3770K Delidded, Lapped, Liquid Pro TIM. What a difference! (Lots of Pics!)

chrisoverson · 14 Jan 2013 at 19:06

Hi guys

. You might have seen my post in which I announced my successful de-lidding of my CPU. I decided I could have done more, so today I took everything apart again and decided to have a go at lapping.

Lid off, TIM removed:

Heatsink marked (badly):

CPU marked:

And away with the lapping... working up from 240 grit.

Went all the way up to 1200 grit, then used some "Lapping Paper" I found at the local tool shop. Don't think that made much difference though.

Lapping the heatsink (a little too much water):

Afterwards, both nice and shiny and down to flat copper:

Liquid Pro on both CPU die and base of IHS:

Liquid Pro on Heatsink:

CPU and IHS in place, more Liquid Pro applied:

.............

Final Results!

Clocked to 4.5ghz @ 1.3vcore

Before (standard IHS TIM, AS5 on heatsink):
Idle temps: 45, 41, 44, 40. Avg = 42.5
Prime95: 85, 90, 90, 84. Avg = 87.25

After 1st De-Lid (Liquid Pro on CPU Die, AS5 on heatsink):
Idle temps: 35, 32, 33, 28. Avg = 32
Prime95: 70, 72, 73, 69. Avg = 71

New result (Lapped, Liquid Pro everywhere

):
Idle temps: 25, 22, 22, 20. Avg = 22.25
Under Load:

56, 56, 57, 53. Avg = 55.5

I think it was worth the few extra hours work

.

This was my first time lapping a heatsink/cpu guys, so I probably did some things wrong. Please share your questions/comments though!

chrisoverson · 14 Jan 2013 at 19:06

Woops, images too big, I'll change them ASAP guys.

panyan · 14 Jan 2013 at 19:12

Impressive results

When you said you were going to start lapping i thought you were going to start sanding down the die!

JonJ678 · 14 Jan 2013 at 19:13

So that's 87 -> 71 -> 55 as you do more things? Astonishing. Good results

You've definitely used a lot of liquid pro!

chrisoverson · 14 Jan 2013 at 19:19

Okay, fixed the images.

Yes, it did seem like a fair bit of liquid pro, although I must say that stuff goes miles if you take your time with it, I only used the tiniest bead of it to cover both the top of the IHS and the heatsink in one go!

Got the idea to try it from someone in another thread (can't remember who, or I'd have given them credit). But it was quite a thin layer, and it certainly doesn't seem to have hindered performance.

This is all it took for the IHS: Link

Litmoose · 14 Jan 2013 at 19:26

So lapping = sand papering basically? But the difference from the very first test to the end is huge. If it were a f1 car that's like knocking 5 sec a lap off.

chrisoverson · 14 Jan 2013 at 19:28

Litmoose said:
So lapping = sand papering basically? But the difference from the very first test to the end is huge. If it were a f1 car that's like knocking 5 sec a lap off.

Yes, basically sanding it down

.

It seems to have two beneficial effects... one being that you end up with a flatter mating surface for the heatsink and CPU (the CPU is slightly concave as standard, bending inwards at the middle), and the other being the removal of the nickel coating, which doesn't conduct heat as well as copper does.

Using the liquid pro I think definitely helped, as even if I used too much of it it won't hinder performance anywhere near as much as standard paste, because being 100% metal it conducts so much better. It should be as close as I can get to actually soldering the IHS/heatsink on

.

chrisoverson · 14 Jan 2013 at 19:36

If you're into F1 btw one of the things I often use this PC for is F1 2012, if you want to come play

.

I'll be benchmarking that later, after I try to push the CPU frequency up again. (GPU will probably be the bottleneck but still fun to try).

Winston · 14 Jan 2013 at 20:26

Makes me want to redo mine with liquid pro and get the sandpaper out!
Nice results

chrisoverson · 14 Jan 2013 at 20:29

Winston said:
Makes me want to redo mine with liquid pro and get the sandpaper out!
Nice results

Thanks

.

Be sure to post up the results if you do, I'm quite surprised how much difference it's made.

Now running 4.7ghz @ 1.36vcore, max temps under Prime95 got to 67.25 average. Much better than before; I couldn't even get 4.7ghz without 1.42v.

Nickolp1974 · 14 Jan 2013 at 21:50

Nice results

found my core spread temps to be much better once lapped, within 3c now at full load. Got liquid pro under ihs and gelid ex on top.

Did you lap yours straight on the table??

chrisoverson · 14 Jan 2013 at 23:45

I did... I originally was doing it on an old side from a case, but it kept moving so I gave in and did it directly on the desk.

I know that if I'd done it on a perfectly flat metal surface it might have been better... but I had nothing of the sort to hand. Hasn't damaged the desk, and seems to have worked well

.

hlennie · 15 Jan 2013 at 05:43

nice one going to do mine again and rebond the ihs

Nickolp1974 · 15 Jan 2013 at 13:38

chrisoverson said:
I did... I originally was doing it on an old side from a case, but it kept moving so I gave in and did it directly on the desk.

I know that if I'd done it on a perfectly flat metal surface it might have been better... but I had nothing of the sort to hand. Hasn't damaged the desk, and seems to have worked well .

i'd suggest maybe a piece of glass next time

did mine on an old glass hifi stand!

chrisoverson · 15 Jan 2013 at 23:21

Nickolp1974 said:
i'd suggest maybe a piece of glass next time did mine on an old glass hifi stand!

I hunted everywhere for a spare bit of glass... didn't have one to hand.

I've got an old dead printer/scanner though, I'll probably take the glass plate out of that for next time

zepzeppelin · 16 Jan 2013 at 05:28

I'm wondering if anyone else is willing to try the procedure as described above, and post the results? because I can tell you I did exactly that, and it made no difference to load temperatures whatsoever, so I'm calling the Bluff! unless others can state otherwise. Also It's not actually copper on the heat sink, it's a copper nickel alloy, which is why it probably doesn't work

steve258 · 16 Jan 2013 at 05:56

32C drop

55C with 1.3V is quite unbelievable!

I removed IHS and went die direct contact and the temp drop was 16C from stock TIM. It was also at 4.5Ghz but with lower volt at 1.168V, also I didn't lap mine and no liquid pro.

Nickolp1974 · 16 Jan 2013 at 07:43

zepzeppelin said:
I'm wondering if anyone else is willing to try the procedure as described above, and post the results? because I can tell you I did exactly that, and it made no difference to load temperatures whatsoever, so I'm calling the Bluff! unless others can state otherwise. Also It's not actually copper on the heat sink, it's a copper nickel alloy, which is why it probably doesn't work

From my thread, http://forums.overclockers.co.uk/showthread.php?t=18451271&highlight=username_Nickolp1974

steve258 said:
32C drop 55C with 1.3V is quite unbelievable!

I removed IHS and went die direct contact and the temp drop was 16C from stock TIM. It was also at 4.5Ghz but with lower volt at 1.168V, also I didn't lap mine and no liquid pro.

Did you have to remove the retainning bracket?

zepzeppelin · 16 Jan 2013 at 08:57

Nickolp1974 said:
From my thread, http://forums.overclockers.co.uk/showthread.php?t=18451271&highlight=username_Nickolp1974

Ok, but I meant the actual sanding down to the copper alloy, that's what made no difference to the temps I had before, they stayed exactly the same, maybe even a touch higher.

Anyway I have a question? and maybe you can answer it I don't know? but here goes.

When I bought my 3570K I actually screwed up my other board, as some Thermal Paste dropped onto the CPU pins, and I bent them trying to get it Off! this resulted in the PC looping into a continuous reboot and not posting. Also at the PC store they said it looked like the back of the CPU had a Flash burn (looked like a small oil slick), but it still works on this new board. I'm wondering whether the Flash Burn has impacted it's performance. I reset the BIOS with the latest firmware from ASUS, and tested a slow increase from stock x40 hit 71c on Prime @1.224v and x42 hit 80c @1.272v nothing changed except the ratio every other setting as standard. Either the Flash burn did! indeed lower the CPU performance or it's just a duffer of a chip, what do you think?

chrisoverson · 16 Jan 2013 at 12:12

I think one of the big issues with these chips is how varied the results are.

I've seen some people getting 4.5ghz on silly low voltages, others need 1.3v or higher. The inconsistancy is just crazy.

It stands to reason that because the chips aren't soldered, but are just slathered in crappy intel TIM, that that process isn't always equal also... so some people will benefit far more from delidding than others. Also, there's difference in motherboards, ambient temperatures etc.

As for lapping, because heatsinks aren't always flat, and the IHS isn't always flat, and it seems that some are shaped differently (some concave, some convex, some flat to start with), then different people with different heatsinks will have different experiences with lapping.

Also, the more you take off the IHS and heatsink the closer you get to contact between the die and heat pipes... so thats a factor too. If you spent longer with lower grit sandpaper and took more of the IHS away... then perhaps it conducts heat better? It's not really a precise art.

I'm currently writing this on my laptop because I'm playing with voltages etc. on my PC. If you're not believing some of the results I'm getting... then by all means I'll do further testing and show some more evidence.