The Truth About CPU Soldering

[EarthDog]

Dont know. Im still waiting to hear an actual problem that needs to be fixed.

 

[cadaveca]

Also, I forgot to point out that Intel used solder for their Clarkdale processors, which has a die size about half that of Coffee Lake or Skylake consumer processors

https://en.wikipedia.org/wiki/List_of_Intel_Core_i3_microprocessors#"Clarkdale"_(MCP,_32_nm)

If Micro-cracking was not an issue with these small dies, they most certainly would not be an issue for any other Intel processor. Micro-cracking is a bigger issue on smaller dies but you can use different metal layering for different processor lines to offset this issue. If they knew the correct layering for small die CPUs, why not just use that on newer processors?

The micro fissures are a result of the added layer on the IHS leaching away the gallium/indium. It's a well-documented problem with these types of solders. The larger the patch of gallium/indium, the more prone to being problematic, not the other way around.

 

[GoldenX]

To me it should be segregated, similar to what AMD is doing. I don't care that a Celeron, Pentium o normal iX has TIM, but an unlocked K should always be soldered.

 

[Solaris17]

especially when overclocked more than makes up for it

Id think not since

cents x CPUs = millions of dollars and

especially when overclocked more than makes up for it

Intel isnt targeting us at all and enthusiast users are literal grains of rice to the thousands of chips OEMs are ordering by the pallet. The chips run at or under TJ max during full load during testing and this checks all the boxes.

Clarification: Not biased, just thinking logical biz sense. Currently own what might be the most expensive consumer level CPU from them.

 

[R0H1T]

Id think not since

cents x CPUs = millions of dollars and

especially when overclocked more than makes up for it

Intel isnt targeting us at all and enthusiast users are literal grains of rice to the thousands of chips OEMs are ordering by the pallet. The chips run at or under TJ max during full load during testing and this checks all the boxes.

Clarification: Not biased, just thinking logical biz sense. Currently own what might be the most expensive consumer level CPU from them.

Sorry but Intel doesn't sell billions of unlocked chips, that equation doesn't make sense.

Again, Intel only sells unlocked chips in the tens of millions ~ even assuming a dollar or two per enthusiast chip the figure won't breach the 8 digit mark. The locked chips can still be their usual TIM but there's no good reason why unlocked chips should be using the same TIM, of course as Dave said the others dry out faster but there must be some other TIM(s) that can last their warranty period & perform better.

 

[Solaris17]

Sorry but Intel doesn't sell billions of unlocked chips, that equation doesn't make sense.

Again, Intel only sells unlocked chips in the tens of millions ~ even assuming a dollar or two per enthusiast chip the figure won't breach the 8 digit mark. The locked chips can still be their usual TIM but there's no good reason why unlocked chips should be using the same TIM, of course as Dave said the others dry out faster but there must be some other TIM(s) that can last their warranty period & perform better.

maybe but even so it would'nt matter in the slightest. If we go by your logic they don't even sell enough unlocked chips for it to make a difference. It also wouldnt make a ton of sense to dispute it because if soldering the IHS which they have done in the past did improve it to a point that more people wanted to bu unlocked CPUs then why did they switch away from it?

Its clear that the segment is too small for it to make a difference. Unlocked chips by nature may be over clockable but on white paper only offer increased default clocks and boost rates most of the time. Intel does not support overclocking. My argument still stands. If even unlocked CPUs hold thermal specification under load in the lab then they wont change it.

Arguing that Intel has to change it because there "perceived" target audience for unlocked chips is overclockers is a fallacy. If its too hot to OC thats not their problem.

 

[R0H1T]

maybe but even so it would'nt matter in the slightest. If we go by your logic they don't even sell enough unlocked chips for it to make a difference. It also wouldnt make a ton of sense to dispute it because if soldering the IHS which they have done in the past did improve it to a point that more people wanted to bu unlocked CPUs then why did they switch away from it?

