Kaby Lake crap TIM , WHY ?

gregorvich

Limp Gawd
Joined
Mar 6, 2004
Messages
130
So I am reading a lot of articles about a de-lidded Kaby lake and a 25 degree improvement in temps with a better TIM. Do we believe that Intel is deliberately using by the sounds of it a sub par TIM to shave a cent of the finished product cost which is near the top end of the price spectrum ?

OR as my inner skeptic thinks that it is to cripple the processor to some degree in order to leave open a path for the next release of Kabylake architecture cpus (with better TIM) to be able to clock faster in standard form and be in theory from a marketing view an upgrade/top of the line processor.

Back in the day 386 and 486 cpus came in SX and DX form. The DX indicated a maths co processor on board the die. The SX came with it as well but it was disabled in order to be able to charge more for the DX.

Cant see them shrinking the die much in the next couple of years due to optical limits

hard to find a big market for 6+ cores on the domestic front. 95% of applications in general use are happy on 1/2

thermal limits restrict standard machines to 4-4.5 GHz without exotic cooling

How are Intel going to entice in the future if speeds become static and nobody really needs more than 4 cores ?

Thoughts ?
 

viscountalpha

2[H]4U
Joined
Oct 16, 2011
Messages
2,546
it's more of an "intel doesn't really care" point. They want it to work and work well. Do you really think they care about the consumer market? No. They don't. The consumer market is such a small part of their overall profit margin that they don't care about TIM or delidding at all.
 

Shintai

Supreme [H]ardness
Joined
Jul 1, 2016
Messages
5,691
Could you link the article you mention?

The reason why TIM is used is because solder isn't some kind of universal savior. Specially not on smaller dies. Cracked solder removes all the benefits and some kind can be so bad that the chip ends up faulty. I would be surprised if Ryzen isn't TIM based as well, unless the die is much bigger than expected.
 
Joined
Nov 28, 2016
Messages
31
RichUK's testing came from here: https://forums.anandtech.com/thread...ies-overclocks-inside-now-with-delid.2493250/

It's not clear to me that he "re-lidded" the CPU after changing out the paste, which could be why there's such a drastic difference.

Here you have an example from germany - that is a retail CPU bought in a shop.

http://www.hardwareluxx.de/communit...gewechseltem-tim-891243-338.html#post25150844


The guy who did the delidding. among dozens of other CPUs also delidded my 3570K. He uses LM between DIE and IHS for re-lidding.
 

pxc

Extremely [H]
Joined
Oct 22, 2000
Messages
33,064
Here you have an example from germany - that is a retail CPU bought in a shop.

http://www.hardwareluxx.de/communit...gewechseltem-tim-891243-338.html#post25150844


The guy who did the delidding. among dozens of other CPUs also delidded my 3570K. He uses LM between DIE and IHS for re-lidding.
Yeah, but this one only differs by 13C. That TIM pad on the IHS looks like crap in any case, but it would probably look different after it had set in at higher temperatures and under some additional pressure from a mounted HSF.
 

RedWagnum

Gawd
Joined
Mar 30, 2007
Messages
589
Ok while the consumer market is small for intel surely the server farms etc would see lower temps as very important due to energy costs in volts and cooling. A 1 degree lower temp translate in to a big saving when you scale up.
Doesn't work that way. The only way to reduce energy costs is to reduce power consumption. A processor that draws 100W will still draw 100W no matter if the TIM is good or bad. It is still going to use the same amount of power and generate the same amount of heat. The only difference the TIM makes is to the efficiency of the heat transfer to another medium. Think of TIM like the clutch in a transmission. The less it slips the more power is transferred to the wheels. If the engine is generating a fixed 100 HP and you have a perfect clutch you'll get 100 HP at the wheels but if the clutch is slipping, the engine will still be generating 100 HP but you may be getting only 70 at the wheels. (Yes, yes, yes there are other losses in a transmission like gearing but I'm talking about a theoretical ideal transmission). The engine will not run any hotter or cooler because it is still generating 100 HP.

So the only way to reduce energy costs is to increase the efficiency of the processor (more work for the same amount of energy consumed) requiring fewer processors or reducing the load (less power required for less work.) The TIM has nothing to do with power usage.
 

CSI_PC

2[H]4U
Joined
Apr 3, 2016
Messages
2,200
It could also be partly down to the firmware as well, with the TIM not optimal for the original settings before a more mature and critically stable tolerance settings pertaining to firmware.

The following chart shows the ProCarbon test using a more recent firmware relating to the motherboard/CPU, and has lower temps:


The new article on the testing and thoughts about the firmware and how they have adjusted it can be seen here (pretty clear when looking at power draw that can be seen in the link):
http://www.tomshardware.com/news/intel-core-i7-7700k-kaby-lake-overclocking-update,33119.html

Worth reading.
It does seem yet again Intel has tried to get away with the absolute minimum required for heat dissipation, which may not bode well for really good OCing.
Cheers
 
Joined
Nov 28, 2016
Messages
31
Yeah, but this one only differs by 13C. That TIM pad on the IHS looks like crap in any case, but it would probably look different after it had set in at higher temperatures and under some additional pressure from a mounted HSF.
sigh - look at the max temps ... ~25°C delta
 
Joined
Nov 28, 2016
Messages
31
It does seem yet again Intel has tried to get away with the absolute minimum required for heat dissipation, which may not bode well for really good OCing.
Cheers
When u use the link I provided earlier and use google translate you can learn the following - rough translation (my own):

"Most CPUs reach 5 GHz at either 1.29V or something like 1.35V ... 30 samples"
 

CSI_PC

2[H]4U
Joined
Apr 3, 2016
Messages
2,200
When u use the link I provided earlier and use google translate you can learn the following - rough translation (my own):

"Most CPUs reach 5 GHz at either 1.29V or something like 1.35V ... 30 samples"
Seems they are quoting something that is not confirmed nor how tested - although I am using a translator.

My point is around that the more recent firmware adjusts the power draw-performance envelope and so TDP, hence why the temps were really bad in the 1st example Tom's Hardware had then improved dramatically with their 2nd sample that clearly shows that change, but also indicates that the TIM solution-spec Intel went with is not really enthusiast level nor worthy of the CPU's price point.
I prefer to use Tom's because they have a much better power measurement system and tools, anyway look at what Skylake can OC to and with what draw/TDP.

Cheers
 
Last edited:

gregorvich

Limp Gawd
Joined
Mar 6, 2004
Messages
130
does resistance not square with the power ? so if the chip is cooler then less resistance and less volts needed to do the same work ?
 

EchoWars

Limp Gawd
Joined
Jan 7, 2010
Messages
497
Dissipated power remains the same regardless. And in a modern CPU, you'll be hard-pressed to find a 'resistor'.

WRT resistance, power squares with current (P=I² x R). Not really relevant when talking about processor pwr consumption.
 

Shintai

Supreme [H]ardness
Joined
Jul 1, 2016
Messages
5,691

Shintai

Supreme [H]ardness
Joined
Jul 1, 2016
Messages
5,691
Good point , the plot thickens ! But would no lid give that sort of change ?
Yep, removing the IHS gives a large benefit. But it comes with a lot of dangers of it own. People got large benefits from delidding soldered parts too.
 

gregorvich

Limp Gawd
Joined
Mar 6, 2004
Messages
130
Thanks for all the informative comments. But back to my original question. Why would you cripple your top of the line consumer CPU which being unlocked and aimed at enthusiasts like us with Tim that's not up to the job. I'm convinced that 6 months on they will release a 5k kabylake with a good time in order to have something to gazump AMD new offering.
 

Monkey God

Mangina Full of Sand
Joined
May 7, 2007
Messages
6,723
Thanks for all the informative comments. But back to my original question. Why would you cripple your top of the line consumer CPU which being unlocked and aimed at enthusiasts like us with Tim that's not up to the job. I'm convinced that 6 months on they will release a 5k kabylake with a good time in order to have something to gazump AMD new offering.
Your question has been answered already.

it's more of an "intel doesn't really care" point. They want it to work and work well. Do you really think they care about the consumer market? No. They don't. The consumer market is such a small part of their overall profit margin that they don't care about TIM or delidding at all.
Intel hasn't "crippled" anything. It just doesn't matter for the vast majority of CPU owners.
 

