delided 4770k - THE thermal paste testing - die/ihs/hs

[Boulard83]

There are the result i have come with after a few days of playing with my delided 4770k. I dont have any Noctua paste left to test the CLP or CLU on die then Noctua on IHS/HS but after some testing ive found that the Phantek paste perform similar to the Noctua paste on the IHS/HS.

I'm no done yet but this is what i can show you right now.

As for now, CLP on die and some good paste on the IHS/HS seems to give me some really good result.

To be continued ...

 

[TuxDave]

Nice. I don't suppose you happen to have a baseline number?

 

[Boulard83]

Nice. I don't suppose you happen to have a baseline number?

baseline number ? You mean a non delided screenshot ? No sorry, i dont have a similar before delid SC if this is what you want.

 

[SunnyD]

This whole de-lidding thing... you de-lid it, clean the crap out of it, put new goop on it, then put the IHS back on it?

Other than the "convenience" of maintaining mounting specifications, what the hell is the point? You delidded it... get proper measurements and put your heatsink right on the damn chip!

 

[Vesku]

With some heatsink mountings it will be a lot easier to deal with a gap reduced heatspreader than a direct die. Just reducing the gap should free up plenty of thermal headroom for the majority of Haswell Ks that don't have good thermals stock.

 

[TuxDave]

baseline number ? You mean a non delided screenshot ? No sorry, i dont have a similar before delid SC if this is what you want.

Yeah, just to have the traditional "before" vs "after". It's a good study nonetheless.

 

[Boulard83]

With some heatsink mountings it will be a lot easier to deal with a gap reduced heatspreader than a direct die. Just reducing the gap should free up plenty of thermal headroom for the majority of Haswell Ks that don't have good thermals stock.

+1. Most of the "temp issue" with ivy and haswell come from the gap the black glue is adding between the die and IHS. The manufacturing process need to be quick so not all CPU end up with the exact ( perfect ) same gap. Some have worst temp, some have good temp. Add to this your luck on the chip you have and you can end up with a pretty toasty chip if you are that much unlucky !

Removing the glue, apply better paste and voila ! You have a good 10 to 20c drop from stock on pretty much any IVY/HW.

Yeah, just to have the traditional "before" vs "after". It's a good study nonetheless.

Thanks. Sorry for this missing screen

I Will update with more paste soon.

 

[Yuriman]

This whole de-lidding thing... you de-lid it, clean the crap out of it, put new goop on it, then put the IHS back on it?

Other than the "convenience" of maintaining mounting specifications, what the hell is the point? You delidded it... get proper measurements and put your heatsink right on the damn chip!

In addition to what others have said, you can also run bare-die once the lid is off.

It's good for a few °C over just getting rid of the gap between the die and the IHS.

 

[hawtdawg]

This whole de-lidding thing... you de-lid it, clean the crap out of it, put new goop on it, then put the IHS back on it?

Other than the "convenience" of maintaining mounting specifications, what the hell is the point? You delidded it... get proper measurements and put your heatsink right on the damn chip!

for reference, my 3770K went from topping out at 88C at 4.5@1.3v down to 69C. It can make an absolutely huge difference even when leaving the IHS on it.

 

[Idontcare]

for reference, my 3770K went from topping out at 88C at 4.5@1.3v down to 69C. It can make an absolutely huge difference even when leaving the IHS on it.

And you can count on your CPU lasting much longer at that OC (4.5G/1.3V) with the lower temperature than at the hotter temperature.

TJmax is only a safe operating temperature if you aren't overclocking or overvolting.

 

[Face2Face]

TJmax is only a safe operating temperature if you aren't overclocking or overvolting.

I didn't know that.......:hmm: Granted, 105c is not a temp I use as a peak for testing (Mine tops out @ 90-92c for stressing) - I just got some CLU - Maybe it's time to Delid..

