Bare-die testing: A delidded 3770k, an H100, and 9 different TIMs

[tw33k]

Finally got to delid my 3770K and the results are fantastic. I used Liquid Ultra on all 4 surfaces. CPU is @ 4.6GHz 1.31v

Before

After

The delta temp dropped 22.5c!

 

[Idontcare]

Finally got to delid my 3770K and the results are fantastic. I used Liquid Ultra on all 4 surfaces. CPU is @ 4.6GHz 1.31v

Before


After


The delta temp dropped 22.5c!

That is awesome :thumbsup:

And I bet with your now lower operating temperature you will be able to decrease your Vcore while remaining stable (try ~1.23V), which will in turn further reduce your operating temperatures even more

 

[Lepton87]

I do. The 2600k would do 5GHz without missing beat. For what I do, I need 3-4 desktop computers so the 2600k was already rotated out of my "fun hobby computer" and over into a "it works for a living" computer, but I can certainly rotate it back if need be.

As for your question - SB or IB - once power usage is removed from the equation, if the individual is not using the iGPU then I'd recommend a SB no question. It is a no-fuss no-muss situation where hitting good overclocks on economical air cooling is completely doable.

Thanks for the info,I've just checked prices and it turns out that buying 3770k is rather stupid in my country because it's 15% more expensive than 2700K, and it's not even that much faster at stock no to mention OC.

BTW. Does NT-H1 suffer from pump-out effect at IHS to Heatsink contact? A few months ago I hit 87C degrees with my mild OC which already seemed very high and right now I hit 98C in IBT. My cooler is zalman CNPS 10X and TIM is noctua NT-H1, those temperature seem much too high I'm begging to suspect there's something wrong with my CPU's IHS. Either that or the cooler is not flat where it sticks to the IHS. It has two fans in push-pull configuration. AFAIR it was only 10C worse then the top coolers like nactua NH-D14.

 

[Idontcare]

BTW. Does NT-H1 suffer from pump-out effect at IHS to Heatsink contact? A few months ago I hit 87C degrees with my mild OC which already seemed very high and right now I hit 98C in IBT. My cooler is zalman CNPS 10X and TIM is noctua NT-H1, those temperature seem much too high I'm begging to suspect there's something wrong with my CPU's IHS. Either that or the cooler is not flat where it sticks to the IHS. It has two fans in push-pull configuration. AFAIR it was only 10C worse then the top coolers like nactua NH-D14.

Yes it does. But so will just about every other TIM out there other than IC Diamond. Liquid metal will, too, but being a liquid at all times means it just reverses the pump-out on cooling so it is irrelevant. But NT-H1 has been reported by a number of people as perhaps having the most pronounced pump-out issue.

 

[DrFezzik]

Just thought i'd toss in my experience. I have delidded 4 3770k's and the 4th one is the one I found to be the best. The first 3 I had to crank up the voltage to 1.45-1.5V to get 5 Ghz stable and even then I could not use all my 32 gigs of ram (8x4) even at 1333mhz. The 4th one I got I am stable prime and linx for 48 hours at 1.39V and my memory is at 2400mhz. with all banks populated. Now the temps on my cpu do not get high. I am custom watercooled and straight to the die with swiftech apogee HD. It works perfect. I put a half rice grain size of liquid Ultra and spread it out. My temps idle are between 29-35C. My water temps is at 28C On full load my temps only get to 68C on average and peak 75 sometimes but rarely. This was at 1.5v. At 1.39V, got the same idle, but I only peak at 63C on full load average 58C.
I tried the stock cooler for the heck of it with the IHS on it and the temps were getting near 100 degrees just the first few mins of prime, so I shut her down. I have the noctua NH-D14 and the temps were about 85-95 degrees at 1.5V straigt to die
This is the cpu delidded and with liquid ultra and the footprint it left on the apogee HD. The nice thing about the apogee HD with delidded is that with the way its designed with the springs it will put even pressure even delidded when screwed down. Great design and cant screw up.

Sorry for the hairy leg shot but you can see very little used. The smudge to the side is where I was removing it from the heatsink.

My case setup if you are curious

Great Job OP on creating this test. It's not easy doing what you did and took a lot of time!

 

[Idontcare]

Welcome to the forums DrFezzik :thumbsup: and thanks for contributing your results

 

[railven]

Figured I'd ask here:

I just got my GF a Ivy Bridge i5 3570K (can't beat Microcenter deals) and plan on OCin'g it. I'm not aiming for any stars.

I picked up an CM EVO 212+ and will aim for about 4.4ghz-4.6ghz range.

Should I delid the processor or are temps in that range acceptable? Since she is getting the chip my main concern is noise/heat vs performance. She isn't as anal as I am about performance haha.

Great thread IDC, really enjoyed reading it.

 

[T_Yamamoto]

You will be fine for a modest oc

 

[Idontcare]

Figured I'd ask here:

I just got my GF a Ivy Bridge i5 3570K (can't beat Microcenter deals) and plan on OCin'g it. I'm not aiming for any stars.