Shintai

Supreme [H]ardness
Joined
Jul 1, 2016
Messages
5,691
Thanks for all the informative comments. But back to my original question. Why would you cripple your top of the line consumer CPU which being unlocked and aimed at enthusiasts like us with Tim that's not up to the job. I'm convinced that 6 months on they will release a 5k kabylake with a good time in order to have something to gazump AMD new offering.
Your question should be, why isn't Intel and AMD selling delidded CPUs to enthusiasts. If you think AMD uses solder in its newer CPUs you will be disappointed.
 

gregorvich

Limp Gawd
Joined
Mar 6, 2004
Messages
130
Monkey God I cant see how my question is answered. My question was why would Intel cripple a chip with a mediocre TIM on their High end enthusiast unlocked CPU. I think it for marketing and this is why: I worked as a Technical Recruitment consultant in the 80s in Ireland and then London. I was given a guided tour of the Mostek chip production facility in Dublin. My guide explained that at the beginning of a production run the cpus coming off the line would mostly test slow but the odd one would clock a lot faster. They were sorted in to bins according to speed. My guide explained that as they gained experience of the production foibles of a new chip they would dial in the process and start to increase the yield of faster CPU. He explained they sold the slower ones first to get the best value for them and reduce stock, and then months on would release the faster variant and charge a premium.

Now today I imagine the dialing in of a new production run of a CPU is much faster and one could expect a much higher initial yield of quick CPU Day one. So if the vast majority of 7700K produced are quick where are you going to get the faster variant from for the next release. I say you cripple this release with poor TIM and then later on release the same CPU with a faster clock using good Tim so it stays cool and keep the price premium for speed. I thought the whole point of unlocked "K" processors was for the like of us . If they give us the best offering of a new line day one there is nothing to tempt us with in six months and consumers expect prices to drop over time.
 

Shintai

Supreme [H]ardness
Joined
Jul 1, 2016
Messages
5,691
Why isn't the soldered parts performing so much better if Intel and AMD is "crippling" the TIM parts? :)

AMD is a good example because they use both types. And I dont think you can tell what chip uses what.

 

gregorvich

Limp Gawd
Joined
Mar 6, 2004
Messages
130
Why isn't the soldered parts performing so much better if Intel and AMD is "crippling" the TIM parts? :)

AMD is a good example because they use both types. And I dont think you can tell what chip uses what.

I didn't include amd, they have had a hard enough time keeping up. Also I only referring to kabylake CPU. Perhaps we need more Tim replace ment data.
 

thesmokingman

Supreme [H]ardness
Joined
Nov 22, 2008
Messages
5,897
Monkey God I cant see how my question is answered. My question was why would Intel cripple a chip with a mediocre TIM on their High end enthusiast unlocked CPU. I think it for marketing and this is why: I worked as a Technical Recruitment consultant in the 80s in Ireland and then London. I was given a guided tour of the Mostek chip production facility in Dublin. My guide explained that at the beginning of a production run the cpus coming off the line would mostly test slow but the odd one would clock a lot faster. They were sorted in to bins according to speed. My guide explained that as they gained experience of the production foibles of a new chip they would dial in the process and start to increase the yield of faster CPU. He explained they sold the slower ones first to get the best value for them and reduce stock, and then months on would release the faster variant and charge a premium.

Now today I imagine the dialing in of a new production run of a CPU is much faster and one could expect a much higher initial yield of quick CPU Day one. So if the vast majority of 7700K produced are quick where are you going to get the faster variant from for the next release. I say you cripple this release with poor TIM and then later on release the same CPU with a faster clock using good Tim so it stays cool and keep the price premium for speed. I thought the whole point of unlocked "K" processors was for the like of us . If they give us the best offering of a new line day one there is nothing to tempt us with in six months and consumers expect prices to drop over time.
Wrong, Krabby Lake and the rest of the lineup is not high end, it is midrange or mainstream. X99/2011 vX socket is high end, not 1151/1150/1155, etc etc.
 

