[SOLVED] Can you use Thermal Paste instead of Liquid Metal when delidding a CPU?

[Iver Hicarte]

Greetings!

I just wanted to know, when it comes to CPU delidding, do you really need to strictly use liquid metal or can you use regular thermal paste like the Arctic MX-4, or both? If both can be used, what's the difference? Is there a huge difference in temperatures when you use liquid metal instead of good ol' thermal compound?

From what I have heard from other forums, liquid metal degrades faster than regular thermal paste, so you will often need to delid the CPU to replace it, at least that's how I understand it, correct me if I'm wrong.

 

[lvt]

I think unless you have a copper IHS, it's not worth delidding the CPU just to replace the thermal paste.

 

[vov4ik_il]

Hi,

Nothing prevents you from using peanut butter or jelly. It is all about thermal conductivity. Copper seems to be the best among cheap metals. Just next to silver. Any non-metal-based material would be significantly less conductive (10 times or even worse).
The thermal compound is only there to fill the rough surface gaps. Liquid metal degrades faster when exposed to air.
I used it since when the lead is back on and sealed, there is no air access. It works and is stable for over a year now.

 

[Phaaze88]

Direct die cooling is a thing, but I can vouch that paste does not work as well for that application, yet a Kraken G12 + Asetek cooler on the gpu works fine with paste...
Both are direct die, but the Intel die(7820X) is smaller than the Nvidia(1080Ti) one.

NH-D15S / Eisbaer Aurora 360, with NT-H2, thermals get worse after a few days.
Those same 2 coolers with Conductonaut, thermals still holding after a few months.



The 2 key objectives of delidding are:
1)The TIM. The stock Intel paste apparently doesn't transfer heat well, but what it has going for it is longevity. Should be ok to use other pastes here.
2)The IHS, or I should say the black silicone used to seal the cpu increases the IHS's height over the die. Trimming it off entirely would improve contact between the die and IHS.

 

[geofelt]

What cpu do you have in mind?
The main purpose in delidding is to be better able to control the high temperatures you get with maximum overclocks.

If you need to ask the question, I would suggest a professional delidding service which should cost about $50.

 

[lvt]

Direct die cooling is the most effective method, we already did that long ago.

 

[CompuTronix]

Greetings!

I just wanted to know, when it comes to CPU delidding, do you really need to strictly use liquid metal or can you use regular thermal paste like the Arctic MX-4, or both? If both can be used, what's the difference? Is there a huge difference in temperatures when you use liquid metal instead of good ol' thermal compound?

From what I have heard from other forums, liquid metal degrades faster than regular thermal paste, so you will often need to delid the CPU to replace it, at least that's how I understand it, correct me if I'm wrong.

Thermal Interface Material or “TIM” (thermal compound) facilitates better heat transfer from one surface to another. TIM can refer to either paste (pTIM) or solder (sTIM) or liquid metal. In 3rd Generation mainstream processors, Intel stopped using solder and instead started using paste between the “Die” and IHS, which resulted in higher Core temperatures, thus creating "The TIM Problem”. Intel resumed using solder in 9th through 11th Generation processors.

Delidding requires that you use only liquid metal TIM between the Die and IHS. Paste (pTIM) will fail in a relatively brief period of time. Due to thermal cycling, a process known as “pump-out” will expel pTIM from between the Die and IHS, whereas liquid metal is very resistant to pump-out. Although Intel's pTIM is formulated to resist pump-out, it still degrades over time, losing its thermal bond with the Die. The most recommended liquid metal TIM is Thermal Grizzly Conductonaut.

Thermal conductivity is expressed in Watts per meter kelvin or W/mk.

For comparison, here’s a short list of TIM in order of thermal conductivity:


IHS to Die - Soldered CPUs

Indium .............................................................. 81.8 W/mk


IHS to Die - Liquid Metal

Thermal Grizzly Conductonaut ........... 73.0 W/mk

CoolLaboratory Liquid Extreme ........... 41.0 W/mk

CoolLaboratory Liquid Ultra .................. 38.4 W/mk

CoolLaboratory Liquid Pro ..................... 32.6 W/mk


IHS to Cooler - Paste

Thermal Grizzly Kryonaut Extreme ... 14.2 W/mk

Arctic Silver 5 .................................................. 9.0 W/mk

Arctic Cooling MX4 ...................................... 8.5 W/mk

Gelid Solutions GC-Extreme ...................... 8.5 W/mk


Silicon Lottery is a company that tests, bins and sells overclocked, delidded "K" CPUs. They also offer professional delidding services. They use Thermal Grizzly Conductonaut and claim it's rare that any processors are returned within 3 years due to loss of thermal performance. They also give the following figures on how much Core temperatures at 100% workload are improved by delidding:

10th Gen ............... Comet Lake - 5 to 12°C

9th Gen ... Coffee Lake Refresh - 3 to 7°C

8th Gen .............. Coffee Lake - 12 to 25°C

7th Gen .................. Kaby Lake - 12 to 25°C

6th Gen .......................... Skylake - 7 to 20°C

5th Gen ..................... Broadwell - 8 to 18°C

4th Gen ........... Devil's Canyon - 7 to 15°C

4th Gen ....................... Haswell - 10 to 25°C

3rd Gen ................... Ivy Bridge - 10 to 25°C

Except for 9th through 11th Generation, Core temperatures on processors with solder between the Die and IHS are typically within about 5°C above IHS temperature, which indicates good thermal conductivity. However, Core temperatures on processors with paste between the Die and IHS can vary up to 25°C above IHS temperature, which indicates poor thermal conductivity and inconsistent uniformity.

Here's how "The TIM Problem" looks on a graph:

Although 9th through 11th Generations use sTIM, the solder is much thicker than earlier soldered processors, which decreases thermal conductivity. Additionally, the Die in 9th Generation is twice as thick as earlier Generations, which further decreases thermal conductivity. Here’s a detailed explanation by Mechatronics Engineer, Roman “der8auer” Hartung:

Core temperatures and IHS temperature converge at idle and diverge as load increases. Here’s how solder and paste differ between idle and 100% workload:

Thermal behavior is relatively uncompromised at idle due to low Power dissipation. But as workload approaches 100%, poor thermal conductivity among processors with paste becomes apparent. Moreover, as Intel's pTIM degrades over time, 3rd through 6th Generation 22 and 14 nanometer processors, (launched 2012 through 2015), may no longer cool as well as when new. Delidding restores and upgrades thermal performance similar to that of earlier soldered processors.

CT

 

[Phaaze88]

@CompuTronix
So it's just using paste under the IHS, period as the problem? That's something I still didn't catch on to.
It doesn't matter if it's NT-H2, Kryonaut, or GC-Extreme?

I already know first hand the conclusion of paste V liquid metal with direct die on smaller chips... so this doesn't change even with delidded chips?

 

[CompuTronix]

That's correct. Several years ago, prior to the arrival of Thermal Grizzly products, I conducted a series of delidding tests spanning over six months, where I tested about a dozen or so various thermal compounds, including NT-H1 and GC-Extreme, then compared the results to liquid metal. Testing was consistently conducted at 22°C standard ambient temperature with the same cooler using Prime95 v 26.6 Small FFT's (no AVX) for its steady-state 100% workload to eliminate variables and validate my findings.

The phenomenon of "pump-out" has been discussed over the years in the forums of the foremost websites dedicated to overclocking. As per their conclusions, I found that certain brands of paste would indeed experience pump-out in less than a week, which was evidenced by Core temperatures that gradually increased. Of the thermal compounds I tested, although Arctic Ceramique 2 showed she slowest rate of pump-out, within 2 months it also failed to maintain Core temperatures.

I've been delidding since 3rd Generation Ivy Bridge and have always had excellent results, both in terms of decreased Core temperatures as well as longevity. I can vouche for Silicon Lottery's claim that it's rare for them to have any of their delidded processors returned to them within 3 years due to loss of thermal performance.

Provided that the delidding and relidding procedures are performed very carefully with attention to detail, where all bonding material is completely removed from the Substrate and IHS, the correct amount of liquid metal is applied to both surfaces, and the IHS is reattached with a very thin application RTV Liquid Gasket and left to dry overnight, then results will always be consistently impressive, with an average decrease in Core temperatures of about 18°C.

I have older delidded processors that were deployed over 5 years ago which are still running at the same Core temperatures.

CT

 

[vov4ik_il]

Just to give a reference point to the numbers above, the thermal conductivity of copper is about 400 W/mk at CPU working temperatures (measurement varies with temperature).

 

[Iver Hicarte]

What cpu do you have in mind?
The main purpose in delidding is to be better able to control the high temperatures you get with maximum overclocks.

If you need to ask the question, I would suggest a professional delidding service which should cost about $50.

I have been planning to delid my i7-7700K due to its abnormal full load temps, it reaches 100 degrees. The degree mark where Intel says the chip might die (not really die, but rapidly shorten its lifespan for sure and in the process, eventually die) and throttle. Almost bought a new cooler because of this issue, but thank God I discovered CPU delidding, I was still hesitant to buy a new cooler since I'm not ruling out the chance that the issue might be caused by a small fixable problem, and ever since discovering CPU delidding I'm confidently sure this is a workaround I can do to fix the issue and make the most out of my CPU since I don't plan on upgrading my beloved system. CPU delidding will surely solve this issue. Since my PC is always well maintained and thoroughly cleaned, I also apply quality thermal paste within the components such as the GPU and the CPU. And I also found out that 7700K users have a common practice delidding that CPU because it runs very hot after years of use, I have mine for 4 years now so CPU delidding WILL solve the issue.