CPU lapping

[P4-630]

is lapping a CPU a good idea?

I did it in the past with a Intel E2200 and a E7200.
Can't remember the temp differences though.

 

[JalleR]

Well if you want to go all the way then do it,


i have done it on 4 CPU's so fare
My brothers 7820k it dropped 10C, my own 5930K dropped 15c and my 6900k dropped 10 ish, and it is so easy to do that i did it for fun on my Home server CPU i7-4771 just because why not. (the last 2 cpu's was done without any motorized tools )

The first 3 cpu's was/is custom loop cooled

If you are handy then it will go well but watch out for those pins,
Reason for not using motorized tools: i launched my 5930k into a wood plate so hard that it bounced back in to the glass plate that i had mounted on a drill (for the sanding process) so it broke into a million pieces, it got a serious dent but it survived
(Maybe the problem is my skills for combining stuff with a motorized tool )

 

[ThrashZone]

I did it in the past with a Intel E2200 and a E7200.
Can't remember the temp differences though.

Hi,
Lapping usually only tightens core temperature spread seeing the top of a cpu can be pretty wonky and flattening it at least give it something better for a cooler to mate too of course some coolers are designed to not be flat.

 

[Valantar]

It's fascinating how people are just blithely responding to the OP and not the necro, despite the OP being from nearly two years ago

 

[P4-630]

It's fascinating how people are just blithely responding to the OP and not the necro, despite the OP being from nearly two years ago

I usually look at the posting dates, this time I didn't...

 

[ThrashZone]

What do you mean by "CPU core"? The IHS? The die? If it's the IHS, it's copper, so that sounds like you've gotten something conductive between it, some voltage rail, and it might be shorting to ground through the heatsink. If you mean the die itself, then yes, sanding that down will eventually expose the conductive innards of the CPU.

It's fascinating how people are just blithely responding to the OP and not the necro, despite the OP being from nearly two years ago

Hi,
Doesn't look like you caught it either from your first post lol yes fascinating indeed

 

[Valantar]

Hi,
Doesn't look like you caught it either from your first post lol yes fascinating indeed

Uhm, I was responding to the necro.... Did you read the post? Did you see who I quoted? Did you check the date on the post I quoted?

 

[ThrashZone]

Uhm, I was responding to the necro.... Did you read the post? Did you see who I quoted? Did you check the date on the post I quoted?

Hi,
Yeah you didn't say anything about how old the thread was so cudos for not noticing too.

 

[Valantar]

Hi,
Yeah you didn't say anything about how old the thread was so cudos for not noticing too.

Oh jeez. Please reread. Here, I'll quote it for you:

It's fascinating how people are just blithely responding to the OP and not the necro, despite the OP being from nearly two years ago

See? I said it's funny that people are responding to the OP and not the necro, not that they didn't notice at all. I responded to the necropost and only that, as I did indeed spot the 2-year-old date and wondered "Hm, I wonder what has happened to bring this up to the recent threads column." Lo and behold, it was the post I responded to. Almost as if I knew, no?

 

[delshay]

What do you mean by "CPU core"? The IHS? The die? If it's the IHS, it's copper, so that sounds like you've gotten something conductive between it, some voltage rail, and it might be shorting to ground through the heatsink. If you mean the die itself, then yes, sanding that down will eventually expose the conductive innards of the CPU.

The modification I did to the CPU has nothing to do with the flat part of the CPU, but yet it now has resistance to ground. This is clearly not normal , but the CPU is working fine.

The IHS is soldered to the heatsink (old experiment), so the complete IHS & heatsink is conductive & gives out a voltage reading. This mean if the heatsink come into contact with ground, something is going to fail or get burnt. The voltage can only be coming from the flat part of the CPU.

If I take out the CPU put the DMM direct to the flat part where we normally put thermal paste (mine has LM still present) & touch almost any pin with the negative terminal on the back of the CPU it now has a resistance reading. I have never seen this before, first time today.

 

[Valantar]

The modification I did to the CPU has nothing to do with the flat part of the CPU, but yet it now has resistance to ground. This is clearly not normal , but the CPU is working fine.

The IHS is soldered to the heatsink (old experiment), so the complete IHS & heatsink is conductive give out a voltage reading. This mean if the heatsink come into contact with ground, something is going to fail or get burnt. The voltage can only be coming from the flat part of the CPU.

If I take out the CPU put the DMM direct to the flat part where we normally put thermal paste (mine has LM still present) & touch almost any pin with the negative terminal on the back of the CPU it now has a resistance reading. I have never seen this before, first time today.