Its clear that the segment is too small for it to make a difference. Unlocked chips by nature may be over clockable but on white paper only offer increased default clocks and boost rates most of the time. Intel does not support overclocking. My argument still stands. If even unlocked CPUs hold thermal specification under load in the lab then they wont change it.

Arguing that Intel has to change it because there "perceived" target audience for unlocked chips is overclockers is a fallacy. If its too hot to OC thats not their problem.

This reminds me of one particular news item ~ https://www.techpowerup.com/forums/...tels-core-i7-7700-i7-7700k-processors.233018/

Also an unlocked chip is for OCers, to be precise targeting them, & that's not a logical fallacy.

 

[cadaveca]

Also an unlocked chip is for OCers, to be precise targeting them, & that's not a logical fallacy.

Actually, it kind of is. I agree with you in a way, but you have to look at the fine print, as always. This is what Intel says about overclocking:

Overclocking your unlocked Intel® Core™ processor, RAM, and motherboard is a way to custom tune your PC. You can adjust the power, voltage, core, memory settings, and other key system values for ultimate performance. It speeds up your components—and your gameplay. It can also help with processor-intensive tasks such as image rendering and transcoding

but then fine print says:

Altering clock frequency or voltage may damage or reduce the useful life of the processor and other system components, and may reduce system stability and performance. Product warranties may not apply if the processor is operated beyond its specifications. Check with the manufacturers of system and components for additional details.

https://www.intel.ca/content/www/ca/en/gaming/overclocking-intel-processors.html

Intel also provides a guide for overclocking.

The second item lsited says:

1.2 - Apply aggressive cooling
Utilize a robust cooling solution which can cool the processor well beyond the minimum processor requirements. Liquid cooling is ideal, and added chassis fans will further improve overclocking potential.

https://www.intel.ca/content/www/ca/en/gaming/how-to-overclock.html


So yeah... they say you need "aggressive" cooling, and liquid is ideal... and if you don't understand what "well beyond the minimum processor requirements" means, well...


So, let's look at the 8700K. Specification (minimums, mind you) calls for a 130W cooler for a 95W CPU. And if you want to overclock, they say you need cooling that is WELL BEYOND minimum, which is 130W. Well beyond. You can find the specifications for cooling in their whitepapers for the platform.

So if you aren't using watercooling, according to Intel, your cooling solution is NOT ideal. Air coolers need not apply.

 

[R0H1T]

Actually, it kind of is. I agree with you in a way, but you have to look at the fine print, as always. This is what Intel says about overclocking:



but then fine print says:



https://www.intel.ca/content/www/ca/en/gaming/overclocking-intel-processors.html

Intel also provides a guide for overclocking.

The second item lsited says:



https://www.intel.ca/content/www/ca/en/gaming/how-to-overclock.html


So yeah... they say you need "aggressive" cooling, and liquid is ideal... and if you don't understand what "well beyond the minimum processor requirements" means, well...


So, let's look at the 8700K. Specification (minimums, mind you) calls for a 130W cooler for a 95W CPU. And if you want to overclock, they say you need cooling that is WELL BEYOND minimum, which is 130W. Well beyond. You can find the specifications for cooling in their whitepapers for the platform.

So if you aren't using watercooling, according to Intel, your cooling solution is NOT ideal. Air coolers need not apply.

The thing is even with aggressive cooling the limiting factor will always be TIM, more so with liquid cooling or the more exotic dry ice, liquid helium, LN2 et al. So you need to go bare die or delid & replace the TIM to get the best results, though it has to be said that's mainly for the top 10% of the enthusiast crowd.

Essentially Intel is gimping their own products, now I can't comment on the economics of it ~ as in how selling a few million unlocked K chips with better TIM affects their bottomline wrt 10x more of the locked variety they'd sell with regualr TIM. In essence this is what I want from them, it's not a necessity nor is it a deal breaker for others, but repaying the enthusiast crowd who've stood by them for the last decade goes a long way in redeeming their brand name, especially these days with the constant stream of bad PR.

 

[RejZoR]

What's the point of using top of the line TIM on top when it's always that turd useless toothpaste underneath IHS making everything suck. You can do whatever you want and that TIM underneath will be the limiting factor, not what you use on top. Given they use that crap even on top of the line processors that cost strong triple digit prices is worrying and crappy. And perfectly explains why my old 5820K at 4.5GHz runs cooler than newest 14nm processors with same core count and same clocks. It's that toothpaste underneath IHS. Coz 5820K was still soldered and the new stuff isn't.

Who cares about micro cracks, chances of happening are equal to winning a lottery, but to avoid that you run an inferior product every single day. How does that make any sense? Solder the damn thing.

 

[rtwjunkie]

So if you aren't using watercooling, according to Intel, your cooling solution is NOT ideal. Air coolers need not apply.

There are plenty of good 200w and higher air coolers that work quite well. Just in our house we use Enermax T-40 Fit and Enermax T-50A., 200w and 250w thermal capacities respectively.

 

[las]

To me it should be segregated, similar to what AMD is doing. I don't care that a Celeron, Pentium o normal iX has TIM, but an unlocked K should always be soldered.

I agree. A CPU meant for OC, should be able to OC well without needing to delid. K models should be soldered for sure.

Not many 8700K/8600K/7700K/7600K comes close to 5 GHz without delid or insane load temps, even on high-end cooling solutions.

People are delidding tho. Warrenty gone. Intel happy. If they break a CPU doing it, they simply buy a new CPU... Win/Win for Intel. Even with all the delid tools, lots of people are fucking up.

 

[eidairaman1]

Better yet just go back to exposed dies like P3 and Athlon

 

[EarthDog]

I agree. A CPU meant for OC, should be able to OC well without needing to delid. K models should be soldered for sure.

Not many 8700K/8600K/7700K/7600K comes close to 5 GHz without delid or insane load temps, even on high-end cooling solutions.

People are delidding tho. Warrenty gone. Intel happy. If they break a CPU doing it, they simply buy a new CPU... Win/Win for Intel. Even with all the delid tools, lots of people are fucking up.

So, wait... 4.5ghz on 10c/20t or 16c/32t is not a good overclock? Thats 900 mhz over its all core boost. I believe 1.1ghz over all core boost on the other chip. 8700k are mostly 5ghz chips. Running one now not delidded on corsair h110i.

 

[IceShroom]

A moment before you jump off the deep end.
1) Thermal energy is conducted 3 ways; convection, conduction, and radiation.
2) Thermal conductivity, as expressed, is conduction. It depends upon an area of surface contact. Thermal pastes, and solder, are designed to maximize functional area of contact. They don't need to be great conductors, only increase surface area at a rate greater than they internally resist thermal energy transfer. This is why materials like toothpaste actually match or beat higher quality TIM in tests. The proof is in the interface not drying out, to maintain that heat transfer.
3) The best thermal pastes are still largely insulation. You bang on about how important the value is, without really understanding the application. Please, don't spread that kind of misunderstanding. Take this as a personal request, as half understanding often leads to wholly incorrect statements.
4) Here's a moment to consider what you're saying. If price is no object I can beat copper. It's simple to suggest platinum and a manufacturing process of fricatively welding rods to platinum plate. That would give us better performance, but the increase in performance doesn't match the huge increase in cost. Likewise, soldering is a huge cost for a few degrees, which "nobody" in their consumer base (namely business, not enthusiasts) will see. Why fix what is cheaper and demonstrably not broken?


To the other end of things.
Solder is a surface coating of liquified metals. These metals form a mechanical bond by flowing into the surface imperfections, and creating a connection. As the surface is being heated, and silicon isn't a great conductor, the components don't fry if soldered. This is different when heat is applied for a long time, and the internal components have enough energy transfer to fry.