Gideon

2[H]4U
Joined
Apr 13, 2006
Messages
2,669
Yep, removing the IHS gives a large benefit. But it comes with a lot of dangers of it own. People got large benefits from delidding soldered parts too.
Umm you cant delid a processor with a solder heat spreader as it will destroy the chip trying to remove it. I have never heard of anyone doing that and having a working chip afterwords. So I would like to see a example of this great result after removing a solder on heat spreader. Starting to wonder how much Intel pr pays you.
 

EuphoricRage470

[H]ard|Gawd
Joined
Apr 15, 2011
Messages
1,108
Umm you cant delid a processor with a solder heat spreader as it will destroy the chip trying to remove it. I have never heard of anyone doing that and having a working chip afterwords. So I would like to see a example of this great result after removing a solder on heat spreader. Starting to wonder how much Intel pr pays you.
https://www.overclock3d.net/news/cp...elids_intel_s_core_i7-6950x_broadwell-e_cpu/1

Before you accuse others of being shills, you should triple-quadruple check that you are correct and that the person you're accusing is wrong.
 

Gideon

2[H]4U
Joined
Apr 13, 2006
Messages
2,669
https://www.overclock3d.net/news/cp...elids_intel_s_core_i7-6950x_broadwell-e_cpu/1

Before you accuse others of being shills, you should triple-quadruple check that you are correct and that the person you're accusing is wrong.
Ok May of this year is the first successful delidding of a solder IHS and it gained a whole 4 degrees which is almost nothing not a large difference. That is a ton of risk for little gain and almost nobody would do that. But I am wrong on someone doing it without success. Large gains was entirely fud tho. Glad you could answer it for him tho.
 

Shintai

Supreme [H]ardness
Joined
Jul 1, 2016
Messages
5,691
Umm you cant delid a processor with a solder heat spreader as it will destroy the chip trying to remove it. I have never heard of anyone doing that and having a working chip afterwords. So I would like to see a example of this great result after removing a solder on heat spreader. Starting to wonder how much Intel pr pays you.
Focus on the ball, not the person.

Ok May of this year is the first successful delidding of a solder IHS and it gained a whole 4 degrees which is almost nothing not a large difference. That is a ton of risk for little gain and almost nobody would do that. But I am wrong on someone doing it without success. Large gains was entirely fud tho. Glad you could answer it for him tho.
You can try google delid and sandy bridge or FX, Thuban and older. Its nothing new and the benefit varies. The benefit for the 6950X in question is small and the gap could be below average as an explanation.
 
Last edited:

Gideon

2[H]4U
Joined
Apr 13, 2006
Messages
2,669
Focus on the ball, not the person.



You can try google delid and sandy bridge or FX, Thuban and older. Its nothing new and the benefit varies. The benefit for the 6950X in question is small and the gap could be below average as an explanation.
I looked at Google and Bing and found almost nothing on it other then pictures of ruined processors and people warning others not to do it and there would almost no gain from doing it. People do it on the paste cpu's cause they always use cheap garbage tim and using a much better tim works wonders, a few brave souls even go bare. They use tim cause easy and cheap and it's more cash in their pocket. I wouldn't say Intel or AMD does it to cripple a processor, it just makes them more cash and it works good enough.
 

noko

Supreme [H]ardness
Joined
Apr 14, 2010
Messages
5,148
To me it really sounds more technical then just a cheap Tim. Different materials expand differently for a given temperature change. Different expansion coefficients. A chip has many different types of materials and if you have an interface material that is rather less giving (temperature transition point to the HSF) it may just cause a lot of cycle stress. Meaning after a period of time you form cracks and stress to the material plus stress to the chip itself as it expands and contracts due to temperature changes. CPU have a lot of temperature changes due to work loads. Delidding may help in heat transfer but it may not help in the longevity of the chip if you succeed doing it. So for Intel, failed chips over time, especially on big servers etc. It sounds like for Intel using solder would not work in the long run.

