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

[goodtofufriday]

1 minute ago, Ryan_Vickers said:

Not at all.  I'm debating their use of it.  They based their percent changes on 0, when 0 in that scale (or any scale really) has no inherent meaning in this context.  The only "0" that I could be convinced makes any sense would be the idle temperature.

But why 0?  We have to think of the physical meaning of this.  0 C is 32 F.  Who's 0 is "real 0"?  Maybe 0 K?  That's generally accepted to be a good point but as has been mentioned, there are reasons you might not want to simply accept it as gospel either.  I'm saying they should have picked a temperature with meaning in the context of the test, like, for example, idle temps, rather than what they did.  In this context, 0 is no more special than any other number.

Okay. I think ibunderstand better. But regardless of the scale used the heat output reduction is still scientifically the same. 

I think most people can easily grasp tge concept when relative to 0, asna lot of other scales work that way. 

Maybe its not the best scale but people understand it.  

[vanished]

1 minute ago, goodtofufriday said:

Okay. I think ibunderstand better. But regardless of the scale used the heat output reduction is still scientifically the same. 

I think most people can easily grasp tge concept when relative to 0, asna lot of other scales work that way. 

Maybe its not the best scale but people understand it.  

Ok, there's a few things to mention here:

• the heat output (ie, wattage) was (in theory) not actually reduced by delidding at all (perhaps slightly due to electronic efficiency at lower temperatures but that's within the noise imo)
• The temperature was reduced, but, as I demonstrated above, the % change in temperature is different depending on what scale you use, (C, F, etc.) if you make your % difference measurements relative to 0 in that scale.

[Drak3]

15 minutes ago, goodtofufriday said:

Being genuine here. I dont really understand? 

If thw cpu was X temp before TIM

And then is Y temp after TIM,  why is it not okay to calculate the percentage differwnt between the two relative to zero? I think that a range of "0 - > max load temp"  makes perfect sense. 

 

But one cannot reach 0c with an air cooler or water loop. The closest they can get is ambient, which is not a constant temperature, especially when one considers that some people keep their systems in certain conditions for various reasons.

 

8 minutes ago, zMeul said:

that's exactly why I say you're debating the Celsius scale and not the temp differences 

@Ryan_Vickers is arguing that using a scale where 0 is an arbitrary figure for an absolute % difference doesn't really work. He then brings up a scale where 0 is an absolute end of an array (temperatures can go up indefinitely in theory), which wouldn't exactly be better, as 100K is -173.15c, meaning we'd be looking at long lists of numbers to get values that look less significant than they are.

 

7 minutes ago, goodtofufriday said:

Okay. I think ibunderstand better. But regardless of the scale used the heat output reduction is still scientifically the same. 

I think most people can easily grasp tge concept when relative to 0, asna lot of other scales work that way. 

Maybe its not the best scale but people understand it.  

But people can be confused if they see thermal compound X reduces temps 25%, but ambient for them is 10% below what their CPU is already experiencing (good cooler) according to thermal compound X's manufacturer's scale.

[MageTank]

4 minutes ago, patrickjp93 said:

So does Kelvin. We've achieved sub-0K in labs many times in the last decade.

 

See, this is why I took a hiatus. This community has gotten so toxic, especially considering the mods still let you crawl around.

The community isn't growing toxic. They just started demanding sources to the "facts" that often gets spouted here. As for why the mods let me "crawl around", it's probably because even when I toy around with people, I do so within the rules. The exception being when I toy with you, mostly because you yourself go outside of the rules when speaking to me. We are both equally childish in that regard, but at least we both have fun while doing it.

 

4 minutes ago, Prysin said:

which Grizzly you referring to? Kryo or Conducto? Kryo is not much better then NH-T1 from Noctua. CLU is worse then Grizzlys Conductonaut. No documentation i can find proves otherwise. Liquid PRO is supposedly the best, but it seems Liquid Ultra is the more popular one... as far as i know, Grizzlys metal TIM is less accessible in the state (going by PCPartpicker)

 

Although, so far i only found two places mentioning thermal values...

http://forum.notebookreview.com/threads/liquid-metal-showdown-thermal-grizzly-conductonaut-vs-cool-laboratory-liquid-ultra-pro.791489/

http://www.overclock.net/t/1351984/coollaboratory-liquid-ultra-vs-liquid-pro-vs-phobya-liquid-metal#post_19106740

 

Also the last part wasn't referring to you, by may have referred to someone else who's username begins with the same letter as yours. Said person may or may not be watching this thread anymore.

Okay, let me clear this up. Pro is better than Ultra, by 1-2C, but it's not used because it's exceedingly difficult to spread and is an absolute pain to clean off. Ultra is hard enough to spread on nickel (it balls up, and hates bonding to it) without sanding the surface down to copper, and Pro is even worse in this regard. 

 

Conductonaut isn't much better at all compared to CLU, and is far more expensive and way harder to find. It's also almost completely liquid, making leaking a problem. I am shocked that your Notebookreview source used it on a laptop. That guy has no fear, even with tape, lol. 

 

The thermal conductivity of paste isn't the be all, end all for how great it will work. Remember, pastes primary function is to fill the gaps in between the surfaces that are conducting/dispersing the heat. CLU having more silicone helps in this regard (not only making it easier to spread, but easier to fill gaps with less material waste). There is also the longevity issue with conductonaut, lasting even less than CLU (which is already far less than Dow Corning 5688). After all, CLP already has 80W/mk, while TGC boasts 73W/mk. Seeing as we only see a 1-2C difference between CLU and CLP, it's safe to assume we won't see any real benefit from using that over CLU.

 

As for the Grizzly I was referring to, that would be Kryonaut, the paste that sub-zero OCers use (since both CLU and TGC fail once below 10C). 

[vanished]

15 minutes ago, zMeul said:

no Celsius, not anything else,  just a X scale from 0 to 100

on that scale is 91 at a 25.(274725)% from 68?

Mathematically, yes, but this isn't pure math.  This is real life where the numbers we apply have to have some physical meaning - they have to represent something, and if they don't do that properly, they're not helping.

 

The problem I have is this: percentages lead directly into interpolation/extrapolation.  If they say this reduces temperature by 25%, then that means if they were able to reduce the temperature by 100%, the CPU would be at 0.  Well, does that make sense?  It still has heat, relative to 0 K, so it's not like the temperature was really reduced entirely, and it could go lower too.  What if we reduce temperature by 150%?  What does that even mean?  If the room was sufficiently far below 0 it could theoretically be done but it just sounds nonsensical.

10 minutes ago, Ryan_Vickers said:

• The temperature was reduced, but, as I demonstrated above, the % change in temperature is different depending on what scale you use, (C, F, etc.) if you make your % difference measurements relative to 0 in that scale.

This in and of itself proves, in my opinion, that it was a poor choice.  Your claims should not be affected by how you chose to measure them.  If I can switch scales and suddenly my % improvement is a different number, something is seriously wrong.

[goodtofufriday]

3 minutes ago, Drak3 said:

But people can be confused if they see thermal compound X reduces temps 25%, but ambient for them is 10% below what their CPU is already experiencing (good cooler) according to thermal compound X's manufacturer's scale.

Very fair point. 

But then it might actually be better to use 0 because of the high variance between different coolers/pastr. 0 being considered an "off" state

[Drak3]

5 minutes ago, goodtofufriday said:

Very fair point. 

But then it might actually be better to use 0 because of the high variance between different coolers/pastr. 0 being considered an "off" state

We'd still have to account for the variances between CPUs (both within a SKU and comparing SKUs), and the % differences could be different for the generalization comparing say, the 7700K, and the 6950X.

[goodtofufriday]

5 minutes ago, Drak3 said:

We'd still have to account for the variances between CPUs (both within a SKU and comparing SKUs), and the % differences could be different for the generalization comparing say, the 7700K, and the 6950X.

I think the max temp before throttling occurs is a good enough max temp. 

Idk,  im not trying to change the system here lol. 

[Drak3]

13 minutes ago, goodtofufriday said:

I think the max temp before throttling occurs is a good enough max temp. 

Idk,  im not trying to change the system here lol. 

% Differences just aren't good unless we're comparing a single CPU to itself, with the start of the scale being an absolute constant (that must be documented and mentioned every time it's discussed), and only one scale should be used (preferably Celsius as we're well established as using that for computing thermals).

[zMeul]

3 hours ago, Drak3 said:

is arguing that using a scale where 0 is an arbitrary figure for an absolute % difference doesn't really work. He then brings up a scale where 0 is an absolute end of an array (temperatures can go up indefinitely in theory), which wouldn't exactly be better, as 100K is -173.15c, meaning we'd be looking at long lists of numbers to get values that look less significant than they are.

fact is Kelvin scale is not absolute

and I shown to you that Absolute Zero, as in 0 deg Kelvin is not the lowest number on the Kelvin scale - negative temperatures on Kelvin scale can be achieved

 

the premise of the Lord Kelvin scale is that nothing is colder than Absolute Zero, and he was wrong

[vanished]

44 minutes ago, zMeul said:

fact is Kelvin scale is not absolute

and I shown to you that Absolute Zero, as in 0 deg Kelvin is not the lowest number on the Kelvin scale - negative temperatures on Kelvin scale can be achieved

 

the premise of the Lord Kelvin scale is that nothing is colder than Absolute Zero, and he was wrong

I always just assumed the purpose of it was that 0 represented a point at which there was 0 thermal energy - no movement of the atoms.  Do you know for a fact that isn't the case?  Just because you can go below 0 K doesn't necessarily mean it isn't; when it comes to quantum mechanics, I've seen stranger stuff than the concept of "negative energy"  

 

Regardless, I've summed up my point and the heart of it wasn't K vs C:

4 hours ago, Ryan_Vickers said:

Quote

• The temperature was reduced, but, as I demonstrated above, the % change in temperature is different depending on what scale you use, (C, F, etc.) if you make your % difference measurements relative to 0 in that scale.

This in and of itself proves, in my opinion, that it was a poor choice.  Your claims should not be affected by how you chose to measure them.  If I can switch scales and suddenly my % improvement is a different number, something is seriously wrong.

 

[PRTI]

16 hours ago, DarkBlade2117 said:

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

If its already cool and silent i see no reason to try and lose $320

[PRTI]

11 hours ago, Lays said:

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

Linus' video on a 6700k got a ~5c difference, while this experiment got a ~23c difference

 

So the difference is the same?

[MageTank]

8 minutes ago, keNNySOC said:

Linus' video on a 6700k got a ~5c difference, while this experiment got a ~23c difference

 

So the difference is the same?

This is why we don't follow Linus with blind devotion. His video was awful, and everyone has already told him that. Even the more... reputable enthusiasts have pointed it out. If you actually go to that video's official thread, you will see all of the reasons why his methodology was flawed, along with many examples of proper delidding and the results that come of it. 

[PRTI]

6 minutes ago, Kumaresh said:

Linus messed up and didn't follow the proper technique. There is already a thread on LTT regarding the delidding video which has many replies detailing what he did wrong......

 

2 minutes ago, MageTank said:

This is why we don't follow Linus with blind devotion. His video was awful, and everyone has already told him that. Even the more... reputable enthusiasts have pointed it out. If you actually go to that video's official thread, you will see all of the reasons why his methodology was flawed, along with many examples of proper delidding and the results that come of it. 

Ok thanks for the info! wasn't up to date  ty for informing me

[MageTank]

3 minutes ago, keNNySOC said:

 

Ok thanks for the info! wasn't up to date  ty for informing me

No problem. In fact, I'll go ahead and link that thread here in case anyone else needs to see it:

 

I will also link my full response to the video:

 

[Lays]

22 minutes ago, keNNySOC said:

Linus' video on a 6700k got a ~5c difference, while this experiment got a ~23c difference

 

So the difference is the same?

Linus's video was improperly done.

[Bleedingyamato]

On February 3, 2017 at 1:14 AM, DeadEyePsycho said:

Mine was hitting 100C in stress tests while everything was stock. I undervolted my chip just to cool it down.

So Kaby lake runs too hot?  (At least in the case of stress tests.) 

 

Someone in this thread was saying the thermal compound that Intel put under the IHS could or would be only good for maybe 1-2 years.  Is that correct and Kaby lake likely has a rather short like expectancy vs Skylake for example?

 

Or am I misunderstanding the severity of the problem? 

 

On February 3, 2017 at 1:18 AM, Syntaxvgm said:

wow. Kaby lake is...useless. 

Same questions as what I asked DeadEyePsycho.

