      In any case, you may just want to stick with the BIOS upgrade, as this will give you TRIM in RAID0 support (confirmed with trimtool).
      If interested, please let me know!
      Thanks,
      Swick
    • Is your old Celeron cooler ready for the dumpster?
      By timohour
      All started when my brother's ASUS H61 motherboard died and I got his Celeron G540 (socket 1155). I also had an mITX case with a dead custom PSU that I don't intend to fix (Hantol HC ITX02).

      He now needs another PC in his bookshop, so, I decided to buy a cheap mITX 1155 socket motherboard that would use an External power brick. And I found a great deal on a Lenovo m91p motherboard @ 22$ plus shipping. Or so I thought. This motherboard uses non reference cooler holes making it impossible to fit the stock Intel cooler (or any other cooler that I had laying around and would fit the case).
       

       
      So I decided to use the old cooler from a Celeron D 330 (2.66GHz) a friend gave me after his motherboard died...
       


      It seems to have a big amount of aluminum in the middle and have aluminum fins... It just needed some trimming on the side cause there are some capacitors.
       

       
      I kind of hooked it on the motherboard using an 14" fan protector I got from a dead Thermaltake and some M3 screws and nuts.
       

       

       


      Add  some MX2 the old stock Intel cooler on top of it and the results where ~ 26C @ idle and ~52C @ full load (using stress on Fedora 22). Pretty solid!
       

       
      I then decided to add an silent 92mm Scythe fan and the results where even better ~20-21C @ idle and ~ 45C @ full load...
       

       

       


      I lastly tried to use it passively, after an hour idle temperature was merely @ 66C and It took 30min of full load for the temperature to reach 100C.
       
      This means that I can probably run it with one or two case fans only. This will be tested.
       
      The results show that the old Celeron's D (73W rated TDP) had some pretty solid cooling solutions, that you may want to mod and use for cheap projects like this.
      I don't have an 1155 4C to test, but i am pretty sure that this cooler would outperform even the Copper Intel stock cooler shipped with these CPUs.
       
      More pictures with subs...
      http://imgur.com/a/DbOLW
    • Anyone able to spare some M18x time to unbrick a 7970m for me?
      By Blackyhimself
      Hi guys,
       
      I've managed to brick my graphics card trying to install an undervolted vbios. I'm running an M15x and so can't slot in an extra card to run the screen in order to flash another vbios.
       
      I'd be happy to pay to cover anyone's time.
       
      Thanks
    • Installing Clevo GTX 680M in Alienware M18x
      By Brian
      Officially, the upgrade path for the Alienware M18x-R1 ends with the nVidia GTX 580M but is that really going to stop the Tech|Inferno enthusiast from trying the latest and greatest? Absolutely not! So let’s get started.
      Parts & Tools Needed
      Quantity                                                                                                           Part Needed (1) or (2) Clevo GTX 680M Kit from RJTech (1) X-Bracket (also included with 680M kit when purchased from RJTech) (1)Screws Screws (Also included with 680M kit from RJTech) (1)680M SLI cable SLI Cable (optional, can be found on eBay or directly from Dell) (1)IC Diamond ICD7 Diamond(recommended) or MX-4 Thermal Paste (1)Thermal pads Aftermarket Thermal Pads (1) ESD Mat + Strap (1) Precision Screwdriver Set (1) Techni-Tool Spudger  Clevo GTX 680M Upgrade Kit
      When you purchase the Clevo GTX 680M Upgrade Kit from places like  RJTech or MXM-Upgrade, they usually send a full kit that includes the X-bracket, screws, thermal paste, and Clevo heatsink. For the M18x-R1 or R2, the Clevo heatsink will be useless so you can elect to either resell it on eBay or just toss it in the garbage. A point of importance to keep in mind is that whether you opt to purchase the GTX 680M by itself or as a kit, be sure it comes with the x-bracket and Clevo screws (more information on this below). Typical price for an upgrade kit as of September 2012 is around $750-$850.00 USD.
      M18x Disassembly
      The Alienware M18x-R1 is a fairly sophisticated system that requires the user to have a familiarity with taking apart notebooks. Because some of the parts in the M18x are fragile, it is recommended you only attempt this if you feel you have sufficient experience.
      Refer to the Alienware M18x Service Manual to learn how to take apart the system. View Alienware’s Tear Down videos to see how its done. Installation
      Installation is pretty straightforward, be sure to use the ESD mat + strap when handling the M18x and GTX 680M so that no damage occurs to the sensitive electronics.
      Once the M18x has been opened up, remove the existing video cards. Ensure that you are using GTX 560M/580M video card heatsinks for full compatibility with the GTX 680M. You must use the screws (linked in the above table) designed for the Clevo GTX 680M X-Bracket since the Dell screws do not fit. Remove the retention washers from the 560M/580M heatsink in order to remove the original screws. Clean the previous thermal paste from the GPU using isopropyl alcohol and a cotton pad. Alternatively, coffee filter paper also works well. Place the thermal pads on the video card according to the picture below:  One problem you may encounter while fitting new pads is that the heatsink does not make full contact with the inductors on the video card. To alleviate this, a thicker aftermarket thermal pad can be used or copper shims + ICD7 such as the photo below:  
      Once the thermal pads and/or shim have been placed, do a test placement with the heatsink to ensure all the thermal pads make contact with the heatsink. An example of proper contact is illustrated in the photos below: After proper pad contact has been determined, place a thin line of thermal paste across the GPU, seat the heatsink and fasten it with the screws. Both the right and left heatsinks require different fitting of the pads so always check for proper contact before installation.   Driver .inf Modification
      The Clevo GTX 680M may require the proper device id to be added to the nvcv.inf file found in the Display.Driver directory.
       
      Credit: jhmonroy
      Instructions
      The latest beta drivers require more extensive modification that go beyond the scope of this article at this time. However, there are modified nVidia beta drivers available at LaptopVideo2Go that have the .inf files already modified for both the desktop and notebook and thus are a simple point and click installation–which is what we recommend.
       
      Update 9/13/12: You can grab the latest 306.23 modded drivers from our forum to use with the M18x-R1 and Clevo GTX 680M.
        Clevo GTX680M Gallery
           













    • GPU fan speed under OSX
      By thekobi
      Hi,
      how can I control the fan speed of my GPU under OSX?
      At the moment my GTX 970 is always running on the same fan speed. Is this normal? How can I monitor my temperature under OS X? Under Windows the fan speed changes automatically.
      Thanks a lot for your help guys!