I picked up an CM EVO 212+ and will aim for about 4.4ghz-4.6ghz range.

Should I delid the processor or are temps in that range acceptable? Since she is getting the chip my main concern is noise/heat vs performance. She isn't as anal as I am about performance haha.

Great thread IDC, really enjoyed reading it.

Not sure how close an EVO 212+ compares to an NH-D14, but I was able to OC to 4.5GHz with my NH-D14 before delidding/lapping and the temperature was 82°C (1.18V).

So I have to imagine you can hit 4.4-4.6 with the 212+ while staying under TJmax.

The thing that is a real concern as you mention is the noise. The following is data for my 2600k, but it applies just as well to my 3770k:

Depending on your mobo's fan control profile, the noise will get noticably loud once your CPU's temperature crosses a threshold (70°C for mine).

That alone is where the benefit of delidding and reducing your peak temperatures by 20°C yields dividends.

 

[railven]

Not sure how close an EVO 212+ compares to an NH-D14, but I was able to OC to 4.5GHz with my NH-D14 before delidding/lapping and the temperature was 82°C (1.18V).

So I have to imagine you can hit 4.4-4.6 with the 212+ while staying under TJmax.

The thing that is a real concern as you mention is the noise. The following is data for my 2600k, but it applies just as well to my 3770k:

Depending on your mobo's fan control profile, the noise will get noticably loud once your CPU's temperature crosses a threshold (70°C for mine).

That alone is where the benefit of delidding and reducing your peak temperatures by 20°C yields dividends.

I'm pretty sure the CM 212+ isn't as good as the NH-D14, so this info is some good food for thought.

She's coming off a 930 i7, so even at stock clocks the upgrade should be noticeable, perhaps I won't go as aggressively on the clocks.

Perhaps I'll look into the delidding if she isn't happy with the noise levels after setting it up.

EDIT: Maybe I can be a weasel and talk her into swapping with me on the "I might break it if I screw up" argument haha.

 

[Ed1]

Yes, but what normal app will get those high temps, non probably can come close to prime95/IBT these are just not realistic test points for noise/operation results.
They are great for pushing thermals in stability tests but your running temps will probably be approx 10-15c below those prime95/IBT .

 

[Lepton87]

Yes it does. But so will just about every other TIM out there other than IC Diamond. Liquid metal will, too, but being a liquid at all times means it just reverses the pump-out on cooling so it is irrelevant. But NT-H1 has been reported by a number of people as perhaps having the most pronounced pump-out issue.

What do you recommend, should I reapply Noctua TIM after just 6 months? Obviously there is something wrong with my CPU and/or Cooler because those temps are way too high for a modest overclock on SB. Right now temperature hits over 70C even at stock settings (3.7GHz). That temperature would be fine for a stock cooler not a tower cooler that trails Noctua NH-D14 by 10C.

 

[Idontcare]

What do you recommend, should I reapply Noctua TIM after just 6 months? Obviously there is something wrong with my CPU and/or Cooler because those temps are way too high for a modest overclock on SB. Right now temperature hits over 70C even at stock settings (3.7GHz). That temperature would be fine for a stock cooler not a tower cooler that trails Noctua NH-D14 by 10C.

Yep, reapply. And clean out your HSF fins while you are at it.

If you don't want to deal with the rinse-repeat cycle down the road then you could go with some IC diamond and do the "once and done" dance. That won't prevent the dust build-up though.

 

[Lepton87]

Yep, reapply. And clean out your HSF fins while you are at it.

If you don't want to deal with the rinse-repeat cycle down the road then you could go with some IC diamond and do the "once and done" dance. That won't prevent the dust build-up though.

I just ordered a tube of IC DIAMOND 7 CARAT + is it the same TIM you're talking about? I can't find any other IC diamond TIM in my country. Doing something in my computer is such a pain, it's very cramped, 3 cards being the main culprit, I want to mess with it as little as possible. Thanks for the help. I'll post temps with IC diamond and see if that helped.
BTW. How much of that TIM should I apply on the IHS? Your pictures show bare die which is certainly different then IHS.

 

[Idontcare]

I just ordered a tube of IC DIAMOND 7 CARAT + is it the same TIM you're talking about? I can't find any other IC diamond TIM in my country. Doing something in my computer is such a pain, it's very cramped, 3 cards being the main culprit, I want to mess with it as little as possible. Thanks for the help. I'll post temps with IC diamond and see if that helped.
BTW. How much of that TIM should I apply on the IHS? Your pictures show bare die which is certainly different then IHS.

Yep, thats the TIM. The carat amount is just the volume you buy, 7 carat is a smaller tube than 24 carat, the diamond per volume is the same.

Apply the same as you would any other TIM. You can get too little though, so be careful about trying to spread it too thinnly.

These days these TIMs are all about finding the goldilocks amount - not too much and not too little - so a nice dollop is in order.

 

[Lepton87]

Yep, thats the TIM. The carat amount is just the volume you buy, 7 carat is a smaller tube than 24 carat, the diamond per volume is the same.

Apply the same as you would any other TIM. You can get too little though, so be careful about trying to spread it too thinnly.

These days these TIMs are all about finding the goldilocks amount - not too much and not too little - so a nice dollop is in order.

Thanks for the advice, that test with insanely high amount of NT-H1 showed that it's better to apply too much then too little.

 

[DrFezzik]

I would also suggest to not use ic diamond straight to the die. Would it not scratch it?

 

[aigomorla]

IDC can u do a Delta T of ambient.

Take another probe and take the ambient temperature at the air passing though the radiator in the time you take a load temp.
And give us a Delta T.

I would like to use your data, but would need it to be constant.

 

[Idontcare]

I would also suggest to not use ic diamond straight to the die. Would it not scratch it?

It does scratch it, here is my pristine die before tests:

And after testing:

I noted the scratches after the IC diamond test but did not take a good enough picture to highlight them until after the Liquid Ultra testing was complete.

They did not appear to effect cooling performance afterwards.

IDC can u do a Delta T of ambient.

Take another probe and take the ambient temperature at the air passing though the radiator in the time you take a load temp.
And give us a Delta T.

I would like to use your data, but would need it to be constant.

Checkout this post Aigo, all my data are already captured in that manner and normalized to 25°C ambient (footnote on all tables).

So you should be able to use the data to your hearts content. Any data you come across from me in the past 3 yrs has been collected and normalized in this manner, so you are good to go :thumbsup:

 

[DrFezzik]

I dont care. I'm about to do this to another 3770k but with the noctua nh-D14. I'll post some pics. I'm going to use your numbers. ALso there was a post on here about someone doing this on a noctua nh-d14 so I thought great it will seat perfectly on the die. I did a mark transfer to see how it would seat and guess what. No marks. If i put a little paste on there it will mark, so there is not a good contact. So The stock plastic washer are probably 3mm too high. I'm going to use washers and drop the distance 3mm since i is spring loaded and give it a try.

Edit here it is. I saw this was referenced on anandtech here to this website about this cooler and the delidded process.
http://www.overclock.net/t/1249419/pcevaluation-intel-i7-3770k-temperature-measured-without-ihs

 

[Idontcare]

I dont care. I'm about to do this to another 3770k but with the noctua nh-D14. I'll post some pics. I'm going to use your numbers. ALso there was a post on here about someone doing this on a noctua nh-d14 so I thought great it will seat perfectly on the die. I did a mark transfer to see how it would seat and guess what. No marks. If i put a little paste on there it will mark, so there is not a good contact. So The stock plastic washer are probably 3mm too high. I'm going to use washers and drop the distance 3mm since i is spring loaded and give it a try.

Edit here it is. I saw this was referenced on anandtech here to this website about this cooler and the delidded process.
http://www.overclock.net/t/1249419/pcevaluation-intel-i7-3770k-temperature-measured-without-ihs

Yeah you definitely have to replace the stock mounting shims that come with the NH-D14 with washers or springs.

Just be careful to not gouge your mobo with whatever you do put in contact with it. Even if using springs, I recommend putting a washer at the base (between the mobo and the bottom of the spring) and also putting tape on the bottom of the washer.

We don't want to have dead mobos from abrasion like we have seen dead CPUs from nicked PCBs.

Good luck :thumbsup:

 

[DrFezzik]

Yeah. i was going to put some electrical tape at the base to protect the MB, and use the washer method like you did on here. But if you look at the guy who did the test with the Noctua he's using the stock base plastic hubs. There is no way unless he modifyed it and there is no mention he did this.

Pic is from pcevga forum

 

[Idontcare]

Yeah. i was going to put some electrical tape at the base to protect the MB, and use the washer method like you did on here. But if you look at the guy who did the test with the Noctua he's using the stock base plastic hubs. There is no way unless he modifyed it and there is no mention he did this.

Pic is from pcevga forum

True, but unless I read the thread wrong his temperatures were crappy, and no doubt they were crappy because he left a HUGE gap between the bare die and the NH-D14, a gap that had to be filled in with TIM.

So no question you can make it work with the stock standoff, but that kinda destroys the point of removing the IHS in the first place though given that the IHS (being made of copper) is going to be far better as a thermal conductor versus a pile of TIM.

(for example, see this post and this post where a gap was intentionally engineered and explored)

 

[DrFezzik]

Again. thanks for doing this and finding the drop in losing the IHS being 3mm. Saves a bunch of time and headache.

 

[USER8000]

Absolutely fine. Noctua knows what they are doing and it shows whether you are looking at their NH-D14 coolers or their NT-H1 TIM. I even used NT-H1 on my GTX460 when I delidded it and remounted an Accelero XTREME Plus.

(I sound like a damned spokesperson for Noctua D: but I swears I'm just calling it like I sees it)

Thanks! So,the only time it has issues,is when you are using it on a bare IB die??

The reason I am asking longterm is because I tend to build small form factor systems,so I would prefer to not have to take the system apart every six months, to re-apply the thermal compound.

BTW,I have some of the IC Diamond stuff,and it is like spreading clay!