I'm no longer a believer of MX-4

[steve258]

So this month I got myself a nice shiny new i7 build with the Megahalems CPU cooler. As we all know that the Ivy's heat problem seems to be the number one limiter of overclock, and the only way to solve this it seems is to remove the IHS and replace the terrible thermal paste that Intel use.

So one night I finally worked up the courage to try this mod, and below are just some comparisons of temps both before and after, as well as between MX-4 and the Prolima PK-1 paste which was included with the Megahalems.

But before that, here is a shot of the delidded i7 3770K:

Before De-lidding
Thermal paste between IHS/Heatsink - PK-1

Screenshot taken roughly 10 minutes into running prime95 with AVX

After De-lidding (IHS still on)
Thermal paste between Die/IHS - MX-4
IHS/Heatsink - MX-4

Screenshot taken roughly 10 minutes into running prime95 with AVX, just to make sure it wasn't a messed up application I re-applied again with same result. IHS still on.

After De-lidding (IHS still on)
Thermal paste between Die/IHS - PK-1
IHS/Heatsink - PK-1

Screenshot taken roughly 30 minutes into running prime95 with AVX.


Arctic Clean 1&2 is used on all 4 sides (Die/IHS both sides/heatsink) before each application. Application is the same line method for all tests. All three tests are with IHS still on, just with different thermal paste.

Conclusion:
- Removing IHS and replacing the stock thermal paste really works, other forums all suggest similar results. It is relatively easy if you have the right tool (I just used a stationary cutter so it took me a bit longer) but obviously due to the voided warranty and the potential risks involved it is hard to recommend.
- I must be a semi-expert on thermal paste removal and application by now.
- Stulid's excellent thermal paste testing thread here suggests PK-1 to be slightly better, I just didn't realise it could win by such a large margin.

Also there is an excellent thread on IHS removal over at extreme forum, I don't think I'm allowed direct link but if you search for "3770k IHS Removals" the it should be at the 1st result.

Hope this helps someone.

 

[Admiral Huddy]

So the last screen shoot is without the IHS whereas the the first two are with?

I have been very keen on this idea for some time. However, I wasn't sure if the motherboard retaining plate would prohibit the HSF making any contact where the IHS is missing.

 

[steve258]

Oops I should have stated clearly all three are tested with IHS still on, the first is before delidding, 2nd and 3rd are after delid, but IHS is put back after thermal paste is replaced. so the temp difference between 2nd and 3rd is purely down to thermal paste used.

I have been thinking die direct contact as well. Socket retention bracket needs to be removed all together and modifications are needed for the Megahalems mounting bracket. Also Athlon-XP style foam pads need to be placed at the 4 corners of CPU to prevent crushing the die.

 

[Helios1234]

MX-4 was never that great. It's not even as good as MX-3 so the results aren't too surprising. An average 3 degree difference does seem quite high but if you were to see it as a gain of 1.5C between the die/IHS and another 1.5C gain between the IHS/heatsink, then it seems to make sense. Of course, that's assuming your ambient is constant.

 

[jaybee]

So you went from the original:
70, 74, 71, 69 (average = 71)

To:
65, 72, 73, 71 (average = 70.25)

Less than a single degree average temp lost. That's not worth voiding a warranty for. You should test it naked with no IHS since you have already gained access. This might show better results.

 

[steve258]

MX-4 was never that great. It's not even as good as MX-3 so the results aren't too surprising. An average 3 degree difference does seem quite high but if you were to see it as a gain of 1.5C between the die/IHS and another 1.5C gain between the IHS/heatsink, then it seems to make sense. Of course, that's assuming your ambient is constant.

I suppose, both tests were done last night so both results are maybe 1.5hrs apart. The first pre-delid screenshot was taken a few days before though.

So you went from the original:
70, 74, 71, 69 (average = 71)

To:
65, 72, 73, 71 (average = 70.25)

Less than a single degree average temp lost. That's not worth voiding a warranty for. You should test it naked with no IHS since you have already gained access. This might show better results.

Hmm ok maybe I should just put it down in numbers:

Before delid:
79, 83, 79, 76 Avg: 79.25

After delid, IHS on (MX-4):
67, 75, 76, 72 Avg: 72.5

After delid, IHS on (PK-1):
65, 72, 73, 71 Avg: 70.25

So I went from 79.25 to 70.25 just replacing the stock thermal paste, that's exactly 9 degrees difference.

Like I said, lots of mod needed for direct die contact. But from reading other people's results you can potentially shave additional 5C off for removing IHS all together.

 

[mattup]

Okay. Good temps, but how does it affect the max overclock?

If you can get to 5Ghz it might be worth the warranty issue but for a 10+ % performance increase is it really worth it?

Don't get me wrong I have a knife ready to go!!

Matt

 

[steve258]

Tbh realistically I'd say maybe 100-200Mhz extra on air depending on the cooling you have and the voltage required. Although if you have a decent WC setup that 9C difference is going to stretch a lot further.