Page 1 of 2 12 LastLast
Results 1 to 25 of 42

Thread: Proper Quadcore water block mounting technique

  1. #1
    Xtreme Member
    Join Date
    Jan 2007
    Posts
    209

    Proper Quadcore water block mounting technique

    Ok I have done 25 mounts and remounts in the last 12 hours straight, trying to figure out a consistent way of mounting the waterblock onto a quad core. Since the quad core has a non centralized heat spreader system we need to figure out which quadrants of the heat spreader cool which core. I have been studying the distribution of thermal paste on the cpu and water block after mounting and figured out a map of the cores. Since the new quad core cpu water blocks have a convex shape base the pressure is good in the middle but horrible to the sides. Our job is to balance the pressure of the cpu water block on the quad core cpu IMHS, just like balancing a titter totter.
    You can balance the cooling load to all four corners of the quad core by using the four tension screws on each corner of the cpu block. Using a tool like core temp beta in windows you can see which core is the coolest and which one is the hottest. Your goal is to get an even temp across all four cores. Here is a map of the squares on a Quad core CPU:




    Notice that pin one is on the upperleft hand side to orient yourself(I put three pics to show you the different angles you can see how to figure out the squares just pay attention to pin one). You will have to look at your motherboard and see where pin one is and then mark the approiate tension screw for that particular square you want to apply more pressure to. So square one will influence core 1 so turn the tension screw you labled as square one to either increase the load(lowers the temp on core 1) on that square or decrease the load(increases temp on core 1) to that square.

    The picture of above shows the 4 squares of where the cpu heat spreaders must lie. These squares do not signify that there is a core under that square and the quadrants dont have definite lines as shown above. What the squares do show is when more pressure is applied in square 3 the temps of C3 go down and when square 0 is applied with more pressure C0 goes down in temps.

    Lets do an Example:
    In this screen shot you can see all four cores are not even in temps: C0 and C2 are higher in temps which indicates squares 0 and square 2 are looser in screw tension. The screen shot also shows C3 and C1 are cooler so this means they have more screw tension and can be loosened if necessary to give pressure back to C0 and C2 tension screws.

    To remedy this screen shot scenerio what you do is losen the screw closest to square 3 and square 1 and tighten the tension screws for square 0 and square 2. Shut down your computer and turn the screws a 1/4turn. Reboot and check your temps again. Keep making small adjustments till your temps even out as shown below.





    If you were to take cpu water block off the cpu you would see this kind of thermal paste distribution for the screen shot above. See how the square 0 and square 2 have collected unused thermal paste and thats why you want to tighten the screws down on those areas.



    After doing the technique above you will see this type of heat distribution, notice the overall temp of the chip went down in nvmonitor and the cores are all nearly even.

    Idle temps:


    Loaded temps:


    The pic above shows all the cores are the same except for C3 in idle state. This is because the base of the cooling block and the shape of the cpu heat sreaders favors pressure for that core. I could loosen the screw in square 3 to raise the temps to match the others but what for? the other three cores are as low in temp and as even in pressure as I can get due to the fact the tension screws dont have much more lee room for tightening. The overall shape of your cpu block and cpu heat spreader will determine your limits. Also by initially releasing some of the pressure from square three it allowed me to tighten down the tension in square 0 or lower right quadrant for more even pressure so that is why the overall temps went down even though I loosened the screw to square three.

    A note about the screen shots, the first screen shot was taken at 26C ambient room temperature then after the proper waterblock mounting procedures were done the screen shot was taken at 25C ambient temperature but still you can see a drastic decrease in properly mounting the water block on the cpu.

    To test my theory I have unmounted and remounted successfully with same results 4 times in a row and when looking at the thermal compound it was evenly distributed with more compound squirted off the sides of the chip showing good pressure. Without the procedure and just tightening all the bolts as far as they can go I could only reproduce these results randomly one out of 8 tries.

