[Ivy-Bridge] IHS removal extreme Test

[der8auer]

PC-Watch published a test few days ago about changing the TIM on Ivy Bridge CPUs. 5-10°C better was the result. But what about LN2? I tested it for you!

 

My test is based on the test of PC-Watch: IHS removal, change the TIM and glue the IHS back on. First I tested the CPU on air to see the changes. To show the full profit I used LN2 aswell.

 

Step 1: Pretesting on air

Setup:

Intel Core i7-3770K

GIGABYTE Z77X-UD3H

4 GB G.Skill 2133 9-10-9-28

Scythe Katana 3

Gelid Extreme

der8auer Fusion rev3

 

For the test on air I clocked the CPU to 4200 MHz at 1,224 Volt and tested for 20 minutes using Prime95 large FFTs. To avoid mistakes I mounted the Scythe Katana 3 times. After 20 mins the maximum temperature reached 83°C. I straight noticed that Core 0 was a lot cooler than the rest.

 

 

 

 

Step 2: IHS removal

Ivy CPUs are not soldered anymore as PC-Watch already showed. Using a razorblade the IHS was removed after a couple of minutes. Cleaning the chip I straight noticed that the stock TIM is realy bad quality and already completely dry. That's not how a thermal paste should look after few months.

 

 

 

 

 

 

Step 3: Change to GELID Extreme

GELID Extreme seemed to be one of the best pastes in the past for extreme overclocking. So I put a little bit in the middle and spread it over the chip.

 

 

Step 4: Glue the IHS back on

The IHS is not only protecting the DIE - it's also important for the correct mounting. So I decided to glue the IHS back on to the CPU PCB using some normal silicone glue. To get the correct pressure I put the CPU back in the socket and left it there for about 24 hours.

 

 

 

Step 5: Retesting on air

I used exactly the same settings and also 20 minutes of prime 95 to get a compareable value. The difference in cores 1-3 is about 5 kelvin. But core 0 is much cooler with about 8-9 Kelvin. I'm sure this is one of the reasons why the CPUs clock much higher with only core 0 active.

 

 

Step 6: Let's get the LN2!

I tested this CPU already last week on wPrime to get a comparable result. 6293 MHz were 1024m stable. More voltage did not result in higher clocks.

 

[hwbot=2281599]submission[/hwbot]

 

Now after using the GELID Extreme between IHS and DIE I was able to push the CPU to 6385 MHz. Almost 100 MHz more and ranked 4th here on HWBot

 

[hwbot=2286149]submission[/hwbot]

 

 

Benching wPrime32m I could push the CPU to 6437 MHz - ranked 2nd global!

 

[hwbot=2286147]submission[/hwbot]

 

Conclusion:

I would not recommend this to 24/7 users. 5-9°C more is not worth losing the warranty of a 300 € chip. However extreme overclockers can get some extra MHz changing the thermal paste

 

Note: This is just a shorted and translated version of my original Test at PCGHX Forums: http://extreme.pcgameshardware.de/extreme-kuehlmethoden/218944-extreme-test-wlp-wechseln-bei-ivy-bridge.html

Edited May 23, 2012 by der8auer

[Olesius]

oops, very good idea bro

 

nice job as always

[Vivi]

Thanks dude!

You just answered the biggest question,

 

to remove or not remove. I think yes

[cyclone]

to remove or not remove. I think yes

 

Nooo! You should have already got a lucky ~7.1-7.3 GHz chip, so extra 500 MHz will kill almost everybody inside the community!

[Christian Ney]

Very nice roman, thanks

[Eeky NoX]

Yup very nice test

 

But even if it's here on the bot, you should warn about such mod for non extreme guys like most sites does... (warn explicitly ^^)

 

Gives me some ideas btw

 

Thank you Der8auer

[Splave]

this is awesome nice work

[just_nuke_em]

Good work . So how far off is the die from the IHS? Did you do a mounting test to see how much paste is needed?

[der8auer]

Thanks guys

 

Good work . So how far off is the die from the IHS? Did you do a mounting test to see how much paste is needed?

 

I don't know whether this is just an issue with my IHS but the inner surface is not as planar as it could be. The edges of the DIE have a little more space to the IHS than the center. That's why I used "quite much" thermal paste. The result shows that it can't be that bad

[DOM.]

looks like ill be trying this

[sin0822]

Why kelvin?

[Hondacity]

kelvin is just another unit of temperature. nice presentation roman!

[Massman]

Using Kelvin makes the entire thing more scientific.

[xoqolatl]

Very nice! Now remove it again and try LN2 without IHS

[Massman]

Very nice! Now remove it again and try LN2 without IHS

 

I think with IHS is actually the best way because:

 

- the mounting mechanism presses on the IHS to make sure the CPU is perfectly mounted

- without IHS, you will need a perfect mount to avoid cracking and optimal contact

 

Imho, the way der8auer did it is the best approach.

[Christian Ney]

Yes, too much pressure on the die will crack it or the system will simply not boot. When I do overclock laptop chips using cold I always use a P4 IHS otherwise the system doesn't boot due to too much pressure on the die

Edited May 24, 2012 by Christian Ney

[McZonk]

I think with IHS is actually the best way because:

 

- the mounting mechanism presses on the IHS to make sure the CPU is perfectly mounted

- without IHS, you will need a perfect mount to avoid cracking and optimal contact

 

+ you have to remove the entire mounting system of the socket. Otherwise the pot would not even get in contact with the die, but would rest on the metal frame due to the reduced height of the CPU.

 

Moreover Intel just seems to have raised the forces for the ILM to a minimum of 317 N (maximum at about 600N !). For applying those forces a heavyweight monster-pot giving pressure onto a few square-centimetres of brittle silicon is not the best idea i think.

 

@Roman: talked about the topic at the EOS - few days later it is done Thanks for sharing your experiences with lots of pics!

 

Edit:

Why kelvin?

Since in sience temperature differences are always indicated by using the kelvin-scale. (fyi: 1 degree celsius equals 1 Kelvin)

Edited May 24, 2012 by McZonk

[der8auer]

Why kelvin?

 

Using Kelvin makes the entire thing more scientific.

 

Actually telling differences in temperature in Celcius is just wrong. Kelvin is the correct measuring unit for it That's how university changes you

 

@ Chris: Yea I know Just could not resist trying it.

[SF3D]

I did same kind of testing this week. Here is the article where you can check the pictures

http://www.sf3d.fi/artikkelit/Hy-dytt-k-Core-i7-3770K-l-mp-tahnan-vaihto

 

It had a huge impact to temperatures, which was a bit of a surprise to me. I did not expect that much difference on watercooling.

So 7Ghz runs are with replaced thermal paste?

[djzatorze]

If we are removing the IHS, we can make it better - I think polishing the inside of IHS to the mirror would decrease the temperature a little more. And maybe Coollaboratory Liquid Pro would be better than Gelid Extreme in such place?

[Splave]

[riska]

Here is my chip..

 

It dropped 10c in idle from 39 to 29c at 4.2ghz havent tested it further..

[der8auer]

Here we go

 

Nice work guys!

 

riska it will be interesting how liquid metal performs sub zero.

[ME4ME]

Liquid metal is no-go for subzero, when directly in contact with the die. Between metalls, down to -100 it is superior, but then stiffens up en give terrible results.

[der8auer]

I also thought that. So GELID Extreme should be good enough.