Your wording is still vague and hard to grasp. By "the flat part of the CPU", do you mean the die? As in the little dark square in the middle of a pic like this (yes, I know that's an Intel chip)? That's the die.

If you've been sanding and scrubbing the die, it's no wonder you're getting a voltage reading off it. You've likely worn down the diffusion barrier on top of the silicon sufficiently for it to be really, really thin, but not thin enough to actually cause a short, and then liquid metal has probably diffused into this to provide a path for voltage to flow. I would expect this to mean the CPU will die relatively shortly, as a process like this once progressed this far is unlikely to just stop and stay where it is. If the diffusion barrier fails entirely, you'll effectively have exposed the inner workings of the die, making the surface conductive and making it highly likely that your LM will diffuse into the core itself, causing it to fundamentally change its molecular makeup.

 

[JalleR]

It's fascinating how people are just blithely responding to the OP and not the necro, despite the OP being from nearly two years ago

LOL yes i didnt look at the date of the OP

Well is sounds like your liquid metal has worked its way to the backside of the IHS and is creating conduction between the core and the IHS or maybe the Cpu pcb and IHS.

 

[delshay]

Your wording is still vague and hard to grasp. By "the flat part of the CPU", do you mean the die? As in the little dark square in the middle of a pic like this (yes, I know that's an Intel chip)? That's the die.

If you've been sanding and scrubbing the die, it's no wonder you're getting a voltage reading off it. You've likely worn down the diffusion barrier on top of the silicon sufficiently for it to be really, really thin, but not thin enough to actually cause a short, and then liquid metal has probably diffused into this to provide a path for voltage to flow. I would expect this to mean the CPU will die relatively shortly, as a process like this once progressed this far is unlikely to just stop and stay where it is. If the diffusion barrier fails entirely, you'll effectively have exposed the inner workings of the die, making the surface conductive and making it highly likely that your LM will diffuse into the core itself, causing it to fundamentally change its molecular makeup.

Yes, I mean the die.

I don't go out of my way to sand down the die, I just that I use the brillo pad to clean it before I apply new LM compound.

The CPU die itself is outputting a voltage. & because the IHS is soldered to the heatsink, the whole complete assembly now has a voltage reading.

 

[Valantar]

Yes, I mean the die.

I don't go out of my way to sand down the die, I just that I use the brillo pad to clean it before I apply new LM compound.

The CPU die itself is outputting a voltage. & because the IHS is soldered to the heatsink, the whole complete assembly now has a voltage reading.

Yep, as I suspected. Like I said, you've likely worn down the diffusion barrier sufficiently that it's no longer acting as an insulator. I would expect the chip to start throwing random errors and die (no pun intended), though I have no idea how quickly or slowly that will happen.

 

[mouacyk]

dooit

 

[delshay]

Yep, as I suspected. Like I said, you've likely worn down the diffusion barrier sufficiently that it's no longer acting as an insulator. I would expect the chip to start throwing random errors and die (no pun intended), though I have no idea how quickly or slowly that will happen.

I don't think all what you are saying is 100% correct, as I'm not constantly using the brillo pad. Yes I do use it to shave a bit off the die, but only when cleaning which is not that often.

This is what I think has happen.

When mounting the cooler I may have cracked or chipped the edge of the CPU die. This could have also happened when using the brillo pad. I have to remove the LM & check the die for micro cracks to confirm.

The CPU is on test for more that a day now, all seems well. ..I can't tell if the latest new hardware modification is working, because it could also be because the IHS/heatsink has a voltage reading, It could be rejecting outside interference.. ..I can confirm, "stability is up".

Look on the bright side, have I accidently created the worlds first CPU/heatsink combo that has shielding technology built-in. ..Have a good giggle here.

 

[bobbybluz]

As a former machinist I have to say getting the top of a heatspreader anywhere near close to perfectly flat by hand is impossible. At one time years ago I discussed the idea of using a milling machine on an AMD FX-4000 with my former supervisor. We checked to top with a Starrett dial indicator with the CPU mounted in a test fixture for checking specs on small precision parts. The variance was so tiny we concluded it'd be a total waste of time and effort. You'll never come close to what a milling machine can do by hand. Also, thin surfaces flex under pressure and they don't do it evenly.

 

[hat]

As a former machinist I have to say getting the top of a heatspreader anywhere near close to perfectly flat by hand is impossible. At one time years ago I discussed the idea of using a milling machine on an AMD FX-4000 with my former supervisor. We checked to top with a Starrett dial indicator with the CPU mounted in a test fixture for checking specs on small precision parts. The variance was so tiny we concluded it'd be a total waste of time and effort. You'll never come close to what a milling machine can do by hand. Also, thin surfaces flex under pressure and they don't do it evenly.

My own personal experience seems to disagree with this. There's always gains to be had by lapping, even if they're not always the greatest. Plenty of crappy surfaces have been fixed by hand as well.

 

[delshay]

I have found the problem. It looks like I chipped the edge of the die when testing different heatsink, but it may have also been there before & I may of not notice it. I only carried out a DMM test for the first time due to CPU hardware changes, so it may have been present there for many years.
Anyway there is a tiny chip in one corner but the die does not look cracked, so I completed the CPU modification in full & I have it on test again checking for errors & all functions/features of the CPU are working.

 

[ThrashZone]

I have found the problem. It looks like I chipped the edge of the die when testing different heatsink, but it may have also been there before & I may of not notice it. I only carried out a DMM test for the first time due to CPU hardware changes, so it may have been present there for many years.
Anyway there is a tiny chip in one corner but the die does not look cracked, so I completed the CPU modification in full & I have it on test again checking for errors & all functions/features of the CPU are working.

Hi,
Didn't even take an image of the damage :/

 

[Nuckles56]

As a former machinist I have to say getting the top of a heatspreader anywhere near close to perfectly flat by hand is impossible. At one time years ago I discussed the idea of using a milling machine on an AMD FX-4000 with my former supervisor. We checked to top with a Starrett dial indicator with the CPU mounted in a test fixture for checking specs on small precision parts. The variance was so tiny we concluded it'd be a total waste of time and effort. You'll never come close to what a milling machine can do by hand. Also, thin surfaces flex under pressure and they don't do it evenly.

My experience also disagrees with this, I've flattened surfaces to very impressive tolerances with just sandpaper and a sheet of glass by hand and a lot of time. I've never done it with a CPU but there's no reason it wouldn't be the same with them as long are you're careful.

 

[delshay]

Hi,
Didn't even take an image of the damage :/

I was thinking of uploading image to the GHETTO THREAD as the modification does not belong in this thread. If & when uploaded you can see everything. Modification is something never been done before to this extreme. So it will never be copied (one off).

CPU is working fine sofar, so the chip in the corner may have always been there.

 

[Deleted member 185158]

As a former machinist I have to say getting the top of a heatspreader anywhere near close to perfectly flat by hand is impossible. At one time years ago I discussed the idea of using a milling machine on an AMD FX-4000 with my former supervisor. We checked to top with a Starrett dial indicator with the CPU mounted in a test fixture for checking specs on small precision parts. The variance was so tiny we concluded it'd be a total waste of time and effort. You'll never come close to what a milling machine can do by hand. Also, thin surfaces flex under pressure and they don't do it evenly.

We always did it to

1. remove nickel plating
2. better mating surfaces ( sand the cooler and cpu plates )
3. show and tell.

Started de-lidding back in 09. I don't think I've sanded a cpu (IHS plate) since then. Direct Die, just about every single Cpu I've had since.

 

[lexluthermiester]

I have a good AIO, the Corsair H100x, and using TG Kryonaut TIM. There is plenty of ventilation also. I am running the AMD R7 2700 at 1.425v. Clocked at 4.2Ghz all cores. Temperatures are in the low 80's. I saw a video by Der8uer also where he lapped a cpu and gain about -5c! Thats major in terms of cooling for overclocking.

It would be easier and far more effective to drop the ambient temps in the room you run the PC in. Get a window mounted AC and crank it down to 55 or 60 degrees F then turn the fans on the cooler up a few notches to get a better cooling effect going on.

Lapping that CPU will have very little effect on the OC you're running.

 

[bobbybluz]

It would be easier and far more effective to drop the ambient temps in the room you run the PC in. Get a window mounted AC and crank it down to 55 or 60 degrees F then turn the fans on the cooler up a few notches to get a better cooling effect going on.

Lapping that CPU will have very little effect on the OC you're running.

I have central A/C and in the summer having my AIO cooled PC's in front of the A/C vents works great. If I want them cooler I turn the A.C temp down. Extremely effective. Now that's it's winter here I have the central heat set at 64 degrees F. My overclocked Xeons (and non-overclockable E5 2683 V4) and i7 Extreme Editions stay nice and cool under load even though my wife complains it's too cold in the house once in a while.