Intel 18c HEDT Die Pictured, Not Soldered (7980XE)

[HKZeroFive]

9 hours ago, snortingfrogs said:

Xeon chips has a much smaller temperature variation due to how they are used, ie servers with much better cooling (albeit much louder) than consumer/enthusiast computers so it's not really comparable.

Unless I've been completely unaware, Broadwell-E chips (which are soldered) didn't have any such problems/failures that crippled the platform. On top of that, AMD's 14nm Ryzen chips are soldered and while you can argue that they're not as dense, there has been no evidence that the difference matters at all unless you can provide a source that says otherwise.

 

The excuse might be valid for Skylake-S, but when you're talking about the large dies of Skylake-X, that same excuse doesn't hold up at all.

[8uhbbhu8]

7 hours ago, MageTank said:

Yay, another MageTank prophecy has come to be true! Bow down to me mortals, for I know all!*

 

*MageTank does not know why kids love the taste of cinnamon toast crunch, nor can he provide the number of licks required to make it to the center of a tootsie pop. These are the only known limitations of his powers. 

7 hours ago, MageTank said:

I plan on casting a spell on an actual tank. A tank that shoots magic missiles. I'll also be wearing plate armor under the robes, doubling as a MageTank with a magical tank. 

 

As for the topic, @done12many2 was already spot on. While thermals would certainly be better with Solder, it's safe to assume the larger die chips are going to be better than the smaller die chips, even at similar clocks with higher core counts. Hopefully Der8auer can minimize risk with his delid tool, and people will at least be able to salvage that aspect of them. It's going to be tricky, but we've already seen delidded results that survived on HWbot. I can certainly say those that did it, got a whole lot more out of it. 

Don't forget your obligatory Fireball!!

4 hours ago, MageTank said:

And trust their weak stress tests?!? No thank you. They are far too afraid of causing a literal fire under Linpack MKL. Meanwhile, I keep a bag of marshmallows on hand just in case. 

Oh? Didn't know about that! lol Want to know more

[Dabombinable]

2 hours ago, snortingfrogs said:

One of the reasons Intel doesn't use solder is due to micro cracks.
Micro cracks are tiny cracks that create small "bubbles" in the solder that are formed due to the constant variation in temperatures and those "bubbles" are not good for heat dissipation at all and will make the CPU last a shorter time.

 

It wasn't such a big problem before, but with Intel Speedstep, smaller chips, higher clocks (creating more heat) the shift between temperatures are much bigger and with Intel Speedstep the shift between hot and cold is much smaller making the micro cracks appear faster thus shortening the life of the CPU.

My Pentium 4 630 and 631 have far smaller dies and run very hot. They to this day still work like new, with no loss in their ability to dump heat and stay cool at the same time.

[Jito463]

8 hours ago, MageTank said:

Once upon a time, I knew those answers, but the moment I hit 18, the knowledge was stripped from me. Along with my ability to consume Trix. Life is cruel.

Silly MageTank, Trix are for kids!

 

8 hours ago, snortingfrogs said:

It's cute how some think they know better than Intel's thermal engineers...

It's possible they were overridden by corporate to save some money.  No matter how smart an engineer may be, they still have to answer to the guys running the company.  As I understand it, the process of soldering the IHS could lead to failed chips (someone correct me if I'm mistaken); and given the size of die they're working with, the yields probably aren't that exceptional to begin with.  Thus I could completely understand if cost was their motivating factor in not soldering the IHS to the die.

 

I don't approve necessarily, but I can understand.

 

By the way, throwing around the term "engineer" doesn't mean they're the smartest people on the planet.  Even smart people make mistakes.  That's not even remotely an argument.

[mr moose]

1 hour ago, Jito463 said:

By the way, throwing around the term "engineer" doesn't mean they're the smartest people on the planet.  Even smart people make mistakes.  That's not even remotely an argument.

 

I think the idea behind using the term "engineer" is to illustrate the difference in education and experience between them and internet plebs, especially those of us on the forums who are guilty of making derisory comments in such a manor that would indicate we think we could do a better job.

 

 

[Coaxialgamer]

11 hours ago, jaysangwan32 said:

Why is this a significant problem? I understand that heat dissipation would be better if soldered, but if it's hitting 4.8 on water anyway (not sure if it was delidded), does it matter?

 

It's more of a problem by principle. Even if it performs well enough, intel shouldn't be cheaping out on a 2000$ cpu. 

It shouldn't be that a modern chip that costs 2 grand is worse than last generation's 400$ 6 core chip in any way.

[Coaxialgamer]

3 hours ago, Jito463 said:

Silly MageTank, Trix are for kids!

