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:

http://img166.imageshack.us/img166/4...ofcores6gw.jpg
http://img443.imageshack.us/img443/6...agram33hg0.jpg

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.

http://img213.imageshack.us/img213/5...ckidle15vb.jpg

http://img443.imageshack.us/img443/1...exposedgy7.jpg

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.

http://img166.imageshack.us/img166/9...mpleone7wj.jpg

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:
http://img244.imageshack.us/img244/7708/idle77fmm2.png

Loaded temps:
http://img145.imageshack.us/img145/8...munbersyy1.jpg

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

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?

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.

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.

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.

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.

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.

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.

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.

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...

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

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.

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.

http://img443.imageshack.us/img443/6...agram33hg0.jpg

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.

I think somebody has to much spare time on their hands...
________
Chevrolet lacetti

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

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:

http://img244.imageshack.us/img244/7835/idle77fym1.png

http://img145.imageshack.us/img145/8...munbersyy1.jpg

now thats good temps for a Kentsfield fully loaded, how long did you run prime? Did you see and increase after 30 mins?

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.

I am confused..what exactly is hard mount?

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.

Quote:

---

Originally Posted by serialk11r

I think its when you do away with the springs and use bolts directly.

---

Yes, it is.

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.

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!

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

http://www.hardware.info/images/news/quad_core.jpg

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

Very nice little writeup. Evened out my core temps nicely. Thanks Redcorn!!

Quote:

---

Originally Posted by Sideroxylon

Very nice little writeup. Evened out my core temps nicely. Thanks Redcorn!!

---

Ive done this procedure on two other installed quad cores and they had even temps all across. Both had to install the back plate though and no springs we hard mounted them.

Why run large ffts ? small ffts can give 4-8C higher load temps even with priority set at default

Also different boards seems to recognise the order for cpu0 to cpu3 for quad cores differently from what i tested with Q6600 on intel d975xbx badaxe vs p5b deluxe and asus 680i striker.

interesting nonetheless... will try without springs and see as my first die seems to be always hotter

this with swiftech g4 storm stock mount

http://fileshosts.com/intel/Asus/P5B...ime95_load.png

http://fileshosts.com/intel/Asus/P5B...s_24hrmark.png

Quote:

---

Originally Posted by redcorn

Ive done this procedure on two other installed quad cores and they had even temps all across. Both had to install the back plate though.

---

Well I don't have the back plate to install but I got all my cores within 2c of each other. I was somewhere between 3-6c difference between cores before.

Now I'm wondering if I should remove the springs and hard mount it.

Quote:

---

Originally Posted by Sideroxylon

Well I don't have the back plate to install but I got all my cores within 2c of each other. I was somewhere between 3-6c difference between cores before.

Now I'm wondering if I should remove the springs and hard mount it.

---

2c difference between the cores with the springs installed is not to bad, I guess if you didnt want to mess with it anymore just stop there until you put a back plate on later.

Man that 975 chipset can really push the quad core well. I wish the 680i chipsets could go past 300 fsb. I wish intel would do a 975 chipset with nvidia SLI. Here is my small TTFs which is probably the equivalent to what TAT does to increase temps fast. This is at 78F or 26C ambient temps with fan on 5/10 and pump half power. One hour on prime 95 small TTF

http://img407.imageshack.us/img407/5223/smallttfsl9.jpg

IM SORRY FOR RESURECTING THIS THREAD!

But i think its a very good thread, and should be bumped back up to the top.

Thanks for this, it's interesting :)

I will try this soon as my temps can vary up to 10C between the hottest and coldest core.

i so needed this hardmount it is!!

Quote:

---

Originally Posted by Pete

http://www.hardware.info/images/news/quad_core.jpg

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

---

And if you go here it shows the Kentsfield quads oriented differently,

http://tweakers.net/reviews/614/18

This is somewhat confusing, not the procedure, just the initial orientation of the waterblock, by the way Redcorn, great guide, should be in the FAQ section.

Utnorris

hardmounted my fuzion and all cores are 1c of eachother :)

Quote:

---

Originally Posted by disruptfam

hardmounted my fuzion and all cores are 1c of eachother :)

---

Sweet, very nice disruptfam!!!

Utnorris

Oh yea?

Sounds logical...

Ill try and even out my 6-8ºC Difference...

any pics of the hardmounted fuzion?

Hey, great guide, I can tell it took a LONG time to get it all done, :clap:. I spent the last 2 hours trying to duplicate your results, and have not had success. I've re-mounted my Apogee GT, removed the little plastic spacers, and after a LONG time trying to get it right, I don't think I can with my proc. I took my temps form something erratic and 8C difference to:

C0-54
C1-54
C3-49
C4-49
This is with the processor under load at 3.6ghz with a lapped IHS, and BOWED apogee GT.
The first 2 and bottom two tend to stay close to each other no matter what (this is after 2 re-applications of Arctic Silver), so I'm guessing that one core is just a bit of a dud. Either way, I shaved off 3C off with this whole exercise, and stabilized my temps some, so again good job. If you have any advice for my setup, i'll gladly accept it.

Redcorn, great guide. Just spent about an hour and a half remounting my apogee gtx to my q6600. I used liquid metal pro, so it took me forever to clean everything off. I actually had to use a no-scratch scotch-brite pad and turtle was chrome polish, followed by articclean. Load temps with my q6600 B3 are more uniform across the board 54,54,54,52 with ambient of 72.2 F. No matter what I do, my idle temps have cores 0, and 2 higher. Just must be the processor, cause it's done this with many remounts with any TIM.

Anyone looked it the D-tek pro-mount kit?? I get a 10c split across my cores on my G0 which i can't seem to get down, but i managed to get them within a 1-2c on my B2 using the same technique