    I could not get consistent results using the springs, All testing and procedures were more successful with hard mounting. Make sure to install a back plate to your mother board or else results will not be the same due to bowing of the motherboard.

    http://www.xtremesystems.org/forums/...79#post1972379
    Last edited by redcorn; 02-02-2007 at 11:58 PM.

  2. #2
    Xtreme Addict
    Join Date
    Jul 2006
    Location
    Vancouver, BC
    Posts
    2,061
    Did this improve load temps in any way? Are they more even across cores as well? Is the average core temp lower overall at load?

  3. #3
    Xtreme Member
    Join Date
    Apr 2005
    Posts
    372
    just set infiniti to certain cores via task manager, and get a couple of instances of Orthos, prime, and Super Pi running it will peg them all out.
    Core 2 Duo E8400 @ 4ghz
    2GB PC2-6400 @ 445/890 4, 4, 4, 12
    ATI HD4850 @ stock
    Gigabyte P35-DS3L
    El Cheapo 500w PSU with dual 12v rails
    16X DVD+/- R/W

  4. #4
    OC Jedi (on stand-by)
    Join Date
    Jan 2005
    Location
    Germany
    Posts
    5,576
    I didnt read all posts in detail. Will do later, but a quadcore doesnt have a core in each quarter of the CPU.
    Have a look at one of the pics google finds here:
    http://images.google.com/images?svnu...00&btnG=Search

    It becomes pretty obvious which parts are the hottest, so getting a good mount in that area would result in best possible temps, considering the ihs is making good contact on the inside.
    That would be the left half of 1 and 0 and the cache is located in the right half of 3 and 2.
    オタク
    "Perfection is a state you should always try to attain, yet one you can never reach." - me =)

  5. #5
    Xtreme Member
    Join Date
    Jan 2007
    Posts
    209

    I know

    Quote Originally Posted by Fr3ak
    I didnt read all posts in detail. Will do later, but a quadcore doesnt have a core in each quarter of the CPU.
    Have a look at one of the pics google finds here:
    http://images.google.com/images?svnu...00&btnG=Search

    It becomes pretty obvious which parts are the hottest, so getting a good mount in that area would result in best possible temps, considering the ihs is making good contact on the inside.
    That would be the left half of 1 and 0 and the cache is located in the right half of 3 and 2.
    I know the quadcore does not have a core at each quadrant but for some reason pressure in certain parts of each quadrant can effect that core as mapped. Im not saying there is a core under each quadrant in the pics just that there must be some kind of heat spreader. Im sure each chip will vary but from my tests on two quad cores that map is accurate for the heat spreader the lines are not definite and blurr that is why turning more than one screw sometimes may require turning two or letting up on one and tightening another kind of like spokes on a bicycle wheel.

  6. #6
    Unoriginal Macho Energy
    Join Date
    Dec 2002
    Posts
    3,158
    Quote Originally Posted by redcorn
    I could not get consistent results using the springs, All testing and procedures were more successful with hard mounting.
    I'm right there with ya. I havent used springs in years, I find they suck rhino dung. Hard mount on IHS is the only way to go IMO.

    *:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*
    GTZ --> MCW-NBMAX --> EK FC --> PA 120.3 --> PA 160.1 --> 2x DDC Ultras in Series --> Custom Clear Res
    "Artificial intelligence is no match for natural stupidity."
    *:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*'``'*:-.,_,.-:*

    Quote Originally Posted by ranker View Post
    Did you just get hit in the head with a heavy object? Because obviously you're failing at reading comprehension.

  7. #7
    Xtreme Cruncher
    Join Date
    Nov 2006
    Location
    'burb of Cleveland, Ohio
    Posts
    2,871
    Im not positive, but I think Ive seen a picture saying this is how the cores are laid out in a quad core Intel CPU. Im not sure about the numbers other than 3 was on the end.

