[H] delided a Kaby Lake i7 7700K, the results are .. interesting

[Funtoink63]

I have found this video an thought it would be fitting to this thread:

Spoiler

At the end of the video, he's changing CPU clock speed on Win XP and this change shows up in Win 7. My only guess is that he's running two OS at the same time.

 

Would it be possible to overclock 7700k to 5.5+ Ghz (or any CPU to "extreme" frequencies) after replacing stock TIM with liquid metal and cooling it with "setup" like this?

[Lays]

5 minutes ago, Funtoink63 said:

I have found this video an thought it would be fitting to this thread:

 

  Hide contents

At the end of the video, he's changing CPU clock speed on Win XP and this change shows up in Win 7. My only guess is that he's running two OS at the same time.

 

Would it be possible to overclock 7700k to 5.5+ Ghz (or any CPU to "extreme" frequencies) after replacing stock TIM with liquid metal and cooling it with "setup" like this?

 

People are already benching them at 5.4 ghz on all cores in things like Cinebench R15, validation at 5.5-5.6 should be easy with 1 or 2 cores.  I'm sure there's a few chips out there that can run Superpi 32m at 5.5.

[Master Disaster]

Who'd have though this 5 years ago...

 

AMD making cool hardware

Intel making thermonuclear hardware

[Dabombinable]

59 minutes ago, Lays said:

I expect better from you man

I have to sell a lot of my old stuff before I can actually get at least 1x 4GB stick of DDR3 1333 (which I can easily overclock to 1400MHz, aka a 5GB/sec speed boost). Then I have to get a new cooler since this Pentium 4 stock cooler has poor PWM control and my Siedon 120V died, and then I have to worry about the thermal compound-and I don't want to risk shorting any thing else out (recently discovered that I shorted+fried the onboard audio).

At any rate, even a 10^oc drop would be enough to get my 4790K running at 4.8GHz 24/7-if the new cooler is equal to the 120V (which it won't be-I'm getting a high end air cooler this time around)

[Lays]

1 minute ago, Dabombinable said:

I have to sell a lot of my old stuff before I can actually get at least 1x 4GB stick of DDR3 1333 (which I can easily overclock to 1400MHz, aka a 5GB/sec speed boost). Then I have to get a new cooler since this Pentium 4 stock cooler has poor PWM control and my Siedon 120V died, and then I have to worry about the thermal compound-and I don't want to risk shorting any thing else out (recently discovered that I shorted+fried the onboard audio).

At any rate, even a 10^oc drop would be enough to get my 4790K running at 4.8GHz 24/7-if the new cooler is equal to the 120V (which it won't be-I'm getting a high end air cooler this time around)

If you take clear nail polish and put it over the FIVR on your chip, there is no risk with using the liquid metal TIM's, and the difference is huge with liquid metal on the die compared to normal pastes. 

[Dabombinable]

Just now, Lays said:

If you take clear nail polish and put it over the FIVR on your chip, there is no risk with using the liquid metal TIM's, and the difference is huge with liquid metal on the die compared to normal pastes. 

I wasn't intending to get a normal paste though.

[Bouzoo]

11 minutes ago, Lays said:

If you take clear nail polish and put it over the FIVR on your chip, there is no risk with using the liquid metal TIM's, and the difference is huge with liquid metal on the die compared to normal pastes. 

Would it work with putting clear nail polish around the CPU and GPU die on a laptop MBO, and then adding CLU?

Mind you I'm getting 10-15 C lower temps when I just used NT-H1 instead of the generic crap Asus used (even with 130 MHz OC on my GPU), would it be possible to go even/much lower with CLU?

 

EDIT: the heatsinks are copper. 

Edited February 3, 2017 by Bouzoo

[Lays]

23 minutes ago, Bouzoo said:

Would it work with putting clear nail polish around the CPU and GPU die on a laptop MBO, and then adding CLU?

Mind you I'm getting 10-15 C lower temps when I just used NT-H1 instead of the generic crap Asus used (even with 130 MHz OC on my GPU), would it be possible to go even/much lower with CLU?

 

EDIT: the heatsinks are copper. 

If you're paranoid about components near the die/gpu die getting CLU on it sure I guess.

 

Really though if you just apply a thin layer on the die, it won't spill over.

[Bouzoo]

2 minutes ago, Lays said:

If you're paranoid about components near the die/gpu die getting CLU on it sure I guess.

