Lapping the Heatkiller 3.0 [Top tip: DON'T DO IT!]

[DemonEyez]

I just got myself a HK 3.0 ( did not lap it). Now this is what it looks like with TIM on it:



Pressure is max, according to the info in booklet it's around 250N

From the looks of it the TIM does make very good contact in the middle but not around the edges.
Do I need to lap the IHS of my Q9650? my hunch says yes. I get the feeling the TIM has to be much more equal distributed over the CPU block for optimal contact and performance

[eRazorzEDGE]

i would try lapping the CPU, if you don't plan on warrantying it, that is. it'd be interesting to see the TIM imprint after wards.

[defect9]

it looks fine. you want the best pressure in the middle anyway, since that's where the cores are, and where the heat is generated first. by lapping it and the CPU you run the risk of making it so that the edges have the best pressure, and the middle doesnt get what it needs for pressure.

this thread already stated that lapping the HK and the CPU will give you NO IMPROVEMENTS. so... don't worry about that 1c you might get. You're better off adding more fans in push-pull, or more powerful fans, over lapping.

[gOtVoltage]

Well ,,
From the look of the STOCK HKBase," it puts preshure in a line". Arnt Intel Cores lower and this preshure is now above the cores instead of in the middle? If so i guesse the line is designed to run only One way for Intel correct. So Then you have the Phenoms they need a Nickel or Quarter size CIRCLE of preshure in the middle.

HK is designed to put preshure where the Intel chips need it. Make shure that you Install the Block so this patch/line runs across all your cores for Intels. If its mounted Pacth going opisite of cores , you may see higher temps because two cores wont get cooled as much. You can clearly see that. Phenom/PhenomII Die is Dead center not a line down the center or lower.

I wonder if it can be modded to apply a better circle contact patern instead of a line. This W/b is designed for Intel but they also sell it for PhenomII. Im tempted to get one but am afraid it wont give me the correct contact patch im looking for.

Note:
If you look at the Base "bolt pattern" it explains why it leaves a line insted of a circle.
Two short width sides =2x screw at each end, Thus less force than the long sides x3 screw.
I woulda made the block with 8 or 4 bolts for the Base instead of 6. This would make a perfect circle .

Beside that ,
Thermal paste,,how much and why

Even with a UNLAPPED W/B Hes using too much paste. about 75% is sqeezed where its not needed.

Once its lapped(OP) you will use a lot less . Actually Half or even "less than" half the Amount you would normally use.

A lot of people still use a Grain size dot. That is actually too much with these ne W/B's and better IHS's . Even a half grain dot can be too much depending on surface of W/B alone .

I usually use half a half dot ,,then spread to a almost see through haze method. MXII, AS5 NT1 and that G stuff will all use half the normal amount or less.

I did experiments a while back when the V1/V2 came out . I took the Bow out to use on Stock IHS . It actually performed the same or better and well with a lot less paste. The best gains were with the W/B and IHS lapped .


Experiments come in handy ,, Try this method before you alter the W/B back to a bow. You might find it interesting.


1st
Put on amount of Paste youd normally use,,

2nd
place and mount W/B.

3rd
Remove and wipe off W/B only not IHS.

4th
Do not wipe the IHS,,Respread the paste thats left on IHS.
(plasti wrap finger) works well for this or glove.

5th
I do the above two times back to back . The C's should drop each time.

If the temp goes up , instead of down ,,you only need to perform method #3ONCE.

I origanally used the drop meathod for about 10years , the drop and Squish method with w/b. Ive since found this to be Too much on a lapped surface while using the New paste over last few years.

Some paste will react better than others on a smoother surface. Even though the IHS is stock and rough. The W/B still has twice as fine and prabally 4x as many Micro pores from lapping .It will only use half the amount to do the same job. Thus when you apply the normal amount it just gets turned into a insulation and sqeezed out the side.

A sloppy or well applied paste makes or breaks a great Lap Job

The Heatkiller is growing on me .Its a bit bulky and Bolts coulda been resessed more or use of shorter head. Overall nice rugged type block it seems.

Well im Old school ,,
what my little V1 gets me with lapped base(stock IHS 720BE).
Just a quik little run. My CPUvid 1.55and NBvid is 1.45v @1.56vcore here




I dont know the best method , but i do know this does work with any W/B regardless of it Being lapped.

gOtVolTage

[DemonEyez]

Originally Posted by gOtVoltage

Well ,,
From the look of the STOCK HKBase," it puts preshure in a line". Arnt Intel Cores lower and this preshure is now above the cores instead of in the middle? If so i guesse the line is designed to run only One way for Intel correct. So Then you have the Phenoms they need a Nickel or Quarter size CIRCLE of preshure in the middle.

HK is designed to put preshure where the Intel chips need it. Make shure that you Install the Block so this patch/line runs across all your cores for Intels. If its mounted Pacth going opisite of cores , you may see higher temps because two cores wont get cooled as much. You can clearly see that. Phenom/PhenomII Die is Dead center not a line down the center or lower.

...
gOtVolTage

That's what I was thinking also. The cores of a quad are just below and above the horizontal line, I get the idea that the outer area's are getting to little pressure from the looks of it, and the two outer cores are running warmer, this seems to be the case because two cores run like 10 to 12 degrees hotter than the other two. With my former D-Tek Fuzion rev1 block the differences were more like 5 degrees.

The placement of the block is correct i think. Also when i look at these pictures:




The quadcore manifold is vertical. If I were to turn the block 90 degrees, so that the pressure line becomes vertical,then it would indeed put the most pressure on the the cores.
But then the quadcore nozzle would become horizontal and you don't have the water injected above each core-package, which will result in higher temps.

I'll try you method of applying the TIM. Sounds you know what you're talking about. I agree that it looks i've used too much TIM

[Marci]

T'would appear common knowledge is being lost into the depths of time - haven't checked but I guess these links have been omitted from the stickies by now... http://www.ocforums.com/showthread.php?t=329453 & http://www.xtremesystems.org/forums/...ht=anatomy+bow

[DemonEyez]

Originally Posted by gOtVoltage

...

1st
Put on amount of Paste youd normally use,,

2nd
place and mount W/B.

3rd
Remove and wipe off W/B only not IHS.

4th
Do not wipe the IHS,,Respread the paste thats left on IHS.
(plasti wrap finger) works well for this or glove.

5th
I do the above two times back to back . The C's should drop each time.

If the temp goes up , instead of down ,,you only need to perform method #3ONCE.
....

Okay I did as you suggested, now when i remove the block it looks like this:


I'll fire up my sytem and see if reapplying the TIM did some good.

[gOtVoltage]

Yup thats what it should look like..


See how it looks like a haze and that it has filled the pores in the center W/B with out wasting Paste. The Smoother a surfacr the less it will use no matter what.


These new paste are pretty amazing and are so close in performance . It really comes down to How much and the shape contact of W/B.


Try it, .....
The HK is built nice for Intels i tell ya. Little tweaks can make a differance.

Hmm maybe a hair more PRESHURE to left . Im just picky LOL. It can take a few tries to get it just right. Once you have tried this method it gets way easier with each CPU Change/W/B change. It allows you to see exactly How the paste is needed by IHS/W/B. Its also makes it easier to Remove W/B during CPU changes too.

Less paste = Less prybar to remove.

Youll be surprized at how much less paste this actually takes. I know i was

This is the type of detail i do with every build , no matter the cost of system.

[DemonEyez]

Well my temps with my former D-Tek Fuzion have gone down 2 degrees. So kudos to you for pointing out that paste issue.
I also lapped the IHS on my Q9650 this afternoon, but it made no difference for temps with the HK 3.0 block. Also the pressure line looked the same before and after lapping the IHS.

However I have rather conflicting results with my HK 3.0 block. I was under the impression that the HK 3.0 performs better than the D-Tek Fuzion.
So I did some test runs with both blocks, I jotted down idle temps and the highest temps (LinX running ten runs). This is what I got

HK 3.0 + Q9650 with stock IHS: idle 36 / 35 / 37 /28 full load 79 / 79 /67 /67
HK 3.0 + Q9650 with lapped IHS: idle 36 / 35 / 37 /28 full load 79 / 79 /67 / 67

D-Tek + Q9650 with stock IHS: idle 35 / 35 / 37 /31 full load 77 / 77 /67 / 69
D-Tek + Q9650 with lapped IHS: idle 33 / 32 / 37 /28 full load 73 / 73 /65 /65

I actually could't believe the results at first. Then I started thinking, could it be that the IHS on my Q9650 was already flat to begin with. Before lapping I drew a black cross on the IHS to see I if the IHS was bowed , when I just started lapping the cross equaly dissapeared, to that does indeed point to an already flat IHS.

My D-Tek Fuzion block is just a little bowed (using the washer that come with the nozzle kit), the HK 3.0 is far more bowed. So could one say that an bowed block is only useful when you got an bowed IHS? I was not expecting the HK 3.0 to perform this much worse opposed to my Fuzion block. I bought the block for improving my temps, not make them worse.

[gOtVoltage]

It happens sometimes,,Ive kept the V1/V2 because it perform on just about anything.

I wonder how the HK would perform with out the Nozzel plate. I almost wonder if the plate restricts it too much.

You could even try loosing the TWO outer bolts on each side a hair to where there not as snugg as the Single bolt side on each end . that will make the Contact area grow too.