  8. #8
    Xtreme Member
    Join Date
    Jan 2007
    Posts
    209
    Quote Originally Posted by nikhsub1
    I'm right there with ya. I havent used springs in years, I find they suck rhino dung. Hard mount on IHS is the only way to go IMO.
    I think the reason why the springs suck and were inconsistent was due to heat. When cold they would give you a constant temp but after 2 hours the springs must soften and let up on tension. Not only that you could really never figure out if each spring was equal in pressure. Its almost impossible to reproduce two spring mounts back to back.

    Nikisub I never did any type of benchmarking for this new block and was not anal about testing with TAT and such, I figured the guys have you for that comparo test your doing. Right now I am very happy with the Fuzion. Of course after putting in a back plate, lapping my cpu and figuring out how to mount it on the quad core all made major improvements from my prior setup.
    Last edited by redcorn; 01-25-2007 at 04:03 PM.

  9. #9
    Xtreme Member
    Join Date
    Jan 2007
    Posts
    209
    Quote Originally Posted by Polizei
    Im not positive, but I think Ive seen a picture saying this is how the cores are laid out in a quad core Intel CPU. Im not sure about the numbers other than 3 was on the end.
    I knew putting up a grid on the CPU would screw people up. The grid does not signify where the cores are or where the heat spreaders are. The grid merely shows where to put pressure to influence temp changes to each specific core. You do this by tightening the screw that is closest to the square or the core you want to cool down or loosen the screw closest to the core you want to raise in temp. You simply cannot tighten every screw and hope they will be equal this doesnt work due to the limitations of the cpu IHS shape and the shape of the cpu block base. You have to loosen one inorder to balance the needs of the other cores. Its very transitional all Im saying is putting pressure on square one affects the C1 and putting pressure on square 3 puts more pressure on C3.

    Your pic of the heat spreader is exactly like my pic except its not in squares. If you were to mount the block on and just tightened all the screws you would still hit all four heat spreaders of each core due to the fact that its in the middle of the cpu heat sink. However by applying specific pressure in each square I mapped you will get more of an even pressure and can fine tune it. Im not disagreeing with your pic infact it corresponds to what I mapped just flip that pic so the pin one is in the upper left hand corner. I can tell you that by balancing pressure in all four squares of the map your overall cpu temp will be down and each core will have the same temp. YOu can do this by using the core temp beta program and adjusting your tension screws.

    I found that 1 in 8 mounts randomly did I get the block to sit on the cpu right and the temps on all four cores 1C from each other if not the same. On my random mounts all I did was lower the cpu block straight down and tighten each screw till they couldnt be tightened in a star pattern. By doing the method outlined in conjuction with using core temp beta you can find which corners to tighten or loosen to give all cores the same pressure on each and every mount no guessing. It does require you to reboot alot since you never want to adjust those screws with the system on.
    Last edited by redcorn; 01-25-2007 at 04:12 PM.

  10. #10
    Xtreme Addict
    Join Date
    Jul 2006
    Location
    Vancouver, BC
    Posts
    2,061
    Redcorn, It might be easier for people to understand if you put a pic of the mounted water block with labels on each mounting screw as follows...

    "Tighten this screw to decrease temp on core 0"
    "Tighten this screw to decrease temp on core 1"
    etc...

  11. #11
    Xtreme Cruncher
    Join Date
    Jan 2004
    Location
    Philadelphia, PA
    Posts
    782
    also it is much easier to adjust the nuts while the cpu is loaded and you can watch coretemp realtime and see the effects of your tightening and loosening


  12. #12
    Xtreme Member
    Join Date
    Jan 2007
    Posts
    209

    It might be bad for your system.

    Quote Originally Posted by Hassan
    also it is much easier to adjust the nuts while the cpu is loaded and you can watch coretemp realtime and see the effects of your tightening and loosening
    this could be dangerous to your system espically if you accidently hit a water connection and it comes off and your system is live, thats why I always shut it down did my fine adjustment then rebooted and checked temps. It only takes two to three reboots before I dial in the correct temps. I did alot of reboots in 12 hours of testing let me tell ya.