 

Really though if you just apply a thin layer on the die, it won't spill over.

Eh, I always prepare for the worst, even though that's what usually happens anyway. Just when I decided to use Noctua, and applied it for that matter, now I'm itching to try CLU. Dang it.

[VagabondWraith]

I would love to delid my CPU but I'm scared to death, which is weird because I have no problems taking apart a $1200 dollar GPU to liquid cool lol

[FPSwithaWacomTablet]

5 hours ago, Ryan_Vickers said:

( ಥـْـِـِـِـْಥ) what!?  Please for the love of god... *facepalm* someone needs to explain to them that's not how temperatures work.  For example, 40 C is not "twice as hot" as 20 C since 0 in whatever scale you're using, unless it's absolute is just some arbitrary value with no meaning in this context.

Q = mcT

 

Where Q is the energy of the object in Joules, m being the mass of the object, c being the heat capacity for a given unit mass of said object and T being the temperature in Kelvin.

 

It's a quick tour of high school physics guys. Though if you really wanna dig down deep into temps you might as well start learning thermodynamics.

 

TL;DR It just means it's easier to make light things hotter than heavy things, all else being equal.

 

Should be dQ = mc(dt) but can't be bothered to find the delta symbols and calculus

[goodtofufriday]

5 hours ago, TheRandomness said:

You're getting it all wrong, the Powercolor R9 295x2 air cooled edition with an FX9590 is what you need. AMD does solder all their IHS' to their chips, so why doesn't Intel do that?

Nuhnuhno. Try my old build. 9590 +2x 6990s. 

I hadnt paid for heating for a few winters. 

[DarkBlade2117]

6 hours ago, keNNySOC said:

If your temps are already cool with the 6700k i see no reason to do it...it's not that massive of a improvement as compared to the 7700k + you could ruin your chip, and you might not be able to just send a email to Intel saying "Hey can i get one for free" like linus can

 

 

One of the cores on my 6700k is 8C hotter than the others

[Dabombinable]

Just now, goodtofufriday said:

Nuhnuhno. Try my old build. 9590 +2x 6990s. 

I hadnt paid for heating for a few winters. 

It was the same for me running a Xeon X5450@ 4GHz+ a very poorly binned GTX 970 (it requires the power limit to be at 112% to run stable at its stock clocks). I couldn't have the leave the room closed at all....

Just now, DarkBlade2117 said:

One of the cores on my 6700k is 8C hotter than the others

Its still better than the 10^oC difference between die 1 and 2 of my X6850.

[michaelocarroll007]

6 hours ago, Syntaxvgm said:

wow. Kaby lake is...useless. 

 

5 hours ago, MageTank said:

Why in the world is this "interesting"? This has been common for a very long time now, and is nothing new. They are not telling us anything we didn't already know, and plenty of us have been preaching these results on this very forum for a long time now. Even after Linus released that terrible delidding video of his, we were still saying this. I even had a thread about it: 

 

My results? The exact same as PCper's. 22C difference on my 6700k.

 

Yes, the glue they use is the biggest culprit for the drastic change in heat. It's also why you see large variances in per-core temps (one core being 8C+ hotter than another under uniform load) because the glue's thickness isn't perfectly uniform around the perimeter of the IHS. 

 

 

Yeah But This article even quotes the below 

 

Quote

 

The Bottom Line

That was a whole lot of effort for not much result. But it was fun as hell. I am testing now to see where we get backing the memory clock down.

 

Feel like you all are taking that graph a little out of context it was on a 4.9ghz at 1.35V 

[Sauron]

Not surprising, after Sandy Bridge Intel's thermal interface design went down the drain. I still don't understand why they stopped soldering the die to the lid, I can't think of a good reason to do it any other way. I'm glad I got on that Sandy hype train while it was hot. Even the non-k chips like mine are excellent and can overclock with the multiplier to reasonable levels and with the h100 I got used it sits below 75° even at full load.

[Bouzoo]

11 minutes ago, michaelocarroll007 said:

Feel like you all are taking that graph a little out of context it was on a 4.9ghz at 1.35V 

 

We have people (a person) here running it @ 5.2 Ghz 1.424 with 3600 MHz RAM with all cores enabled so the graph is not really out of context, it is fairly reasonable. And you can buy a guaranteed OCed 5.1 GHz 7700K, either alone or in a system. Considering they "dropped" ~25%, I wouldn't call that not much result. I'm not calling anyone out, but it seems to me like like they were scarred to pump up the voltage a bit more.

[suits]

Not surprising. Kaby lake seems to run really hot. I've got the voltage at 1.25 and it runs hotter than my 6800k at 1.4v with the same cooler. 

[Devin92]

7 hours ago, Ryan_Vickers said:

( ಥـْـِـِـِـْಥ) what!?  Please for the love of god... *facepalm* someone needs to explain to them that's not how temperatures work.  For example, 40 C is not "twice as hot" as 20 C since 0 in whatever scale you're using, unless it's absolute is just some arbitrary value with no meaning in this context.

I dont really understand this. However my knowledge is still from high school so correct/update me. From my understanding of specific heat capacity, temperature change and the amount of energy needed to absorb (dissipate) has a linear ratio, given the specific heat does not change with temperature. so a 20C chip do has double the amount of energy compares to a 40C chip.

[zMeul]

1 hour ago, Sauron said:

Not surprising, after Sandy Bridge Intel's thermal interface design went down the drain. I still don't understand why they stopped soldering the die to the lid, I can't think of a good reason to do it any other way. I'm glad I got on that Sandy hype train while it was hot. Even the non-k chips like mine are excellent and can overclock with the multiplier to reasonable levels and with the h100 I got used it sits below 75° even at full load.

because it exfoliates the die as the solder solidifies, it will shrink and tear the edges inwards

[Funtoink63]

3 hours ago, Lays said:

 

People are already benching them at 5.4 ghz on all cores in things like Cinebench R15, validation at 5.5-5.6 should be easy with 1 or 2 cores.  I'm sure there's a few chips out there that can run Superpi 32m at 5.5.

I meant achieving "extreme" clocks on air using one of those server fans to cool CPU. Like in the video.

29 minutes ago, Sauron said:

Not surprising, after Sandy Bridge Intel's thermal interface design went down the drain. I still don't understand why they stopped soldering the die to the lid, I can't think of a good reason to do it any other way.

Someone correct me if I am wrong.

 

LGA 2011-3 CPUs have soldered IHS, smaller CPUs don't. it has to do something about physics. I do believe that the solder shrinks/expands during the temperate changes. Smaller dies are sensitive to this and they would crack whereas bigger dies won't.

Because the solder "holds" CPUs die, as the temperature rise, solder expands, die does too. Similarly with shrinking. Eventually this will cause die to crack. You  probably could have CPU with smaller die and soldered IHS but you would need really good cooler to avoid temperature changes. CPUs with bigger dies have... well... bigger die surface, it is capable to withstand this stress. Because thermal paste doesn't create this issue. You can have it on smaller die and not cause to crack it during temperature changes. It either:

1. doesn't expand/shrink

2. simply doesn't hold die tightly as solder so even if it expand/shrink, it doesn't affect the die

 

Atleast this is how I understand it.

[patrickjp93]

9 hours ago, Ryan_Vickers said:

( ಥـْـِـِـِـْಥ) what!?  Please for the love of god... *facepalm* someone needs to explain to them that's not how temperatures work.  For example, 40 C is not "twice as hot" as 20 C since 0 in whatever scale you're using, unless it's absolute is just some arbitrary value with no meaning in this context.

Reduction in Temperature =/= Reduction in Heat/Thermal Energy. The statement is correct.

[stealth80]

7 hours ago, Technicolors said:

so you sold the 3770k on ebay? 

Yep, but I clearly stated the single channel only in title and desciption

[Lays]

11 hours ago, keNNySOC said:

If your temps are already cool with the 6700k i see no reason to do it...it's not that massive of a improvement as compared to the 7700k + you could ruin your chip, and you might not be able to just send a email to Intel saying "Hey can i get one for free" like linus can

 

 

The difference is the same on literally every architecture.  Haswell, ivy, skylake, kaby, they all benefit that ~20c drop with CLU.  

[Lord Nicoll]

10 hours ago, Lays said:

Nobody uses those non IHS things for LN2, everyone always puts the IHS back on, too much risk for shorting + crushing die without the IHS.

No they don't I know a few who uses bare die. The risk of shorting is very small as long as you keep the pot below ambient all the time, and insulate correctly, bare die gives much better temperatures.