 

I'm curious what your opinion of this heat issue and mention of a 1-2 year reduced lifespan (if that's what was being implied/stated here) of Kaby lake?

 

 

[DarkBlade2117]

13 hours ago, keNNySOC said:

If its already cool and silent i see no reason to try and lose $320

When they're all 75C and it is 83-85C kind of annoys me

[Lays]

Was curious what settings and test they were using in the test, so I found their picture and tried their settings for myself.

 

 

My result, can't get my voltage to go to 1.368, closest it goes is 1.376, or 1.362 I think was the other step.

 

 

 

My result at 5.2:

 

 

7 hours ago, Bleedingyamato said:

So Kaby lake runs too hot?  (At least in the case of stress tests.) 

 

Someone in this thread was saying the thermal compound that Intel put under the IHS could or would be only good for maybe 1-2 years.  Is that correct and Kaby lake likely has a rather short like expectancy vs Skylake for example?

 

Or am I misunderstanding the severity of the problem? 

 

Same questions as what I asked DeadEyePsycho.

 

I'm curious what your opinion of this heat issue and mention of a 1-2 year reduced lifespan (if that's what was being implied/stated here) of Kaby lake?

 

 

Kaby lake doesn't run too hot, anyone hitting 100c on stock settings has improper cooling, or a horrible mount with their cooler. The thermal compound is the same shit they've been using for years, dow corning or w/e.  It's designed to last forever basically.  The stuff people were talking about being good for 1-2 years is CLU, the liquid metal compound people use when they replace intel's TIM.

[Bleedingyamato]

54 minutes ago, Lays said:

Kaby lake doesn't run too hot, anyone hitting 100c on stock settings has improper cooling, or a horrible mount with their cooler. The thermal compound is the same shit they've been using for years, dow corning or w/e.  It's designed to last forever basically.  The stuff people were talking about being good for 1-2 years is CLU, the liquid metal compound people use when they replace intel's TIM.

That's a relief.  I was thinking about getting a 7700K at some point for a new build and was worried that some of what's been mentioned in this thread was suggesting that there was a serious flaw with Kaby lake.

 

 

Why do people replace stuff that maybe doesn't work as well with this liquid metal stuff that doesn't last nearly as long?

 

Isn't there anything better to replace the stock compound with that would work better and last longer than barely a couple years at most?

[vanished]

4 minutes ago, Bleedingyamato said:

That's a relief.  I was thinking about getting a 7700K at some point for a new build and was worried that some of what's been mentioned in this thread was suggesting that there was a serious flaw with Kaby lake.

 

 

Why do people replace stuff that maybe doesn't work as well with this liquid metal stuff that doesn't last nearly as long?

 

Isn't there anything better to replace the stock compound with that would work better and last longer than barely a couple years at most?

I guess they'd trade lasting forever (down to a few years) for improving load temps by 20+ degrees

[patrickjp93]

4 minutes ago, Bleedingyamato said:

That's a relief.  I was thinking about getting a 7700K at some point for a new build and was worried that some of what's been mentioned in this thread was suggesting that there was a serious flaw with Kaby lake.

 

 

Why do people replace stuff that maybe doesn't work as well with this liquid metal stuff that doesn't last nearly as long?

 

Isn't there anything better to replace the stock compound with that would work better and last longer than barely a couple years at most?

There are a number of non-conducting thermal pastes which fit this image, but most people tend to choose the "max temp reduction NAO!" approach.

[patrickjp93]

1 minute ago, Ryan_Vickers said:

I guess they'd trade lasting forever (down to a few years) for improving load temps by 20+ degrees

You can generally get a very similar reduction with other pastes. The problem is the spacing the glue causes.

[Lays]

6 minutes ago, Bleedingyamato said:

That's a relief.  I was thinking about getting a 7700K at some point for a new build and was worried that some of what's been mentioned in this thread was suggesting that there was a serious flaw with Kaby lake.

 

 

Why do people replace stuff that maybe doesn't work as well with this liquid metal stuff that doesn't last nearly as long?

 

Isn't there anything better to replace the stock compound with that would work better and last longer than barely a couple years at most?

Because 20-30c drop is worth the 1-2 year lifespan of the liquid metal that can be replaced with 2 minutes of work