    Hassan where did you get that program prime95 v25 that loads all the cores. Could you email it to me I would like to run some tests on it and post results. www.colnago31@yahoo.com

    Also post your screen shot of your quad core and Ill tell you which screws to tighten and which to lossen. I know you got yours dialed in pretty close. I think you had two cores at 69 and two at 66C under load, I just foreget which cores they were.
    Last edited by redcorn; 01-25-2007 at 04:49 PM.

  13. #13
    Xtreme Member
    Join Date
    Jan 2007
    Posts
    209

    will do.

    Quote Originally Posted by virtualrain
    Redcorn, It might be easier for people to understand if you put a pic of the mounted water block with labels on each mounting screw as follows...

    "Tighten this screw to decrease temp on core 0"
    "Tighten this screw to decrease temp on core 1"
    etc...
    I ll do that tonight. I have an EVGA motherboard that is inverted but it should still get the point across.

    Ill try to explain it one more time with this diagram: So if I had the following C0=55, C1=51, C2=54 and C3=48 here is a diagram with what screws to turn to adjust the load on the block to cpu. Make sure to find pin one or the gold triangle and then set up the map based on what your motherboard lay out is.



    I would loosen the screw by C3, tighten the screw at C0 and tighten the screw by C2, I would leave C1 screw alone since its in between in temps of the other three.
    Last edited by redcorn; 01-25-2007 at 07:38 PM.

  14. #14
    Xtreme Mentor
    Join Date
    May 2006
    Location
    Croatia
    Posts
    2,542
    I think somebody has to much spare time on their hands...
    ________
    Chevrolet lacetti
    Last edited by XS Janus; 05-12-2011 at 06:26 PM.

  15. #15
    Xtreme Cruncher
    Join Date
    Jan 2004
    Location
    Philadelphia, PA
    Posts
    782
    Quote Originally Posted by redcorn
    Hassan where did you get that program prime95 v25 that loads all the cores. Could you email it to me I would like to run some tests on it and post results. www.colnago31@yahoo.com
    http://www.xtremesystems.org/forums/...d.php?t=126820


  16. #16
    Xtreme Member
    Join Date
    Jan 2007
    Posts
    209
    Quote Originally Posted by virtualrain
    Did this improve load temps in any way? Are they more even across cores as well? Is the average core temp lower overall at load?
    Thanks to Hassan I was able to load all four cores here are the new numbers and tell me what you think if lapping the cpu, installing the Fuzion block, installing the backplate to the motherboard and correctly mounting the cpu changed the temps. Ambient room temperature was 77F or 25C. Fan speed was 5 out of 10 and pump is 50%. Man look at how even the cores are working in temp loads.

    here is after in idle and load:



    Last edited by redcorn; 01-26-2007 at 01:45 PM.

  17. #17
    Xtreme Cruncher
    Join Date
    Jan 2004
    Location
    Philadelphia, PA
    Posts
    782
    now thats good temps for a Kentsfield fully loaded, how long did you run prime? Did you see and increase after 30 mins?


  18. #18
    Xtreme Member
    Join Date
    Jan 2007
    Posts
    209
    Quote Originally Posted by Hassan
    now thats good temps for a Kentsfield fully loaded, how long did you run prime? Did you see and increase after 30 mins?
    I forgot and left prime 95 running even though the prime 95 window was closed it was still running in the startup bar for three hours while I was surfing the net and no increase in temps. Max load at the end of three hours was 61-62C. I think I can even do better by tweaking more pump speed and always increase the fans but I dont like the noise level past 6/10 on fan settings. Artic Silver must be starting to cure also. This set up now is about 40db due to the noise of the PCPOWERANDCOOLING 1KW SR. I cant hear my system just the power supply so thats pretty good for ambient temps in my room of 77-78F or 25-26C without openning a window. The Fuzion block is definitely doing its job! My last water block setup could never tame the quad efficiently like this but I probably didnt mount it right and didnt have the back plate or lapped the cpu.