 

[taq8ojh]

You might be surprised how different results you would possibly get if you dropped those temperatures. My 3770K before delidding maxxed at around same numbers, and I couldn't get it stable (by my standards) no matter what I tried. With the same settings but 20°C cooler, it suddenly passed all I threw at it without errors.


...just sayin'.
cpt. Obvious

 

[Boulard83]

You might be surprised how different results you would possibly get if you dropped those temperatures. My 3770K before delidding maxxed at around same numbers, and I couldn't get it stable (by my standards) no matter what I tried. With the same settings but 20°C cooler, it suddenly passed all I threw at it without errors.


...just sayin'.
cpt. Obvious

+1 !

 

[Face2Face]

You might be surprised how different results you would possibly get if you dropped those temperatures. My 3770K before delidding maxxed at around same numbers, and I couldn't get it stable (by my standards) no matter what I tried. With the same settings but 20°C cooler, it suddenly passed all I threw at it without errors.


...just sayin'.
cpt. Obvious

Nice! Well I have two options - Keep my overclock and enjoy the decreased temps with a delid OR shoot for 5GHz. I am stable now @ 4.8Ghz @ 1.3v to my standards. Zero WHEA errors and Prime95 stable for 6Hrs. This machine is pretty much for gaming (2-3 hrs a week) and a little photo and video editing , so It will never see 100% load for 24hrs or more..

 

[cytg111]

You might be surprised how different results you would possibly get if you dropped those temperatures. My 3770K before delidding maxxed at around same numbers, and I couldn't get it stable (by my standards) no matter what I tried. With the same settings but 20°C cooler, it suddenly passed all I threw at it without errors.


...just sayin'.
cpt. Obvious

So how much larger headroom would you say that you got by dropping temps by 20 degrees ? 200-300Mhz ? I think it is extremely interresting to see if it is the same deal with haswell or that the same 20 degrees will yield a bigger headroom.

 

[taq8ojh]

I don't know, and I am too lazy to try, but I probably should.
I am also pretty sure I would need more serious cooling if I went above 1.2V. Of course, 1.140V for stable 4.4GHz looks somewhat promising. And since it's delidded, temperatures might actually not raise THAT much. Who knows.

 

[caution]

The Noctua paste works indeed great - but you'll lose performance after a month or so.

Been there, before going for CLU.

Very intersting graphs though!

 

[taq8ojh]

The Noctua paste works indeed great - but you'll lose performance after a month or so.

That's the opposite of what IDC wrote in his delid thread...

 

[pimpin-tl]

I ordered this stuff. Is this better than the Noctua stuff sent with my new Noctua cooler?

ARCTIC MX-4 (4g) Carbon-Based Thermal Compound, Non-Electricity Conductive, Non-Capacitive

 

[caution]

That's the opposite of what IDC wrote in his delid thread...

I misread the #1 post :$

Noctua's perfectly good for IHS-to-HS - and to my experience it's pretty good for GPU's too.

But when it come to die-to-IHS, it's performance deteriotates far too quickly.

 

[taq8ojh]

Noone in his right mind would use anything but Liquid Ultra for that purpose

 

[THUMP1er]

Will the IHS be "loose" when you put it back on? Or it will stick and be stable?

 

[Concillian]

Will the IHS be "loose" when you put it back on? Or it will stick and be stable?

silicon + liquid Ultra + IHS = IHS is "stuck" to the silicon, but it can slide around. It needs to be pried or pulled significantly to separate the two, but the IHS can slide around in-plane.

Once clamped in the socket, it's not going anywhere.

When you bring the socket lever down, it will slide the IHS, so you need to start with the IHS moved off-center if you want it to end up centered after everything is clamped down.

 

[pimpin-tl]

I ordered this stuff. Is this better than the Noctua stuff sent with my new Noctua cooler?

ARCTIC MX-4 (4g) Carbon-Based Thermal Compound, Non-Electricity Conductive, Non-Capacitive

Guess no one knows.

 

[caution]

MX-4 never gave me better result than Noctua's staff. We are talking 1-2C here, but still...

( on IHS tested, only )