 

It's possible they were overridden by corporate to save some money.  No matter how smart an engineer may be, they still have to answer to the guys running the company.  As I understand it, the process of soldering the IHS could lead to failed chips (someone correct me if I'm mistaken); and given the size of die they're working with, the yields probably aren't that exceptional to begin with.  Thus I could completely understand if cost was their motivating factor in not soldering the IHS to the die.

 

I don't approve necessarily, but I can understand.

 

By the way, throwing around the term "engineer" doesn't mean they're the smartest people on the planet.  Even smart people make mistakes.  That's not even remotely an argument.

Issues with soldering chips are only apparent on very small dies that don't have the necessary area to deal with the actual soldering process and are more likely to be damaged by indium's thermal expansion. It isn't an issue on large chips. 

[TidaLWaveZ]

6 hours ago, Jito463 said:

-snip-

By the way, throwing around the term "engineer" doesn't mean they're the smartest people on the planet.  Even smart people make mistakes.  That's not even remotely an argument.

A single thermal engineer didn't just decide not to solder these one day, a large team of well paid engineers did. By well paid I mean hired due to excellent credentials.

 

That team of engineers is collectively smarter than anyone on this forum when it comes to the topic at hand.

 

The only way the original statement doesn't stand is if the thermal engineers opinions/decisions are superseded to save time/money, which very well may be the case.

[Bananasplit_00]

nice, now lets see how it destroys VRMs!

[snortingfrogs]

A good read regarding soldering cpu's: https://overclocking.guide/the-truth-about-cpu-soldering/

[Kinda Bottlenecked]

8 minutes ago, snortingfrogs said:

A good read regarding soldering cpu's: https://overclocking.guide/the-truth-about-cpu-soldering/

 

Quoting the the good stuff

 

Quote

Void and micro crack occurrence is mainly affected by the solder area – thus the DIE size. Small DIE size (below 130 mm²) e. g. Skylake will facilitate the void occurence significantly. However, CPUs with a medium to large DIE size (above 270 mm²) e. g. Haswell-E show no significant increase of micro cracking during thermal cycling

 

[snortingfrogs]

1 minute ago, Pohernori said:

 

Quoting the the good stuff

 

 

I rather focus on the conclusion

 

Conclusion
Whenever I read sentences like “What a ripoff – Intel doesn’t even solder a 300 USD CPU” or “Why does intel save 2 USD on soldering” I’m thinking

Stop hating on Intel. Intel has some of the best engineers in the world when it comes to metallurgy. They know exactly what they are doing and the reason for conventional thermal paste in recent desktop CPUs is not as simple as it seems.

Micro cracks in solder preforms can damage the CPU permanently after a certain amount of thermal cycles and time. Conventional thermal paste doesn’t perform as good as the solder preform but it should have a longer durability – especially for small size DIE CPUs.

 

Thinking about the ecology it makes sense to use conventional thermal paste. Gold and indium are rare and expensive materials. Mining of these materials is complex and in addition it’s polluting.

After soldering one of my 6700K CPUs I can tell it’s a pretty complex process. I’m still working on it and trying to make it available for extreme overclockers. However, I doubt that Intel will come back with soldered “small DIE CPUs”. Skylake works great even with normal thermal paste so I see no reason why Intel should/would change anything here.

[The Benjamins]

16 minutes ago, snortingfrogs said:

A good read regarding soldering cpu's: https://overclocking.guide/the-truth-about-cpu-soldering/

This give no explanation why these huge dies are not soldered.

[PCGuy_5960]

15 hours ago, ZackBarletto said:

looks like a CPU attached to a socket converter type pcb think. lol

It's probably because Intel is using the same package for the HEDT and server CPUs.

It has 2 PCBs, that's interesting. Did the older generations had this too?

Broadwell-E had 2 substrates

9 hours ago, snortingfrogs said:

One of the reasons Intel doesn't use solder is due to micro cracks.
Micro cracks are tiny cracks that create small "bubbles" in the solder that are formed due to the constant variation in temperatures and those "bubbles" are not good for heat dissipation at all and will make the CPU last a shorter time.

 

It wasn't such a big problem before, but with Intel Speedstep, smaller chips, higher clocks (creating more heat) the shift between temperatures are much bigger and with Intel Speedstep the shift between hot and cold is much smaller making the micro cracks appear faster thus shortening the life of the CPU.

This is true for smaller dies, but based on some rough measurements, the 7900X's die is 319mm^2 and the 7980XE's die is 506mm^2. The 6950X's die size is 250mm^2 (iirc) and it is soldered, so they should have no problem soldering the 7980XE and the 7900X.

 

I believe that yields used to be terrible, so they couldn't solder the CPUs as some of them die during the soldering process. Now, yields have improved, but Intel can't just scrap all the "earlier" 7980XEs, so they keep using thermal paste. Or, they ran into other issues.

[porina]

59 minutes ago, TidaLWaveZ said:

The only way the original statement doesn't stand is if the thermal engineers opinions/decisions are superseded to save time/money, which very well may be the case.

Engineering decisions are only one part of the process to bring a product to market. If you ask the engineer(s) who work on it, the engineering management, and the product management, you'll likely get 3 different answers as to what is the "best" solution.

[done12many2]

4 minutes ago, snortingfrogs said:

I rather focus on the conclusion

 

Conclusion
Whenever I read sentences like “What a ripoff – Intel doesn’t even solder a 300 USD CPU” or “Why does intel save 2 USD on soldering” I’m thinking

Stop hating on Intel. Intel has some of the best engineers in the world when it comes to metallurgy. They know exactly what they are doing and the reason for conventional thermal paste in recent desktop CPUs is not as simple as it seems.

Micro cracks in solder preforms can damage the CPU permanently after a certain amount of thermal cycles and time. Conventional thermal paste doesn’t perform as good as the solder preform but it should have a longer durability – especially for small size DIE CPUs.

 

Thinking about the ecology it makes sense to use conventional thermal paste. Gold and indium are rare and expensive materials. Mining of these materials is complex and in addition it’s polluting.

After soldering one of my 6700K CPUs I can tell it’s a pretty complex process. I’m still working on it and trying to make it available for extreme overclockers. However, I doubt that Intel will come back with soldered “small DIE CPUs”. Skylake works great even with normal thermal paste so I see no reason why Intel should/would change anything here.

 

See the problem with providing an article like the one you shared is those who are looking for a reason to critique will always find that reason.  Aside from die size alone, I'm sure there a couple more variables that impact thermal cycles.  We love to be experts on LTT.  

[Kinda Bottlenecked]

4 minutes ago, snortingfrogs said:

I rather focus on the conclusion

 

Conclusion
Whenever I read sentences like “What a ripoff – Intel doesn’t even solder a 300 USD CPU” or “Why does intel save 2 USD on soldering” I’m thinking

Stop hating on Intel. Intel has some of the best engineers in the world when it comes to metallurgy. They know exactly what they are doing and the reason for conventional thermal paste in recent desktop CPUs is not as simple as it seems.

Micro cracks in solder preforms can damage the CPU permanently after a certain amount of thermal cycles and time. Conventional thermal paste doesn’t perform as good as the solder preform but it should have a longer durability – especially for small size DIE CPUs.

 

Thinking about the ecology it makes sense to use conventional thermal paste. Gold and indium are rare and expensive materials. Mining of these materials is complex and in addition it’s polluting.

After soldering one of my 6700K CPUs I can tell it’s a pretty complex process. I’m still working on it and trying to make it available for extreme overclockers. However, I doubt that Intel will come back with soldered “small DIE CPUs”. Skylake works great even with normal thermal paste so I see no reason why Intel should/would change anything here.

 

Well er... If you want to quote the part that specifically talks about small dies...  

[SpaceGhostC2C]

14 minutes ago, snortingfrogs said:

I rather focus on the conclusion

Yes, let's rather selectively focus on the conclusion of a guy soldering a 6700K, because that's exactly what soldering a huge die would look like, right?

I mean, I get it: selective reading is how you post an article that lends support of the opposite view while thinking it supports yours...

 

The "Intel engineers know what they are doing" is stupid for two reasons. First, because Intel engineers produced both soldered and not-soldered CPUs, and I'm talking about recent generations. So, if anything, Intel engineers seem to think it is perfectly possible to solder a relatively large die (smaller than the 7980XE, mind you) in a commercial product.

Second, because an actual engineer is not a priest: (s)he won't go around shouting how everything must be soldered, always, ever, nor how you should never solder that chip, "that's suicide, what are you doing, the microcracks, oh lord, the microcracks!". Instead, (s)he will tell you: this is the thermal performance with solder. And this is the performance without it. And this is the durability. And this is the frequency of broken CPUs in the assembly process. They may even tell you what the costs are in each case.

It's not a matter of soldered CPUs not working or not being possible. It's not a matter of non-soldered CPUs not working or not being possible. There is a range of possible configurations for the final product, and then a decision must be made. This decision is made by the product management, with input from all areas, including engineers who tell them how the thing is going to work and marketers who tell them what they can possibly charge for it. Whatever decision is finally made, it is by not means the only possible decision. Whether it is soldering, TIM, base clocks, boost states, SMT capabilities, you name it, it all can, and will at times be, different. At the end of the day, someones sits down, does some cost-benefit analysis, and makes a call. Cost-benefit from the point of profits, mind you. Not performance per dollar of cost, as that doesn't guarantee a higher profit margin.

So, did some very smart people at Intel decide that this exact CPU, with all its characteristics, is the most profitable version of the 7980XE for Intel? Yes, most certainly.

Did any smart people at Intel conclude that soldering the IHS to the die was not possible, or not possible without a serious risk for the longevity of the chip (a risk that mysteriously Broadwell-E, Bulldozer, Ryzen, etc don't suffer from)? I would like to see some proof of that, because so far there is no evidence of anyone at Intel claiming such thing (and I mean from an engineering standpoint, not a PR guy telling the press "no, it's not possible").

[The Benjamins]

9 minutes ago, SpaceGhostC2C said:

Yes, let's rather selectively focus on the conclusion of a guy soldering a 6700K, because that's exactly what soldering a huge die would look like, right?

I mean, I get it: selective reading is how you post an article that lends support of the opposite view while thinking it supports yours...

 

The "Intel engineers know what they are doing" is stupid for two reasons. First, because Intel engineers produced both soldered and not-soldered CPUs, and I'm talking about recent generations. So, if anything, Intel engineers seem to think it is perfectly possible to solder a relatively large die (smaller than the 7980XE, mind you) in a commercial product.

Second, because an actual engineer is not a priest: (s)he won't go around shouting how everything must be soldered, always, ever, nor how you should never solder that chip, "that's suicide, what are you doing, the microcracks, oh lord, the microcracks!". Instead, (s)he will tell you: this is the thermal performance with solder. And this is the performance without it. And this is the durability. And this is the frequency of broken CPUs in the assembly process. They may even tell you what the costs are in each case.

It's not a matter of soldered CPUs not working or not being possible. It's not a matter of non-soldered CPUs not working or not being possible. There is a range of possible configurations for the final product, and then a decision must be made. This decision is made by the product management, with input from all areas, including engineers who tell them how the thing is going to work and marketers who tell them what they can possibly charge for it. Whatever decision is finally made, it is by not means the only possible decision. Whether it is soldering, TIM, base clocks, boost states, SMT capabilities, you name it, it all can, and will at times be, different. At the end of the day, someones sits down, does some cost-benefit analysis, and makes a call. Cost-benefit from the point of profits, mind you. Not performance per dollar of cost, as that doesn't guarantee a higher profit margin.

So, did some very smart people at Intel decide that this exact CPU, with all its characteristics, is the most profitable version of the 7980XE for Intel? Yes, most certainly.

Did any smart people at Intel conclude that soldering the IHS to the die was not possible, or not possible without a serious risk for the longevity of the chip (a risk that mysteriously Broadwell-E, Bulldozer, Ryzen, etc don't suffer from)? I would like to see some proof of that, because so far there is no evidence of anyone at Intel claiming such thing (and I mean from an engineering standpoint, not a PR guy telling the press "no, it's not possible").

Also who cares if Intel has the smartest engineers if they have the stupidest (not saying they are) management. if management decides it is not cost effective to go with solder vs paste even if the engineers recommend solder. 

[SpaceGhostC2C]

Just now, The Benjamins said:

Also who cares if Intel has the smartest engineers if they have the stupidest (not saying they are) management. if management decides it is not cost effective to go with solder vs paste even if the engineers recommend solder. 

I just wanted to emphasize that the management can also be the smartest people, and not soldering (along with all other choices) can be the smartest choice for Intel and its profits (which is what the people managing Intel should focus on). That still wouldn't make soldering impossible, nor even bad.

[Jito463]

1 hour ago, TidaLWaveZ said:

A single thermal engineer didn't just decide not to solder these one day, a large team of well paid engineers did. By well paid I mean hired due to excellent credentials.

 

That team of engineers is collectively smarter than anyone on this forum when it comes to the topic at hand.

 

The only way the original statement doesn't stand is if the thermal engineers opinions/decisions are superseded to save time/money, which very well may be the case.

Granted, the engineers working for Intel have much more training and experience than any of us (and well they should, or they shouldn't have their jobs).  That's not in dispute.  My only point of contention is using the argument "engineers are smarter" to try and win the conversation.  That's no argument at all, that's a weak and fallible position to attempt to stand on.

 

And even a "team of engineers" can make a mistake.  I'm not saying that they did, I'm simply pointing out that they could.  Engineers are people, and like all people they are fallible.

[done12many2]

3 minutes ago, Jito463 said:

Granted, the engineers working for Intel have much more training and experience than any of us (and well they should, or they shouldn't have their jobs).  That's not in dispute.  My only point of contention is using the argument "engineers are smarter" to try and win the conversation.  That's no argument at all, that's a weak and fallible position to attempt to stand on.

 

And even a "team of engineers" can make a mistake.  I'm not saying that they did, I'm simply pointing out that they could.  Engineers are people, and like all people they are fallible.

 

Would you also agree that debates among those of us who have no training or experience in manufacturing CPUs is nothing more than pure speculation and a waste of time?

 

It would be one thing if they provided a reason and we argued over that, but we are literally debating over some shit that we have no expertise in.  We're trying to apply logic from previous gens to current gens.  While this works for a great deal of things, we don't realize that it did or didn't until some time later.  

 

I absolutely agree that Intel should have soldered if they could have with certainty of endurance, but I'm in no position to demand as I don't understand.  I'd hate to demand solder and have them comply to only later figure out that they chose not to for a good reason.  That would be a case of giving the customer what they want despite knowing better for them.

 

At the end of the day, these chips are going to clock high and that's really all I'm concerned with.  

[TidaLWaveZ]

1 hour ago, Jito463 said:

-snip

I understand what you're getting at and I don't defend the technicalities against the vagueness of the original comment.

 

I do however defend the stance that the comment was taking which to me is "Intel is not stupid". I can't go a day without seeing at least 15 insinuations that intel is running themselves into the ground.  Years of work and decisions go into every line made from R&D to POS, yet people continually act like the process is as simple as a big fat rich guy saying yes or no like they could run the company better.

 

There are multiple protections/departments/tests/etc in place to ensure that bad ideas/decisions don't make it that far.

[pas008]

so people are bitching about not being soldered but we overclockers are <1% population

but most here are diy and do many other things that risk the product or warranty

 

 

the best comes at a higher prices not just pc components and they still have flaws

if you want piece of mind comes at a cost too like buying with waterblock on, binned, delidded, etc

if you cant risk/afford it you dont buy or mod simple as that

 

negative nancys

can we just be happy we finally are getting a cpu war and next couple yrs are going to be interesting

 

 

[MageTank]

2 hours ago, done12many2 said:

 

Would you also agree that debates among those of us who have no training or experience in manufacturing CPUs is nothing more than pure speculation and a waste of time?

 

It would be one thing if they provided a reason and we argued over that, but we are literally debating over some shit that we have no expertise in.  We're trying to apply logic from previous gens to current gens.  While this works for a great deal of things, we don't realize that it did or didn't until some time later.  

 

I absolutely agree that Intel should have soldered if they could have with certainty of endurance, but I'm in no position to demand as I don't understand.  I'd hate to demand solder and have them comply to only later figure out that they chose not to for a good reason.  That would be a case of giving the customer what they want despite knowing better for them.

 

At the end of the day, these chips are going to clock high and that's really all I'm concerned with.  

No, I would not agree to this at all. It's human nature to question what we do not understand. Without the open dialogue and discussion of these things, we will never get closer to an answer. Does that mean Intel will ever come out and say why they soldered the previously smaller HEDT/Xeon dies, but not the X299 dies? Probably not, but the chances of them doing so improves dramatically if enough people stir up enough dust by discussing it in the first place.

 

The notion that one needs to be an engineer to question the decisions made by engineers is silly. Do I need to be a politician to question a government policy that I disagree with? Bad comparisons aside, people are asking a very simple question. If they were able to solder both Haswell-E and Broadwell-E, along with every Xeon on 2011, why are they unable to solder these even larger dies on 2066? This isn't just the 2066 enthusiast chips that they neglected to solder. They didn't solder the 2066 Xeon-W lineup either. Yet, the LGA3647 Xeon's (bronze, silver, gold, platinum) are all soldered again. Is it seriously a bad thing for people to question why it was okay to solder previous HEDT CPU's, but not the current ones? 

 

I personally cannot think of any reason as to why they refused to solder these, unless they somehow recently discovered that soldering Haswell-E and Broadwell-E was a bad idea, and that they would get far better longevity by going back to Dow Corning on the HEDT platform. The only other alternative (which you and I already discussed in private) is that Intel's marketing/accounting department stepped on the toes of the engineers, and told them not to solder the current lineup. Multiple reasons as to why they would do this, but my personal opinion (and that's exactly what it is, an opinion) is that they are going to refresh this current 2066 lineup with one that comes with solder later on. Basically, another Devil's Canyon but for HEDT. I certainly hope I am wrong, but at this point, I am out of possible reasons for them to go this route. Until we get an official response, this is all we can do.