    Yeah the temps are really good considering there are two 8800GTX cards in SLI in the water loop and those cards get red hot. Although prime95 only tests the cpu the two gpus without water cooling idle at 65C on air. So throw them into the water loop and its got to raise the temps on cpu cores.
    Last edited by redcorn; 01-26-2007 at 04:57 PM.

  19. #19
    Xtreme Addict
    Join Date
    Oct 2004
    Location
    Tampa, FL
    Posts
    2,105
    I am confused..what exactly is hard mount?
    My Heat
    i5 2500k @4.5ghz Raystorm
    Asus P8Z68-V Pro/Gen3 16gb(4x4) G.Skill PC12800 Ripjaws X
    x1900xt MCW60
    Rad: Thermochill PA120.3 3YL SL/ Pump: DDC2 w/ Petra's top 7/16in ID masterkleer
    Corsair 120gb Force GT SSD/ 1TB WD Black Caviar SATA
    X-Fi music/SH-203B/H62L/LH-20A1L
    Corsair HX620w /Acer AL2223W/modded TJ07

  20. #20
    Xtreme Cruncher
    Join Date
    Dec 2006
    Location
    Palo Alto, CA
    Posts
    4,150
    Quote Originally Posted by nealh
    I am confused..what exactly is hard mount?
    I think its when you do away with the springs and use bolts directly.
    Lenovo Thinkpad X220 - Core i5 2410m, 4gb
    waiting on 28nm video cards...

  21. #21
    Xtreme Member
    Join Date
    Jun 2006
    Location
    Corinth, MS
    Posts
    366
    Quote Originally Posted by serialk11r
    I think its when you do away with the springs and use bolts directly.
    Yes, it is.
    Why is it drug addicts and computer afficionados are both called users?

    | AMD Phenom II X6 & Xigmatek Dark Knight
    | ASUS M4A89GTD PRO
    | Sapphire HD 5770 1GB
    | Mushkin BL 2 x 2GB
    | WD Caviar Black 1TB
    | Corsair TX650
    | Lian Li PC-K7 Black

  22. #22
    Xtreme Member
    Join Date
    Jan 2007
    Posts
    209
    Quote Originally Posted by nealh
    I am confused..what exactly is hard mount?
    Yes you mount the nuts directly without the springs which gives you more even loading and precise loading of the block onto the cpu. After 15 remounts I was convinced the springs caused uneven mounting and were never reproducible from one mount to the next as seen in coretemp beta.

  23. #23
    Xtreme Enthusiast
    Join Date
    Aug 2006
    Location
    North Carolina, USA
    Posts
    839
    redcorn--great, great topic. See, this is why I frequent XtremeSystems.

    Do you have a picture of the backplate you're using and how you have it fitted to cut down on the bowing? A pic of the hard mount would be great as well!

  24. #24
    Xtreme Member
    Join Date
    Jan 2007
    Posts
    209
    Quote Originally Posted by Xion X2
    redcorn--great, great topic. See, this is why I frequent XtremeSystems.

    Do you have a picture of the backplate you're using and how you have it fitted to cut down on the bowing? A pic of the hard mount would be great as well!
    http://www.xtremesystems.org/forums/...=1#post1972879

    http://www.xtremesystems.org/forums/...=1#post1970470
    Last edited by redcorn; 01-29-2007 at 09:08 AM.

  25. #25
    Banned
    Join Date
    Apr 2006
    Location
    Portishead, Bristol, England
    Posts
    3,248


    Belive thats the pics someone tried to post of the cores! FCG is going to be the only dude here that knows as he's poped the IHS off loads of random CPU. Och kinda make him sounds like a pimp or soemthing...

    Nice guied, nice testing and lots of your own personal time spent on this. Im not even a quad core man but dual core but shall use this to help me on mine

Page 1 of 2 12 LastLast

Bookmarks

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •