Planning to lap (sand) both i9-9900K cpu die and Der8bauer OC frame for direct die cooling

[DLoa]

My i9-9900k cpu is delided for a direct die cooling setup using Conductonaute liquid metal, foam barriers and Der8auer OC (9th generation) frame in my Eurocome Sky X7c Clevo chassis laptop, which has a unified RTX 2080 GPU - i9-9900k CPU heatsink.  I used a vaccum chamber 3mm thick on top of my cpu die.

 

The  thickness of the intel IHS top is 2.30mm after the STIM has been removed.  The effective STIM tickness is assumed here to be around the reported 0.2mm since I lost the measurement made before delidding.  This means that the original the z-height before deliding was about 2.5mm and therefore using a 3mm tick vaccum chamber increased it by up to 0.5mm. 

Note that, although the temps are fine after deliding, the GPU VRM temperatures are not known and I am conscerned this results in poor contact with heatsick or would require in places much thicker thermal pads rendering less efficient in hear transfers comparared to original setup.

 

Project:

In order to offset this 0.5mm increase in height and to improve both cpu die heat transfer (reduction of silicon top layer) and the contacts tolelerances with the cooling unified system, I am planning on lapping the CPU die using NLap tool by NUDEcnc and checking thichness with micrometer during sanding passes (400-2000 grit).  With a cpu die 0.88mm above PCB and a tolerance of 0.1mm of OC frame below below the silicon die, would require that the OC frame base be sanded down by the same amount in order to preserve optimum direct die contact.  

 

Question:

Roman / Der8auer initially reported temps after sanding/lapping down by 0.2mm but mentioned that the ticker silcon would be sanded down further 0.4 or more.    

Anyone has tried sanding their cpu die down by 0.4mm or more?  Would 0.45-0.50mm be accpetable?

 

Thank you for your inputs.

 

[Papusan]

On 3/15/2022 at 4:52 PM, DLoa said:

Question:

Roman / Der8auer initially reported temps after sanding/lapping down by 0.2mm but mentioned that the ticker silcon would be sanded down further 0.4 or more.    

Anyone has tried sanding their cpu die down by 0.4mm or more?  Would 0.45-0.50mm be accpetable?

 

Thank you for your inputs.

A friend on old NBR (@jc_denton) sanded his 9900K below what der8auer did. End results was a cracked die after he started use the laptop. He had an Clevo P870Tm. Don't go too far with the sanding. As a side note. He also used a Vapor chamber between Heatsink and die.

[DLoa]

@Papusan thanks for the input and the reminder on @jc_denton mishap.  I was in touch with JC_Denton on the NBR forums but lost touch since it closed down.  I hope he will join here as well as Meaker@Sager.  I was plannning on 0.4mm do you remember how much JC_Denton went had removed when it had cracked?

 

The vaccum chamber was added in my system following his example and shared discussion on the NBR thread "P775TM1-G and I9 9900K high temperature problem!" that I fully backed-up to pdf.

 

Any experience sanding down the OC frame to keep the height differential with top of cpu die after lapping?

 

Thanks

[cylix]

This is mine. Works well for over an year with nno problems

[DLoa]

@cylixit looks great.  Is the measurement showing that you removed 0.3mm?

 

Noob question is it ok to use a digital caliper instead of micrometer with the metal parts touching the silicon and the contact on the cpu base?

 

I ordered the Nlap DIE and IHS from Nude CNC for a precise sanding / lapping of my delidded CPU.

 

I suppose that I'll have to also sand the Der8auer direct die frame by the same amount.  Did you have to also?

 

Thanks for feedbacks.

 

[Meaker]

Yes the measurement device is not usually clamped, just lightly touching.

[ryan]

I have an old trick to lower temperatures dramatically, I simply apply tooth paste(mint flavour) on the die(pea method of course) then crush a mentos on it. usually drops the temps by 40-50c similar to nitrogen cooling. on a serious note, whats the best  non conductive paste. And does anyone know of a chart showing all the pastes tested, I used to have one.

[DLoa]

On 7/23/2022 at 11:18 PM, Meaker said:

Yes the measurement device is not usually clamped, just lightly touching.

@Meaker and @cylix

The dust settled down after a busy time the last few months relying daily on my system.  I am now ready to lap my delidded i9-9900K CPU,  which is direct die mounted with a vacuum chamber and OC-frame on the unified cooler of the  P775TM1-G clevo chassis - all with a very thin layer of liquid metal.  I will follow the Der8auer Youtube video below where he sanded 0.4mm off. 

https://youtu.be/O1ed_rBRb7Q?t=515

 

Here are a three points to clarify before I get started:

8:35   Der8auer recommends taking measurement and not just rely on the NLap CPU tool.  The micrometer he used has a 0.001mm resolution and lightly touching the measurement point.  I will also use the tool to lap my cpu at most 0.4mm (and to take measurement with cpu tacked in).

@MeakerDoes it work as well to use instead a caliper with only 0.01mm resolution to track the amount of silicon effectively removed by lightly measuring between the top cpu and the NLap tool underside?

Should I plan to sand the base or the top of the der8auer direct die Intel 9th Gen. OC-Frame (i.e. the top of the anodised aluminium fame facing the sanded cpu silicon side) by the same amount (or more)?  (Top thickness thickness and height differential with pcb concerns me).
If it is also the top, will nail varnish or silicone modified conformal coating work fine to protect it against corrosion from liquid metal contact -and- will the thickness and texture not be an issue for the frame to work as designed ?

Thank you for your inputs.

 

 

[jaybee83]

20 hours ago, DLoa said:

@Meaker and @cylix

The dust settled down after a busy time the last few months relying daily on my system.  I am now ready to lap my delidded i9-9900K CPU,  which is direct die mounted with a vacuum chamber and OC-frame on the unified cooler of the  P775TM1-G clevo chassis - all with a very thin layer of liquid metal.  I will follow the Der8auer Youtube video below where he sanded 0.4mm off. 

https://youtu.be/O1ed_rBRb7Q?t=515

 

Here are a three points to clarify before I get started:

8:35   Der8auer recommends taking measurement and not just rely on the NLap CPU tool.  The micrometer he used has a 0.001mm resolution and lightly touching the measurement point.  I will also use the tool to lap my cpu at most 0.4mm (and to take measurement with cpu tacked in).

@MeakerDoes it work as well to use instead a caliper with only 0.01mm resolution to track the amount of silicon effectively removed by lightly measuring between the top cpu and the NLap tool underside?

Should I plan to sand the base or the top of the der8auer direct die Intel 9th Gen. OC-Frame (i.e. the top of the anodised aluminium fame facing the sanded cpu silicon side) by the same amount (or more)?  (Top thickness thickness and height differential with pcb concerns me).
If it is also the top, will nail varnish or silicone modified conformal coating work fine to protect it against corrosion from liquid metal contact -and- will the thickness and texture not be an issue for the frame to work as designed ?

Thank you for your inputs.

 

 

 

best of luck bud! keep us updated on your progress and be sure to follow up with pictures of this whole undertaking, definitely super interesting stuff! 🙂 

[LasseK1981]

Hello.

I just wanted to share my sanding experience with you guys. I had a 13900K which was delid and ran great with a lapped IHS and liquid metal. I then wanted to proceed to direct die cooling. I wanted to sand the die only to make it 100% straight or even. I taped the substrate with tape leaving the die to be sanded. I sanded the die only to like you could see it all sanded only surface sanding. After my CPU dosent work anymore. There are no visible damage to the cpu and all caps are there. I dont understand it. Its just dead with a motherboard error 00. I dont know if I pressed too hard when sanding, I did it in hand so to speak. pushing on the back of the CPU. Again no visible damage but CPU DEAD! 😞 I may have pressed to hard but I feel I did it correctly!?  This was a very costly mistake, and I ordered a new 13900KS. I must admit that now im scared to do the normal delid on the KS even though before I was fearless.... This is extremely expensive toys. 😕  I have the EK Direct Die waterblock so I will use it. To begin with just stock CPU with a Thermalgrizzly mounting frame to see everything runs ok no error before delid. 

 

I will not attempt to sand die again. 🙂    I guess you have to be very careful and do it with absolutely no force. 

[Meaker]

On 3/12/2023 at 3:41 PM, LasseK1981 said:

Hello.

I just wanted to share my sanding experience with you guys. I had a 13900K which was delid and ran great with a lapped IHS and liquid metal. I then wanted to proceed to direct die cooling. I wanted to sand the die only to make it 100% straight or even. I taped the substrate with tape leaving the die to be sanded. I sanded the die only to like you could see it all sanded only surface sanding. After my CPU dosent work anymore. There are no visible damage to the cpu and all caps are there. I dont understand it. Its just dead with a motherboard error 00. I dont know if I pressed too hard when sanding, I did it in hand so to speak. pushing on the back of the CPU. Again no visible damage but CPU DEAD! 😞 I may have pressed to hard but I feel I did it correctly!?  This was a very costly mistake, and I ordered a new 13900KS. I must admit that now im scared to do the normal delid on the KS even though before I was fearless.... This is extremely expensive toys. 😕  I have the EK Direct Die waterblock so I will use it. To begin with just stock CPU with a Thermalgrizzly mounting frame to see everything runs ok no error before delid. 

 

I will not attempt to sand die again. 🙂    I guess you have to be very careful and do it with absolutely no force. 

 

It can be a micro crack, that sucks 😞

[kojack]

On 8/12/2022 at 10:00 PM, ryan said:

I have an old trick to lower temperatures dramatically, I simply apply tooth paste(mint flavour) on the die(pea method of course) then crush a mentos on it. usually drops the temps by 40-50c similar to nitrogen cooling. on a serious note, whats the best  non conductive paste. And does anyone know of a chart showing all the pastes tested, I used to have one.

All the good pastes are within a degree or two of each other. So any good one will work.  I am using Noctua NT-H2 with great success. From stock, my temps are down alot.  could they go down and extra deg using liquid metal?  probably.  Is it worth it for all the effing around most definitely NOT.