This is the problem, and why soldering is costly. If you solder too long the chip fries, if you don't solder long enough the connection isn't made and thermal performance is poor.



Intel started using thermal paste because it's cheaper by an order of magnitude. Assuming that you can get the IHS to slightly deform under the pressure applied by the heatsink, you wind up with an interface that very closely matches that of the die. Why we're running into issues is that the dies and IHS are separated by too much distance. This is likely due to regular variability in production of the dies, and IHS plates. I'm just conjecturing here, but the push for lower prices generally leads to larger variation in quality. Intel is likely pushing for those sweet profit margins, given the fact that AMD has actually made demonstrable market progress (whether you appreciate Ryzen or not). Intel is always going to seek profit, and if they can cut a large cost while not negatively impacting their largest consumer base (namely business, not enthusiasts) it's a no-brainer.



If it isn't clear, I voted other. Intel needs to better control their IHS production, or their IHS to die bonding process. The TIM being replaced is interesting, but most people are forgetting that the very small amount of spacing from the adhesive (IHS to die) is being removed. That doesn't sound like a lot, but radiation<convection<conduction. I think that if Intel managed a closer bond, with their current paste, we'd see better performance. Not solder performance, but good enough that delidding wouldn't be a thing (the risk to reward would be too low).

I am not spreading wrong information. Maybe you need to learn how solder and TIM works.
You may not see it but, I wrote at same size means same surface area. At same area Solder with high thermal conductivity will perform better than the best TIM, even better than liquide TIM.
And about burning chips, that happen faster with crappy TIM than solder.(And don't give the crappy Xbox soldering(different kind of soldering) problem as example).
And about cost, i5, i7 and i9 are not cheap cpus. They are mid-high to high end cpus. According to intel fanboy intel has best foundary, that means less chip get wested. Also intel's i5 and i7 are samll cpu, which has less chance of being deffective. So how it would cost to intel to solder a cpu? I bet its negligible. Its not like intel is not making any profit.

 

[cadaveca]

There are plenty of good 200w and higher air coolers that work quite well. Just in our house we use Enermax T-40 Fit and Enermax T-50A., 200w and 250w thermal capacities respectively.

A single 120mm rad is capable of 350w -500W. 200, 250, is not ideal. That's how you get rid of temperature spiking problems... you cool the CPU to the point those spikes are well within limits.

People are delidding tho. Warrenty gone. Intel happy. If they break a CPU doing it, they simply buy a new CPU... Win/Win for Intel. Even with all the delid tools, lots of people are fucking up.

At manufacturing time, just attaching solder to CPU can kill it, or create a hot spot because of uneven application. You do understand that attaching solder means melting metal? You get one attempt, and then you may not be able to remove the IHS ever again. A paste doesn't have that problem. Yeah, people are breaking shit with paste... imagine the added deaths from solder being present.


I simply see this as a case of people being the most selfish, self-centered pieces of crap ever. ME ME ME, DO IT MY WAY OR YOU SUCK!

Worst part of humanity, right here, acting like they know better.

 

[rtwjunkie]

A single 120mm rad is capable of 350w -500W. 200, 250, is not ideal. That's how you get rid of temperature spiking problems... you cool the CPU to the point those spikes are well within limits.

At manufacturing time, just attaching solder to CPU can kill it, or create a hot spot because of uneven application. You do understand that attaching solder means melting metal? You get one attempt, and then you may not be able to remove the IHS ever again. A paste doesn't have that problem. Yeah, people are breaking shit with paste... imagine the added deaths from solder being present.


I simply see this as a case of people being the most selfish, self-centered pieces of crap ever. ME ME ME, DO IT MY WAY OR YOU SUCK!

Worst part of humanity, right here, acting like they know better.

Lol, ok, I’m guessing the final 90% of that didn’t apply to me?

First couple yes. We’re fine with our coolers, and don’t have temp spiking. Everything is well within control.

