3570K @ 4.2GHz - Delid Temperatures?

GammaBreaker

Distinguished
Jan 19, 2012
16
0
18,510
[Hardware list at the bottom]

Hello!

I recently got an unstable 3570K, and was able to get a replacement that worked much better - 0.15V and 10C lower at the same clocks. During LinPack (IBT, specifically) At 21C ambient, the hottest core was 71C, average 68C across all cores.

The friend that I am building this for lives in an area where the climate can be upwards of 43C. That's a pretty big jump from my testing conditions, and I was concerned that the long-life use would suffer from the additional temperatures I couldn't test under. So I delidded it with the hammer & block method, applied a slim coat of CL Liquid Ultra to both the IHS and CPU die, then resealed it with Permatex.

The good news is that it appears to be fully operational, so I didn't damage anything. The temp drop wasn't as steep as I'd hoped, though. The hottest core was 62C under LinPack/IBT, average 60C.

This is my first Ivy Bridge build, so I'm not sure what to expect from the temperatures @ 4.2GHz. My main concern is that I did not use enough Liquid Ultra. A few people have mentioned that the point of the "brush coat" application is so that people do not slather it on and drip it on components. Since this is gasket sealed inside the die, that isn't an issue. A few people that have used thicker coats of Liquid Ultra report a notable additional cooling improvement.

The PC does need to last 3-5 years. It's a very large purchase for him, as he does not have a lot of money, so it needs to last (barring general electronics failures that happen). Considering that another 22C ambient will be added where my friend will use the PC, are those kinds of temperatures going to be detrimental?






CPU: i5 3570K [New]
MB: Gigabyte Z77X-UD4H [New]
RAM: G.Skill F3-12800CL9D-8GBRL [New]
HSF: GeminII S524
HDD: WD Caviar Black, 1TB [Lightly used]
PSU: Corsair TX650 V2 [Lightly used]
GPU: RHD 5770 [Lightly used, will be replaced with a non-bottleneck solution as funds permit]
Case: Antec 300
Optical: ASUS DVD+/-RW
 
Solution
Dr. Stick is in,

Good job delidding! I prefer the razor method myself but still good work.

I am 95% sure that you actually use too much CL TIM. Think about how much you would need to cover the die, divide that by 4, and that is how much you actually need. The best way is to take the little brush and stick it into the tip of the syringe so you have just a tiny bit. This should undoubtedly spread across the die. Then use half of what you just used (same method, stick the brush in the tube, and paint it on the underside of the IHS where it makes contact with the die.

Do not reseal with adhesive. If you absolutely must reseal because your friend/customer is a tinkerer and is going to be messing with the chip (taking it in and out of the...

thasan1

Honorable
Mar 27, 2013
1,363
0
11,660
ive got a question here, are you giving away your rig to your friend? or the rig is your friends and you were testing / oc'ing it?

as for the temps. seeing as at 21C ambient your temps barely exceeded 60C at your friends house it shouldn't go over 75C . though intel recommends temps under 70C, 75C wont be much of a problem, and it should last for 3-5 years easily with that temps. but if temp is around or over 80C you/ he can downclock the CPU to its stock configurations. that should keep the temp under 70C on load with that heatsink.


PS: sorry for my bad english :bounce:
 

GammaBreaker

Distinguished
Jan 19, 2012
16
0
18,510
Hello! Thank you for your response.

The rig isn't being given away, it's actually assembled out of about one third my spare parts, one third his spare parts, and one third new parts (RAM, Mobo, CPU) that he has paid for. I want to make sure he gets a good value for his money, so I gave it a mild overclock of 4.2GHz. Hopefully not enough to be stressful to the machine, but a notable boost.

It sounds like the temperatures should be okay. It's currently on hour 16 of Prime95 custom blend, average temperature is 59C, so lower than Linpack/IBT. Small FFT will be next.

That's a good idea to find a way to down clock just in case. I think I will set up a second profile in the BIOS for him to load if it becomes hot or unstable. Something with reduced settings.
 
Did you completely remove the original sealing or just refresh over it?

Although I have not delidded myself I have seriously researched into doing it.

From my discoveries the least amount of the liquid metal is in a lot of cases too much, so did you test and inspect the contact footprint you had under the lid with the die before resealing?

If you did not clean off all the original sealing, then your gap between the die and bottom of the heat spreader is now probably even greater than it was before the delidding.
 

GammaBreaker

Distinguished
Jan 19, 2012
16
0
18,510

Yes, the old sealing was removed as well, which was probably the most complicated part. I weighed the fresh sealant down with a 2LB lead weight (heaviest I had that wouldn't make it rock or crush the underside of the LGA). After it had cured for 4 of the 24 hours, leaving it in a semi-solid state, I clamped it down into the socket so that it would not move under the latch but would still compress under pressure from the heatsink. Then I left it to set for the remaining 20 hours before firing it up, lest the CPU's heat interfere with the Permatex's curing.

I made sure to clean the die and interior IHS quite thoroughly with my usual cleaning regiment of a AC 1 & 2, as well as 97% alcohol, with lint-free materials. The TIM under the IHS was...not like any thermal grease I've worked with before. It was half way between thermal slime and a thermal pad. Very rubbery, flexible, a tad stretchy. Much like the gasket material around the rim, but not adhesive. Very much like the thermal pads I had to use on an old 8800 GTS once.

I ended up having to do it twice, actually. The first time, I slipped with the weight and the IHS slid while the Permatex was still gelatinous. It gave me a chance to look at the footprint, which looked alright. I suspect the IHS is a tad convex on the inside, as the center of the footprint looked a bit more distinct than the top and bottom ends. So the second time around, I used a bit extra on the ends. Not too much can be done when the inside is the misshapen part (the exterior makes excellent contact with the GeminII).

It didn't get hotter, so it seems to be working, I guess I just expected a bigger drop from all the reports of 15-30C reductions.

I could do it again, as I have plenty of both Permatex and Liquid Ultra remaining. Though my friend will probably groan at another 24 hours of delay, better safe than sorry if you all recommend a third try.


EDIT: As a reference, it's currently testing on a CPU PLL of 1.550V and core voltage of 1.110V with a DVID of -0.010V. Much better than the first CPU that was pushing 1.250V @4.2GHz, and showing weirdness on one core.
 
Dr. Stick is in,

Good job delidding! I prefer the razor method myself but still good work.

I am 95% sure that you actually use too much CL TIM. Think about how much you would need to cover the die, divide that by 4, and that is how much you actually need. The best way is to take the little brush and stick it into the tip of the syringe so you have just a tiny bit. This should undoubtedly spread across the die. Then use half of what you just used (same method, stick the brush in the tube, and paint it on the underside of the IHS where it makes contact with the die.

Do not reseal with adhesive. If you absolutely must reseal because your friend/customer is a tinkerer and is going to be messing with the chip (taking it in and out of the socket) then this is what you need to do.

Go get yourself some loctite quickset epoxy. (Tiny tube of the clear stuff)

1) Mix up the two compounds of the epoxy on a clean surface or the tray they supply you.
2) Get a toothpick and put the smallest amount possible on 4 corners of the IHS's lip.

Allow 20 minutes to set. This will hold the IHS on.

The problem with the Ivy chips is more-so the adhesive than it is the crappy generic TIM. When they slather the adhesive on the chip it creates a gap filled with TIM between the die and IHS. You want the IHS to sit directly on the die with a thin layer of liquid metal. The IHS should theoretically be able to spin around like a top on the die if done correctly.

Basically what I'm saying is, don't use an adhesive, it is counter-productive.
 
Solution

GammaBreaker

Distinguished
Jan 19, 2012
16
0
18,510
Thank you so much for chiming in! While I somewhat loathe doing it again (the GeminII requires the board be taken out to mount properly), that's great detailed info. Though I was extremely gentle with droplets of the Permatex, it is so very viscous, I can see where even under compression it would increase the gap between the IHS and die. I work with epoxy every day, so I know just how much thinner it is than that rubbery Permatex.

My friend is not a tinkerer (quite the opposite, really), but the reason I wanted it lightly adhered was to keep it from sliding during shipping. UPS can be brutal on these packages sometimes. I will follow through with the dabs of epoxy. I think the only kind I have is the 2-ton/24 hour Devcon though, if that's also acceptable. The only concern left is: in the highly unlikely event that I do need to pop the lid one more time, will the epoxy let go with some proper coercion like the gasket, or will it take a piece of the PCB with it?

The razor method seems a lot cleaner, but...let's just say that I don't have the best luck with edged instruments. Mine came off with all of three gentle taps, thanks to the good vice we have at the shop. Didn't even mar the IHS.

Again, I appreciate the help. Done lots of tinkering over the years, but never have I deigned to pop a heatspreader off (mostly because they were soldered). It's new territory for me.
 
On my 4th delid I was in the process of doing the hammer and vise method and it dinged up my IHS pretty bad and I never got the IHS off (probably wasn't hitting it hard enough). So I just finished it off with a razor.

I would use a weaker epoxy like the loctite so it doesn't damage the PCB. You can do the same thing with RTV sealant I suppose. Just put a few dots on the lip of the IHS and it should hold fine. Also you might not even need a sealant if he's not ever going to remove the chip. That CPU latch clamps the IHS down so tight that I don't see it being an issue.

But one things for sure, depending on your voltage, that is way too hot. I was using a CM 92mm Vortex cooler on a 3770K for a while and I was getting those same temps at 4.6GHz (1.31v)

4.2GHz can be had with most 3570Ks at 1.15v or less.
 

GammaBreaker

Distinguished
Jan 19, 2012
16
0
18,510
The current one is at an effective 1.100V after the negative DVID. The temps seemed high, just like you're saying, for that low a voltage. I know the GeminII isn't the most potent cooler, but it still does moderately well. So it seemed off.

The gap increase from the gasket would make sense, as well. Enough though I have lower load temps, something that's been bothering me during re-testing has been higher idle temps. If the metal is expanding ever so slightly during high heat, but contracting and opening the gap when it cools off, might that explain it?

It has to cool off for a bit before being opened back up, as it's just finished a 24hr Prime marathon.

Edit: Got it out. I can see now, that even after compression, there is most assuredly a gap from how thick the Permatex is.