For Enthusiasts, the possible shorter life span of the cpu or damage resulting in having to buy another one is not a concern - go for it.
 

thesmokingman

Supreme [H]ardness
Joined
Nov 22, 2008
Messages
5,897
It sounds like for Intel using solder would not work in the long run.

For Enthusiasts, the possible shorter life span of the cpu or damage resulting in having to buy another one is not a concern - go for it.
Huh? Solder is the traditional method and it is the method for high end chips. Solder doesn't transfer temps terribly like the crap paste and it lasts forever. Where you got "not work in the long run" does not compute. The fact is they sell a helluva lot more 350 and under chips and using paste instead of the more difficult and expensive solder probably saves them a lot of time and money. I'm sure they have been counters that have calculated the financials to make the paste worthwhile for them.
 

Nenu

[H]ardened
Joined
Apr 28, 2007
Messages
19,200
I have another take on this.

A chips lid is attached to the substrate with a solidified gel.
The lid is held off the CPU leaving a gap that the TIM fills.

When you delid, the pressure on the lid is moved from the substrate directly to the CPU core in the centre.
This can bend the CPU substrate such that the core warps (my delidded 6700K does this, I got a pm from somebody else with the same problem on his 6700K).
The thinner substrate and weaker sockets have made this easier to occur.
This is a reason Intel use the lid, as stress relief for the CPU core.
It could be why they dont like soldering, there isnt enough stress relief from the core with it being solid.

The delidded CPUs lid now sits directly on the core and gets the full pressure from the heatsink, it has practically no space for the TIM.
This will make it run a stack cooler.
As long as the Silicon doesnt bend which can cause some cores to be hotter than others due to TIM moving and leaving a gap.
I am using a socket 1155 cooler that can apply too much pressure to socket 1151. I found this out the hard way.
I found all cores remain very cool at first but after a while, 2 or 4 (of 8) cores get up to 30C hotter than the others.
The number of hot cores and temp rise varies with how much pressure I put on the heatsink and the exact position of the lid.
I now fear that the socket has bent a little because I always get 2 (of 8) cores that rise a lot higher than the others after a few days when it has been used at temperature.
This is with a tiny amount of pressure from the heatsink.
I can shift which 2 cores by changing the position of the lid on the core.

Thats it.
In order to provide stress relief there has to be a gap.
The TIM fills that gap.
Remove that gap, cooling will improve substantially. But the chip might bend.

I'm going to bed soon so may reply to any lambasting in the morning :p
 

SomeGuy133

2[H]4U
Joined
Apr 12, 2015
Messages
3,447
it's more of an "intel doesn't really care" point. They want it to work and work well. Do you really think they care about the consumer market? No. They don't. The consumer market is such a small part of their overall profit margin that they don't care about TIM or delidding at all.
they rather save 1 dollar and force me to spend 50 to pay someone to put a decent TIM in that still isn't as good as solder :/....fuckers
 

thesmokingman

Supreme [H]ardness
Joined
Nov 22, 2008
Messages
5,897
I have another take on this.

A chips lid is attached to the substrate with a solidified gel.
The lid is held off the CPU leaving a gap that the TIM fills.

When you delid, the pressure on the lid is moved from the substrate directly to the CPU core in the centre.
This can bend the CPU substrate such that the core warps (my delidded 6700K does this, I got a pm from somebody else with the same problem on his 6700K).
The thinner substrate and weaker sockets have made this easier to occur.
This is a reason Intel use the lid, as stress relief for the CPU core.
It could be why they dont like soldering, there isnt enough stress relief from the core with it being solid.