 

[HTC]

I've never delidded a CPU but my understanding of the problem is that the TIM Intel uses is just bad VS other TIMs currently available. If intel used ... say ... noctua's TIM or MX4 under the IHS, there would be no need to delid @ all, except for those very few that use liquid metal TIMs to squeeze that last bit of performance out of the chip. As such, this whole TIMed VS soldered would be mostly moot, no?

Or is my reasoning flawed? I admit, i don't know much about this topic so i may very well be saying something totally preposterous.

 

[rtwjunkie]

I've never delidded a CPU but my understanding of the problem is that the TIM Intel uses is just bad VS other TIMs currently available. If intel used ... say ... noctua's TIM or MX4 under the IHS, there would be no need to delid @ all, except for those very few that use liquid metal TIMs to squeeze that last bit of performance out of the chip. As such, this whole TIMed VS soldered would be mostly moot, no?

Or is my reasoning flawed? I admit, i don't know much about this topic so i may very well be saying something totally preposterous.

This too is a subject of much debate. Yes, some TIM’s are higher quality than others. All however, really are adequate, and performance is fairly close temperature wise between them.

So sure, some may reduce a couple of degrees more than another. In the grand scheme of things, IMHO, it matter not at all, but is enough for people to have their favorite brands.

To ultimately answer your question, it wouldn’t have made much real difference in the brand used inside the ihs. Just 2-3 degrees at most IMO.

 

[qubit]

8700k are mostly 5ghz chips. Running one now not delidded on corsair h110i.

Are you gonna delid it at some point then? I reckon you're hardcore enough to do it.

 

[HTC]

This too is a subject of much debate. Yes, some TIM’s are higher quality than others. All however, really are adequate, and performance is fairly close temperature wise between them.

So sure, some may reduce a couple of degrees more than another. In the grand scheme of things, IMHO, it matter not at all, but is enough for people to have their favorite brands.

To ultimately answer your question, it wouldn’t have made much real difference in the brand used inside the ihs. Just 2-3 degrees at most IMO.

Only? Since i've heard 15-20 degrees when using liquid metal (not 100% sure of this, nor if this is celsius or farenheit), i was expecting more, tbh. Not as much as liquid metal but something in the range of ... say ... 10 degrees?

 

[DRDNA]

The delid scenario really is for the extreme overclocker and the high end laptop segment where both are worth the risks, there are even Professional Businesses that will do the CPU, GPU, VRM and with very impressive results on the laptops with high end gear in them. That means a High end laptop with NO throttling, some with desktop processors in them and desktop gpu's too.

EDIT, granted it is all stuff most every one here could do as well.

 

[Vario]

I've never delidded a CPU but my understanding of the problem is that the TIM Intel uses is just bad VS other TIMs currently available. If intel used ... say ... noctua's TIM or MX4 under the IHS, there would be no need to delid @ all, except for those very few that use liquid metal TIMs to squeeze that last bit of performance out of the chip. As such, this whole TIMed VS soldered would be mostly moot, no?

Or is my reasoning flawed? I admit, i don't know much about this topic so i may very well be saying something totally preposterous.

The TIM that intel uses is fine, the problem is the gap between the heatspreader and the die that the TIM fills. Some chips have more gap than others. I believe my 8600K has a really ideal minimum gap because my processor sits at 26 idle and 60 load at 4.8 Ghz. Don't have any heat problems with it, it is a great processor.

 

[rtwjunkie]

Only? Since i've heard 15-20 degrees when using liquid metal (not 100% sure of this, nor if this is celsius or farenheit), i was expecting more, tbh. Not as much as liquid metal but something in the range of ... say ... 10 degrees?

From all I have heard of liquid metal, yes, that may be an exception. Cannot personally verify tho, the huge decrease in temps.

 

[EarthDog]

Are you gonna delid it at some point then? I reckon you're hardcore enough to do it.

Nope.