Only loosen enuff to take some flex of the Long side of block. The sides a hair and then leak test it before ya mount it. If it passes the leak test id give it another shot.

It probally has to much bow for your IHS like you were saying.

I know i like to tweak things

Those numbers with the Fusion are pretty nice too.

[DemonEyez]

I tested again, now without the Nozzle plate, and temps are basicly the same, today is a hot day and room temps are about 1.5 degrees higher then yesterday, so it's perhaps a little hard compare, but still , the temps are far higher than with the Fuzion block installed. I also took al look at the baseplate of my Fuzion when switching and it indeed looks like the Fuzion is making far better contact overal, it looks like a circular bow with a tad more presure on the second die-package, which seemes to correspond with my temps being lower for core 3 and 4.

I'm sticking with the Fuzion I think. Will test your suggestion for loosening up the two outer bolts. But I'm a little wary for using that 24/7 eventhough I can run a leek test before.

[DemonEyez]

Originally Posted by gOtVoltage

...

I wonder how the HK would perform with out the Nozzel plate. I almost wonder if the plate restricts it too much.

You could even try loosing the TWO outer bolts on each side a hair to where there not as snugg as the Single bolt side on each end . that will make the Contact area grow too.

Only loosen enuff to take some flex of the Long side of block. The sides a hair and then leak test it before ya mount it. If it passes the leak test id give it another shot.

It probally has to much bow for your IHS like you were saying.

....

Okay , I tried your suggestion. The block seems to make some better contact, but is still bowed in a line. Also I removed the quad nozzle but that doesn't seems to have an impact on the temps. Results are still a few degrees higher than my best temps with my Fuzion block:

HK 3.0 with significant bow + Q9650 with lapped IHS: idle 36 / 35 / 37 /28 full load 79 / 79 /67 / 67

HK 3.0 with less bow with nozzle + Q9650 with lapped IHS: idle 36 / 35 / 37 / 30 full load 77 / 77 /73 / 73

HK 3.0 with less bow no nozzle + Q9650 with lapped IHS: idle 36 / 35 / 37 / 30 full load 77 / 77 /73 / 73

D-Tek Fuzion1 small bow with nozzle + Q9650 with lapped IHS: idle 33 / 32 / 37 /28 full load 73 / 73 /65 / 65

So no HK 3.0 for me anymore. Too bad the shop I bought it from won't take back used watercool products. But I'm sure I get rid of it second hand for a good price.

[WeeMaan]

I'm waiting on my Heatkiller now, but I have a question.
I already have a lapped cpu (i7 920 D0), you think it's worth lapping the block once it arrives?
If it is convex, could lapping it actually hurt the temps?

[Waterlogged]

Originally Posted by WeeMaan

I'm waiting on my Heatkiller now, but I have a question.
I already have a lapped cpu (i7 920 D0), you think it's worth lapping the block once it arrives?
If it is convex, could lapping it actually hurt the temps?

To answer your Q:

Originally Posted by HESmelaugh

. . .snip. . .
Conclusion

As expected, the performance suffered from lapping down the bow. Though I never expected it to suffer this much!

What did we learn?

- The HK 3.0 has a slightly bowed base.
- It's nice to see your reflection in a block's base but not necessarily a good thing.
- Shane is willing to wreck his stuff in the name of science...

Cheers,
Shane

[Jon K]

I know everyone is into lapping them down and making them flat, which obviously isn't always good.

However, has anyone considered simply polishing the surface that's already formed? IE., taking the HK 3.0 to my 8" buffer wheel and using some copper compound and mirror polishing whatever shape they already machined?

[kinghong1970]

Jon, read the first link from Marci above... polishing is overrated... you don't need to see your gorgeous face on the bottom of your block in order to get nice temps.

[Diverge]

Originally Posted by Jon K

I know everyone is into lapping them down and making them flat, which obviously isn't always good.

However, has anyone considered simply polishing the surface that's already formed? IE., taking the HK 3.0 to my 8" buffer wheel and using some copper compound and mirror polishing whatever shape they already machined?

Using any kind of polish compound for something intended on transferring heat isn't gonna be good. The job of tim is to fill microscopic spaces, but if packed with polish, its won't work as well.

[Brodholm]

There is lines left from the mill I think on the hk. So the surface isn't 100% smooth. Maybe just take a 1200grit paper and lap it with your hand so it don't get flat but it gets shiny? Would that be a good idea? Just to get down the "reffels" or how I should describe it?

But so far? A non lapped HK3.0 with a lapped cpu is best?!

[kinghong1970]

Originally Posted by Brodholm

...Just to get down the "reffels" or how I should describe it?...

it's RUFFLES... as in RRRUFFLES have Ridges!