Core i7 4770K Haswell Delidded, 17°C Cooler Under Load With Other Paste

I am only just back from my vacation trip to Tokyo, Japan and already get this very nice email in my inbox. An enthusiasts going by the name "瀬文茶" has tested the temperature difference when delidding and replacing the TIM of a Haswell CP. The process is similar to what Der8auer has done with Ivy Bridge a couple months ago (forum thread). The results are rather impressive. With a Core i7 4770K clocked to 4.4GHz with 1.3V, the temperature under load goes from 90°C to 73°C when swapping out the stock TIM with Cool Laboratory Liquid Pro (82.0W/mk). That's a difference of 17°C! With the Prolimatech PK-3 (11.2W/mk), the result is only 7°C better, which is still quite good of course.

The process of removing the heatspreader does not really seem that complicated. Watch out for the small components close to the die, and perhaps add a little sticker over them to ensure they won't break off.

Source: Akiba-pc.watch.impress.co.jp


14

Belgium Massman says:

Anyone want to try?


United States Redmax says:

I'll be doing this over the weekend.

TaPaKaH says:

overclockers.ru also did a similar article - linky
before - after

Iran Mehdi_FXX says:

why intel don`t use a good paste?!

websmile says:

Mehdi_FXX said: why intel don`t use a good paste?!


To raise their profit :)

Germany der8auer says:

It's a good gain for air or water cooling but I don't see a point why you would do it using LN2. We're not temperature limited like on ivy bridge.

Ukraine IvTK feat. wolfsanek says:

Coollaboratory Liquid Pro - 32.6 W/m*K (earlier written by der8auer)

Belgium Massman says:

der8auer said: It's a good gain for air or water cooling but I don't see a point why you would do it using LN2. We're not temperature limited like on ivy bridge.


Well ...

There is always a difference between HS temp and die temp, especially under load. If the HS has bad contact, under load the die temp would rise faster.

South Africa Vivi says:

Massman said: Well ...

There is always a difference between HS temp and die temp, especially under load. If the HS has bad contact, under load the die temp would rise faster.



tested GC-extreme and LIquid metal on ln2 at -130, i lost about 50mhz overall.

I think we need to find paste that cures, for some reason its always fine with the intel paste but not others..

United States hokiealumnus says:

Mehdi_FXX said: why intel don`t use a good paste?!


According to very knowledgeable people at a very large motherboard partner, it's not the TIM. The TIM is fine. Probably not as good as Cool Laboratory Liquid Metal, but fine. It's the gap, introduced by the silicone sealant that holds the IHS onto the CPU package, that's the problem. Once the gap is eliminated, the temperatures are reduced.

Romania Alex@ro says:

Delidded 4770K,replaced stock tim with liquid pro(4 degrees better than ultra by my tests),a good friend(poparamiro) helped me with lapping the ihs till copper,sticked back with black silicone,gained about 30 degrees in real-temp.Still a very hot chip,before 4500 with 1.2 lynpack 95,now 65 on top air cooling.

Liquid tim is best tim for sure,don't know about it's behavious after -100,but i gained about 150 mhz more on dice(-60) on a 3570K ivy bridge.








Italy Atlas Rush says:

looking forward to someone delidding and freezing his CPU, since if it gives benefits even under LN2 i'll delid as soon as the closest shop sends me a liquid pro :)

Afghanistan AsadSAAD says:

great job. Intel must be think about that.

Singapore Spyboy48 says:

I will delid it too...today!

Please log in or register to comment.

Leave a Reply: (BBCODE allowed: [B], [QUOTE], [I], [URL], [IMG],...)