does AMD have to solder their CPUs

[Nogghan]

now hear me out i dont just want to watch the internet burn, im curious about why AMD uses indium sheets and intel uses TIM and i thought that if AMD didnt have solder on their dies wouldnt they all just thermal throttle? they already seem to run at like 70+c and im not saying intel is any better because you as the end user has to replace the TIM which is overall worse but they seem to be able to get away with it because it seems like their chips run cooler in general 

i know its also hard to compare the temps of any 2 AMD and intel CPUs its really apples to oranges but im curious if AMD didnt use solder sheets like intel does now would they be able to scrape by in the same way intel does by doing the bare minimum of heat transfer? 

 

i do know the raven ridge APUs use TIM but they also seem to be different to a regular CPU and again not really directly comparable because they dont seem to get as hot as the regular ryzen chips anyways 

 

 

[DarkSmith2]

Soldering isnt the best solution thermal wise. You can still delid your soldered CPU and gain about 4°C better temperatures using Liquid Metal.

The real reasons why Intel isnt soldering aren't known, nobody knows. But there are arguments for soldering and against it, nothing is perfect in the end its still a mass produced product with flaws no matter what, no matter if its from Intel or AMD.

[Enderman]

Well, the previous generations of AMD chips were only rated for max temps of like 60-70C, not sure if that's changed much for ryzen CPUs but for the most part intel is fine at higher 80-90C temperatures.

[DarkSmith2]

1 minute ago, Enderman said:

Well, the previous generations of AMD chips were only rated for max temps of like 60-70C, not sure if that's changed much for ryzen CPUs but for the most part intel is fine at higher 80-90C temperatures.

But only on paper 80-90C temperatures arent healthy for any type of semiconductor in long time use.

[Enderman]

14 minutes ago, DarkSmith2 said:

But only on paper 80-90C temperatures arent healthy for any type of semiconductor in long time use.

On paper they are fine up to 105C.

80-90C is fine for regular use.

Of course lower is always better but it will still last a decade at those higher temps no problem.

[DarkSmith2]

5 minutes ago, Enderman said:

On paper they are fine up to 115C.

80-90C is fine for regular use.

Of course lower is always better but it will still last a decade at those higher temps no problem.

well actually not really... it depends but semiconductors for PCs are usually rated for 70°C. They might work with higher temps and wont die within warranty but degrade very fast above specified temperatures. 

[Mira Yurizaki]

There's no definitive reason from Intel or otherwise why they switched to using TIM, but I found an article of sorts that provided some speculation as to why they did it. It basically boils down to Intel's CPUs have concentrated hot spots. This causes uneven heating and exacerbates thermal fatigue. If solder were used, this could create cracks in it, leading to so-called voids that decrease the thermal efficiency. AMD doesn't have this problem because a lot of their CPU dies have hot spots that are more spread out. But using TIM to avoid thermal fatigue issues appears to be supported further by the fact that Ryzen 3 and 5 APUs have TIM in them instead of being soldered.

 

The article also shows that as long as the thickness of the TIM is thin enough, then its thermal conductivity plays even less of a role in heat transfer. Though this was by way of computer simulation, not really of any physical testing (I'd imagine it'd be pretty hard anyway for a person to get within 10um of precision)

 

The article does mention something that's confusing, which is this bit:

Quote

The highest performance compounds do not necessarily have the highest conductivity but have the lowest thermal resistance ...

Which might sound confusing. Isn't thermal resistance and conductivity simply reciprocals of each other? Turns out, no. Thermal conductivity is a different thing than the reciprocal of thermal resistance. That value is called thermal conductance. (see: http://ctherm.com/products/tci_thermal_conductivity/helpful_links_tools/thermal_resistance_thermal_conductance/)

 

The TIM that Intel uses (or has used), Dow Corning's TC-1996, isn't some cheap goop that people like to claim it is. It's a pretty decent performer as far as one review goes: https://www.hardwaresecrets.com/thermal-compound-roundup-august-2011/5/

 

In any case, as far as I know, nobody has done any real formal study on the long term effects of using one or the other. Everyone just has speculation of the end result. But keep in mind that Intel is fully aware that their largest customer base, the professional and server market, is glacially slow to upgrade their hardware. For the sake of keeping customers, they can't just slap on cheapo thermal grease that gives out in three years and causes the processor to fry itself a year later. And heck, we have people here still using Core 2 processors, and that seems to be the earliest Intel started using TIM instead of soldering.

[aezakmi]

maybe they should get rid of the IHS and go back to basics

that way no one will complain about solder vs tim and we'll have lower temps.

[Nogghan]

2 minutes ago, aezakmi said:

maybe they should get rid of the IHS and go back to basics

that way no one will complain about solder vs tim and we'll have lower temps.

nobody will have time to complain about TIM because the torque specs for the screws were slightly off and now my DIE is cracked THANKS INTEL/AMD please... keep the IHS unless you have a frame or an actual reliable torque spec for the mounting force and even then its still a risk

Edited June 4, 2018 by Nogghan

[mariushm]

33 minutes ago, Nogghan said:

now hear me out i dont just want to watch the internet burn, im curious about why AMD uses indium sheets and intel uses TIM and i thought that if AMD didnt have solder on their dies wouldnt they all just thermal throttle? they already seem to run at like 70+c and im not saying intel is any better because you as the end user has to replace the TIM which is overall worse but they seem to be able to get away with it because it seems like their chips run cooler in general 

i know its also hard to compare the temps of any 2 AMD and intel CPUs its really apples to oranges but im curious if AMD didnt use solder sheets like intel does now would they be able to scrape by in the same way intel does by doing the bare minimum of heat transfer? 

 

i do know the raven ridge APUs use TIM but they also seem to be different to a regular CPU and again not really directly comparable because they dont seem to get as hot as the regular ryzen chips anyways 

 

 

Intel uses TIM because they're cheap. They save maybe 20-40 us cents this way and an additional manufacturing process. A very low amount of silicon dies are also damaged by applying the indium solder. AMD uses TIM on the APUs to save costs and because those are only 4 core variants and will not be as hot as the 6 or 8 core variants (and the load on cpu and video won't be max all the time in anything outside benchmarks)

 

Intel uses a TIM that's designed to maintain its consistency over a longer period of time, not necessarily to achieve very good heat transfer. That's why there's huge gains if you remove the TIM and use liquid metal instead, maybe up to 20 degrees Celsius. 

 

Yeah, the indium solder isn't the super bestest ever, but it's really quite efficient. You could lower the heat up to 3-4 degrees celsius by removing the solder and using liquid metal, but are the tradeoffs worth it (like reapplying liquid metal every 6 months or so)? Probably not. You'd probably reduce by 2 degrees celsius just by lapping the lid surface with several sandpaper sheets (or whatever they're called)

 

It's ok for these processors to run at 70c, these modern processes can run fine up to around 95-100 degrees celsius and stay there 24/7 .. the silicon die won't mind.  High temperature isn't good for components around like circuit board and capacitors, and to some degree the insulation on inductor wire (usually rated for >150c), but the die itself will survive.

 

In some cases, it's actually advantageous to run them at a decent temperature, there's a narrow range of temperature and voltage where the transistors inside the silicon die will have the lowest leakage, lowest losses etc, so you get the highest performance for the lowest amount of energy consumed.  At that point, if you cool the cpu more, you're not getting more performance, you're only getting more power consumed for same performance.

 

For AMD processors at least, it's normal and acceptable for them to function at 70 degrees, there's little benefit to force them to stay cooler all the time, so that's why the fan speeds are often "tuned" around that temperature.

[ravenshrike]

8 minutes ago, M.Yurizaki said:

 

The TIM that Intel uses (or has used), Dow Corning's TC-1996, isn't some cheap goop that people like to claim it is. It's a pretty decent performer as far as one review goes: https://www.hardwaresecrets.com/thermal-compound-roundup-august-2011/5/

The issue is that the amount used inside the IHS is thicker than what is used with a proper heatsink on top of the IHS.

[Nogghan]

1 minute ago, ravenshrike said:

The issue is that the amount used inside the IHS is thicker than what is used with a proper heatsink on top of the IHS.

so you think intel should just... use less TIM?...? what?

[Nogghan]

5 minutes ago, mariushm said:

 

It's ok for these processors to run at 70c, these modern processes can run fine up to around 95-100 degrees celsius and stay there 24/7 .. the silicon die won't mind.  High temperature isn't good for components around like circuit board and capacitors, and to some degree the insulation on inductor wire (usually rated for >150c), but the die itself will survive.

 

okay yes 70c is fine but 95c is reaching thermal throttling for some processors today and 100c will throttle ANY processor there are none rated to use over 100c its just flat false what the hell are you talking about  also caps and semiconductors are usually rated for less than 125c but can withstand up to about 150c if they are of good quality which not even every Z370 board uses good quality capacitors most are rated for below 95c actually 

Edited June 4, 2018 by Nogghan

[Mira Yurizaki]

4 minutes ago, ravenshrike said:

The issue is that the amount used inside the IHS is thicker than what is used with a proper heatsink on top of the IHS.

That has been explained by the increased spacing of the IHS Intel has done in the past from using too much IHS binder, which was pointed out by notable delidder Der8auer. At least this was the case for Haswell.

5 minutes ago, mariushm said:

-Snip-

Until someone can find me an actual BOM for the cost of using solder (plus apparently gold plating) vs. TIM, this is purely speculation. Besides, from a consumer standpoint, the TIM Intel has used costs more per mass than even lead free solder. And I'm pretty sure even with gold plating (which actually might not be plating at all), you don't need a whole lot of it for the solder to stick.

 

People continue to speculate but don't seem to have any real hard evidence from the manufacturers themselves. But you're welcome to show me otherwise.

