The Truth About CPU Soldering

[qubit]

According to this article, soldering the heatspreader to the CPU isn't all it's cracked up to be (literally) so using thermal paste can make sense. My question is, in that case, why doesn't Intel use the best thermal paste possible? Surely, the tiny bit extra that it might cost is negligible and the performance increase and lowered temperature, especially when overclocked more than makes up for it? See what you think.

Skylake delidding seems to be very common by now. Every day I read postings from people complaining about Intel and the thermal paste between IHS and die. Even tho Skylake is performing great, people are not satisfied with the temperatures on load. Compared to older generations there is conventional thermal paste between the IHS and the die while Sandy Bridge and older CPUs were soldered. Why did Intel change the production and is the thermal paste really that bad?

https://overclocking.guide/the-truth-about-cpu-soldering

 

[R-T-B]

Hasn't that link been posted in debates like a million times already?

It basically argues it's bad for small dies. It doesn't mean it's bad for big ones like modern 6-core+ CPUs tend to be, however.

And Intel COULD be using better thermal paste, but honestly? They have no real economic reason to do so.

 

[las]

Yes I read that post a few years back and Intel is still having issues with temp spiking. Not sure why it has not been fixed yet. I hope Ice Lake will fix it. New arch / design. If not, Intel will start seeing problems as AMD clocks their Zen arch higher and higher. Current Intel's seem to spike after 4.8. 5 and more requires very good cooling or even delid + repaste.

1st gen Ryzen hits a wall around 4 GHz, but they are still fairly cool. I've clocked a 1600 to 4 GHz using cheap air and 65c max load on hottest core.. Meanwhile 7700K hits like 90c at 4.8 using same cheap air cooler (212 Evo).

Hopefully refreshed Ryzen will do 4.2-4.4 and force Intel's hand to do something about the temp spiking (it also limits boost clocks etc).

 

[dorsetknob]

I have always wondered why they solder the heatspredder on some chips
and yet they will refuse a warranty if the CPU is over heated ????
Surely when soldering the heatspredder on ...>> the temp needed will be far inexcess of anything any average user will ever reach.
Does Intel have Some magic way of Soldering

I have not Dellided since core2 so have no experence of soldered CPU's
IN My Opinion Soldering the heat spredder is a good idea FOR INTEL and the Average user but they also should have a ( Pro line not Soldered for people that wish to Delid and repaste ie Overclockers and high perfomance Benchers).

They Could market Them as ""THE EXTREME RANGE"

 

[delshay]

Hasn't that link been posted in debates like a million times already?

It basically argues it's bad for small dies. It doesn't mean it's bad for big ones like modern 6-core+ CPUs tend to be, however.

And Intel COULD be using better thermal paste, but honestly? They have no real economic reason to do so.

If you want to hear something a little different, I have gone the other direction. I have soldered the IHS to the heatsink under high pressure. I have had this this for a long time with no degrade in performance on my classic PC. It shows a 1C improvement over Liquid Metal Compound. You can see the result in the Motherboard Capacitor Changeover Thread.

 

[Enterprise24]

At first I believe der8auer. But I doubt when I saw what happen to 7980XE (428mm^2).

 

[qubit]

At first I believe der8auer. But I doubt when I saw what happen to 7980XE (428mm^2).

What happened to it?

 

[rtwjunkie]

I voted other, be I want soldering back (what with soldering not being an actual thermal paste).

 

[newtekie1]

Going with the best thermal compound possible, a liquid metal variety, would definitely be too expensive to manufacturer. That stuff is just not easy to work with.

On the other hand, I get why they don't solder, at least on the mainstream products. I remember them having issues with the solder forming voids all the way back in the Pentium D days. Extremists were delidding all the way back then because of the voids that would form in the solder.

The fact is that the TIM they use now is good enough so that 90% of their consumers will never even notice the difference. We are enthusiasts, we overclocked, we check temperatures obsessively for some reason, so we notice the higher temperatures. But Intel doesn't care about enthusiasts. Enthusiasts don't make Intel rich.

That said, at the end of the day I would like to see them use a better thermal compound at least. It doesn't have to be a liquid metal, but the paste they use now is really sub-par. Delidding and replacing with even MX-4 shows a good improvements in temps.

 

[rtwjunkie]

Agreed, even using MX-4 would be an improvement for them, and not cost that much in the grand scheme.

 

[mouacyk]

The answer is YES AND the gap needs to be eliminated.

 

[newtekie1]

The answer is YES AND the gap needs to be eliminated.

I agree. In fact, I wouldn't be surprised if eliminating the large gap between the heatspreader and die isn't the real reason behind the temp decrease when delidding and replacing with normal thermal paste. The paste Intel uses could actually be good, but when it is so thick because of the gap between the die and IHS, it just looks bad.

I thought I read somewhere that when they did the Devil's Canyon refresh of Haswell they switched to using Shin-Etsu thermal paste under the IHS, which is actually some pretty good stuff. And I kind of believe that is what they are using, because the last time I had a processor replaced by them, they sent me a tupe of Shin-Etsu to use with the cooler, so it makes sense to me that they are also using that under the IHS too.

 

[IceShroom]

Intel need to use solder the cpus specially these coat more than $200. Deliding is not a feature, it is a burden.