The delidded CPUs lid now sits directly on the core and gets the full pressure from the heatsink, it has practically no space for the TIM.
This will make it run a stack cooler.
As long as the Silicon doesnt bend which can cause some cores to be hotter than others due to TIM moving and leaving a gap.
I am using a socket 1155 cooler that can apply too much pressure to socket 1151. I found this out the hard way.
I found all cores remain very cool at first but after a while, 2 or 4 (of 8) cores get up to 30C hotter than the others.
The number of hot cores and temp rise varies with how much pressure I put on the heatsink and the exact position of the lid.
I now fear that the socket has bent a little because I always get 2 (of 8) cores that rise a lot higher than the others after a few days when it has been used at temperature.
This is with a tiny amount of pressure from the heatsink.
I can shift which 2 cores by changing the position of the lid on the core.

Thats it.
In order to provide stress relief there has to be a gap.
The TIM fills that gap.
Remove that gap, cooling will improve substantially. But the chip might bend.

I'm going to bed soon so may reply to any lambasting in the morning :p
Ugh, you never want a gap between your IHS, TIM and heat element. The TIM is not the cooling apparatus, it is merely there to fill in the micro gaps between the two surfaces. I repeat again, you never want space between the TIM and heat element. If as you theorize that the IHS compresses onto the TIM each time a mount force is applied, THERE WILL BE NO TIM LEFT the next time a heatsink is applied. Each time a force is applied the TIM will be spread out, and each successive time there will be no TIM left to push down on because its already been spread out. This theory is counter to how TIM is applied and used. There'd have to be a magical force that pushes the TIM back each time for your theory to work.
 
Last edited:

Shintai

Supreme [H]ardness
Joined
Jul 1, 2016
Messages
5,691
I looked at Google and Bing and found almost nothing on it other then pictures of ruined processors and people warning others not to do it and there would almost no gain from doing it. People do it on the paste cpu's cause they always use cheap garbage tim and using a much better tim works wonders, a few brave souls even go bare. They use tim cause easy and cheap and it's more cash in their pocket. I wouldn't say Intel or AMD does it to cripple a processor, it just makes them more cash and it works good enough.
Delidding have been done successfully with soldered parts for over 10 years. Then you can make up whatever excuse you want. That failed delids are more attractive to the internet shouldn't surprise you.

If you notice, TIM tends to be used on smaller dies for a good reason. And its not just about money.
 
Last edited:

Nenu

[H]ardened
Joined
Apr 28, 2007
Messages
19,200
Ugh, you never want a gap between your IHS, TIM and heat element. The TIM is not the cooling apparatus, it is merely there to fill in the micro gaps between the two surfaces. I repeat again, you never want space between the TIM and heat element. If as you theorize that the IHS compresses onto the TIM each time a mount force is applied, THERE WILL BE NO TIM LEFT the next time a heatsink is applied. Each time a force is applied the TIM will be spread out, and each successive time there will be no TIM left to push down on because its already been spread out. This theory is counter to how TIM is applied and used. There'd have to be a magical force that pushes the TIM back each time for your theory to work.
Obvious is obvious, dont get your panties in a wad :p
You didnt consider that my core flexes a little when hot, no magic.

The point is that Intel need to leave a gap with tolerance for problems, I gave a major reason.
It explains why they dont solder.

I detailed my experience so that others might not screw up their own kit after delidding.
When I first apply paste my load temps are low 50C at 4.7GHz on all cores.
After a few days of the core flexing under load, this changes and I have to reduce the overclock to 4.6GHz.

Right now, 2 of my 8 cores are 25C higher than the rest under Prime95.
That is the best I have managed to get it so far with only a very mild tightening of the heatsink.
Shrug.
Its still better than before the delid, my VCore is slightly lower with the same clock speed.
 
Last edited:

Gideon

2[H]4U
Joined
Apr 13, 2006
Messages
2,669
So far everything you linked to is for extreme overclockers that will do anything for a few degrees as they try to set world records. Stress cracking of indium is only a issue when using LN2 due to the extreme temperatures, under normal operation a processor is never exposed to that extreme of a temperature shift. I don't believe this is a mainstream thing being done in the overclocking enthusiast community, the hard core community might have a few people crazy enough. You can try to spin it all you want, but they use paste to save cash and it works good enough, it simplifies manufacturing and those savings add up on mass produced items.
 
Top