Edited June 4, 2018 by M.Yurizaki

[DarkSmith2]

1 minute ago, Nogghan said:

so you think intel should just... use less TIM?...? what?

from what ive discoverd underneath the IHS of my 8700k, hell yea. Its like they put the same amount like some companies sell in a tube LUL

[Nogghan]

2 minutes ago, DarkSmith2 said:

from what ive discoverd underneath the IHS of my 8700k, hell yea. Its like they put the same amount like some companies sell in a tube LUL

interesting when i delidded my 8600k there was just a little bit of spillage hardly more than what id say is the bare minimum it certainly wasnt a whole gram tube of compound and i arguably did not even need to delid the chip it was under 80c at 5ghz anyways so i did it just because i dont like seeing 80c now i see 60c

Edited June 4, 2018 by Nogghan

[aezakmi]

5 minutes ago, Nogghan said:

nobody will have time to complain about TIM because the torque specs for the screws were slightly off and now my DIE is cracked THANKS INTEL/AMD please... keep the IHS unless you have a frame or an actual reliable torque spec for the mounting force and even then its still a risk

back in the day you had to be really dumb to crack a CPU, the 478s had pads and the heatsink mount lacked screws, and the 370 wasn't any different

[DarkSmith2]

5 minutes ago, Nogghan said:

interesting when i delidded my 8600k there was just a little bit of spillage hardly more than what id say is the bare minimum it certainly wasnt a whole gram tube of compound 

really? my die was completely burried couldnt even see anything else than the thermal compound when i opened it up. like half of the IHS was literally filled with it LUL

 

 

and i arguably did not even need to delid the chip it was under 80c at 5ghz anyways so i did it just because i dont like seeing 80c now i see 60c

Same but still LUL

[Enderman]

16 minutes ago, DarkSmith2 said:

well actually not really... it depends but semiconductors for PCs are usually rated for 70°C. They might work with higher temps and wont die within warranty but degrade very fast above specified temperatures. 

Nope.

 

There is no "line" where it magically starts degrading above and not below.

Those are just general arbitrary values.

The higher the temperature is, the more it degrades, it is a sliding scale not a clear line.

 

There's plenty of evidence that intel CPUs can run over 80-90C long term without any damage that would clearly be distinguishable from regular failure.

[Nogghan]

2 minutes ago, DarkSmith2 said:

really? my die was completely burried couldnt even see anything else than the thermal compound when i oppened it up. like half of the IHS was literally filled with it LUL

mine looked crusty already and i could see like drying cracks in the thinness of the application on my 8600k so.... thats something people should look into IMO

[DarkSmith2]

6 minutes ago, Nogghan said:

mine looked crusty already and i could see like drying cracks in the thinness of the application on my 8600k so.... thats something people should look into IMO

mine wasnt crusty or anything, just alot of it i think it havnt hurt anything but was a mess to clean up. Ive heard they changed the thermal paste application of 8700k's during some manufacturing process, maybe i had one of the newer ones..  well i definitely have one of the "newer" ones (46th week 2017) but i mean already with changes.

[ravenshrike]

11 minutes ago, Nogghan said:

so you think intel should just... use less TIM?...? what?

Whatever system Intel has in place for applying the IHS seems less precise than AMD's given the variance they experience in die spacing. They can get away with it because of paste, but it's certainly not optimal. In order to confirm of course someone would have to buy at least 50 raven ridge CPUs preferably from as many different batches as possible to check and see.

[Nogghan]

1 minute ago, DarkSmith2 said:

mine wasnt crusty or anything, just alot of it i think it havnt hurt anything but was a mess to clean up. Ive heard they changed the thermal paste application of 8700k's during some manufacturing process, maybe i had one of the newer ones..  well i definitely have one of the "newer" (46th week 2017) but i mean already with changes.

maybe its just because its an 8700k v an 8600k but IMO they should be the equal thermal paste applications, i did buy the 8600k in the first week of 2018 and it could easily have been an older stock because its a much less popular model 

[DarkSmith2]

Just now, Nogghan said:

maybe its just because its an 8700k v an 8600k but IMO they should be the equal thermal paste applications, i did buy the 8600k in the first week of 2018 and it could easily have been an older stock because its a much less popular model 

Its written in your batchnumber. i had L746D346  meaning:

L Malaysia

7 is 2017

46 is 46th week

D waver

346 waver position

[Nogghan]

2 minutes ago, ravenshrike said:

Whatever system Intel has in place for applying the IHS seems less precise than AMD's given the variance they experience in die spacing. They can get away with it because of paste, but it's certainly not optimal. In order to confirm of course someone would have to buy at least 50 raven ridge CPUs preferably from as many different batches as possible to check and see.

im pretty sure AMD ravenridge APU have a .5mm  thicker plate over the IHS to replace the solder and give the same level of contact as the solder sheet that usually takes up about .5mm of space 

but what do you mean by the system intel has in place for applying an IHS? im almost 100% sure its a factory line with some robot arms that have glue and a pressure plate