One thing all people forgetting is "Thermal Conductivity".
In heat transfer the thermal conductivity of a subtance is an intensive poperty that indicates its ability to conduct heat. It is measured in watts per meter-kelvin(W*m-1*k-1). A subtance with higher thermal contuctivity will transfer more heat that a subtance with lower thermal conductivity.
That is why at same size, a Copper radiator performes better than a Aluminium radiator (cause Copper’s thermal conductivity is 401 W*m-1*K-1 and Aluminium’s thermal conductivity is 237 W*m-1*K-1).
Also intel may have good engineers that doesn't mean other company don't have good engineers.

 

[natr0n]

With solder it's basically like direct die cooling/contact.

That's how I see it.

 

[EarthDog]

OFN....how many times does this need to get rehashed??

As an enthusiast, id like to see solder again. On the other hand the current Goop works fine at stock and allows overclocking. Sure, many are temp limited, but what does delidding and voiding your warranty get you? A whopping 100-200 mhz in most cases if you arent voltage limited.

I ran a 7900x at 4.5 ghz all cores/threads, not delidded. I currently run a 7960x at 4.5 ghz 16c/16t, not delidded (i can run all 32t at this speed w/o issue, just dont need HT threads). What will i gain from 4.7 ghz??? In a 8700k, what will i gain to 5.2ghz? People make waaaaay too much out of this. Ironically, people seem to have zero issues with amd chips using solder overclocking 400mhz from base. Without solder my chips are 1.2ghz and 1.7ghz over base...whats the issue? Just something to complain about?

Again, you are paying a premium for the unlocked multi and higher base clockspeed in the x...

 

[Ebo]

I think the one they are using now is good enough for normal use. If you want high OC or bench extreme, just delid, replace with your own favorite tim and loos your warranty....its simply perfect .

 

[EarthDog]

I think the one they are using now is good enough for normal use. If you want high OC or bench extreme, just delid, replace with your own favorite tim and loos your warranty....its simply perfect .

You technically arent warranted anyway when overclocking unless you buy the extra plan....

 

[cucker tarlson]

ey solder the heatspredder on some chips
and yet they will refuse a warranty if the CPU is over heated ????
Surely when soldering the heatspredder on ...>> the temp needed will be far inexcess of anything any average user will ever reach.
Does Intel have Some magic way of Soldering

I have not Dellided since core2 s

I've got 5775c running at 1.416v in IBT at max preset, it barely touches 70 degrees. Thermal paste really ain't so bad if it's decent thermal paste.

 

[phanbuey]

I do believe usually the one they're applying now's sufficient for typical use. If you would like large OC or counter excessive, only delid, change with your own personal beloved tim and loos your warranty....its merely ideal

i mean... 'sufficient' is tough to say when a stock 7900x can overheat and throttle running avx 5 on a tower air cooler.

http://www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092-11.html


That's at stock... running instructions that are within design specifications (Sisoftware Sandra's multimedia benchmark is even worse);

if you delid and replace with a liquid metal tim you will drop 20C and all of the sudden, normal coolers are capable again... (at stock). I get the argument that "it's fine if you don't OC" but this is really not the case with their $1000! chip... That's really a design/manufacturing flaw now matter how you dice it.

 

[EarthDog]

i mean... 'sufficient' is tough to say when a stock 7900x can overheat and throttle running avx 5 on a tower air cooler.

its a 140w chip....a cheap air cooler wont do it... i believe it. Welcome to HEDT requirements.

Ive reviewed 15 different x299 motherboards using two different 7900x with an h110i cooler able to overclock each cpu to all cores 4.5 ghz before it crapped out temperature wise. Our test uses occt which also uses avx instructions.

 

[Space Lynx]

I'm putting liquid metal on my 8700k next week, my only worry is will I need to re-paste it every 6-12 months, or will it last me a good 2-3 years... hmm i am considering just using Noctua thermal paste for the IHS and the outside... i know it won't be as good, but it should still keep temps down better than stock paste, plus i won't have to repaste it for a solid 3-4 years.

 

[phanbuey]

its a 140w chip....a cheap air cooler wont do it... i believe it. Welcome to HEDT requirements.

I mean I get that... but when you go from a cooler being "not viable" to "viable" with a 15C TIM paste change, on a $1k chip... it's hard not to think "yeah... they should have just soldered this".

I'm putting liquid metal on my 8700k next week, my only worry is will I need to re-paste it every 6-12 months, or will it last me a good 2-3 years... hmm i am considering just using Noctua thermal paste for the IHS and the outside... i know it won't be as good, but it should still keep temps down better than stock paste, plus i won't have to repaste it for a solid 3-4 years.

It should last quite a while... silicon lottery has been selling liquid metal delidded chips that are much older than 2-3 years and they are still going strong.

 

[EarthDog]

See my edit...

Not sure what the deal with Toms sample was, but my two for review are good under a corsair 2x120 aio.

 

[Space Lynx]

See my edit...

Not sure what the deal with Toms sample was, but my two for review are good under a corsair 2x120 aio.

I have my 8700k on a Noctua NH-D14 with 3x 140mm fans in push pull directly attached to the heatsink. --- its quite the beast. i just can't wrap my head around risking water leaking on suhc expensive equipment, i know it most likely will never happen, but i just want worry free.

 

[Tomgang]

In the overclockable CPU's deffently yes and infact they shut be soldered. In non oc CPU's it dosent matter as much.

Its to cheap ass in my oppinian that not even the 2000 USD I9 7980EX is soldered, but used the same cheap ass crap tim as the cheapest CPU´s. And its not like intel cant do it cause my old I7 980X as far i know is soldered.