The Truth About CPU Soldering

[hat]

I voted yes because even using something like MX-4 brings good improvement over their thermal gunk. However, mass production thermal paste applications are rarely good. I know someone who had a jet engine PS4 until he took it apart and re-pasted it... and we all know better paste can bring improvement of a few degrees (unless you use something extreme like coollab liquid pro/ultra), but the biggest improvements are usually credited to the better application you get doing it yourself than done in a factory that does thousands/day.

 

[las]

You get one attempt, and then you may not be able to remove the IHS ever again.

If AMD can manage, Intel should too. Ryzen temps are stable across cores and much lower during load than Intel's paste crap.

Intel has temp spiking and huge variance across cores. Last week I build a 8700K rig, during full load, Core 2 was 13C hotter than the rest. Haha

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

No they are not, many people are struggling to reach 5 stable with DECENT TEMPS on normal cooling, but even with 240mm AIO's people are closing in on 90C load, insane temps if you ask me.
Delid and temp is down to 60-70....

CFL run terribly hot, just like Kaby Lake did.

 

[Dia01]

I just delided my 7820X last night. The thermals were just terrible going beyond 4.5 for my chip with minor increases in voltages. For the price anyone pays for a decent main stream upwards to the HDET platform processors I think at least a better thermal interface should be used or simply solder and be done with it. Besides, why make 'K or X' models if you were not expecting them to be overclocked?

 

[evernessince]

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.

Bad TIM under the IHS has an exponential affect on temperature. You are adding another bottleneck preventing maximum heat transfer. If the paste under the IHS can't effectively transfer all the heat to the IHS then in turn your paste on top will be less effective and so will your CPU cooler. Intel's stock paste is designed with stock settings in mind. Once you start overclocking the stock paste is very ineffective at transfering heat over the rated TDP of the processor.

There's really no need to guess at this either, there have already been plenty of people who have replaced Intel's stock paste with higher quality paste and gotten 12-18c temperature drops. I personally got a 25c drop with liquid metal and 17c drop with GC-1.

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.

Do you have anything to support this? "I believe" isn't really objective data or a professional review.

 

[rtwjunkie]

Bad TIM under the IHS has an exponential affect on temperature. You are adding another bottleneck preventing maximum heat transfer. If the paste under the IHS can't effectively transfer all the heat to the IHS then in turn your paste on top will be less effective and so will your CPU cooler. Intel's stock paste is designed with stock settings in mind. Once you start overclocking the stock paste is very ineffective at transfering heat over the rated TDP of the processor.

There's really no need to guess at this either, there have already been plenty of people who have replaced Intel's stock paste with higher quality paste and gotten 12-18c temperature drops. I personally got a 25c drop with liquid metal and 17c drop with GC-1.



Do you have anything to support this? "I believe" isn't really objective data or a professional review.

Thing is, it’s not a scientific conclusion. We have no way of knowing if the better temps at replacing the TIM are due to TIM differences (of which there really are very few differences) or in fact just better application of TIM that effectively closes the gap between chip and IHS.

Even cpu interior aside, most TIM repastes which result in better temps are just as likely to be from proper application as from “better” TIM.

 

[EarthDog]

If AMD can manage, Intel should too. Ryzen temps are stable across cores and much lower during load than Intel's paste crap.

Intel has temp spiking and huge variance across cores. Last week I build a 8700K rig, during full load, Core 2 was 13C hotter than the rest. Haha



No they are not, many people are struggling to reach 5 stable with DECENT TEMPS on normal cooling, but even with 240mm AIO's people are closing in on 90C load, insane temps if you ask me.
Delid and temp is down to 60-70....

CFL run terribly hot, just like Kaby Lake did.

AMD really has little to do with it. Temps may be lower for more reasons than just paste, you know.

Also, 90c is fine for these cpus. They do not throttle for another 10c. Ive also never seen a more than 20c drop. Thats being incredibly optimistic...10c-20c is the average.

 

[trog100]

it is how its applied more than the tim.. all tim is bad.. nothing is better than direct metal to metal contact.. tim is there to get rid of any air gaps.. if properly put together there should not be many air gaps..

intels mass production dollop of tim and dollop of glue and then pres together isnt that clever.. it often ends up with too much tim to fill too much of an air gap..

also 90C isnt insanely hot for these intel chips.. its normal for the high end ones.. intel set the throttle point at 100 C.. they set it there for reason.. they think its okay..

4.8 seems a sweet spot.. both for kabylake and coffeelake.. 5 or just over is doable but pushing it from a temp point of view..

a relid is probably good for another 200 mhz.. but i cant be arsed the difference isnt noticable..

trog

 

[las]

AMD really has little to do with it. Temps may be lower for more reasons than just paste, you know.

Also, 90c is fine for these cpus. They do not throttle for another 10c. Ive also never seen a more than 20c drop. Thats being incredibly optimistic...10c-20c is the average.

90c is fine, lol. I prefer my builds cool and quiet. Your system probably runs twice as loud as it would have been with delid and repaste.
Also, you would be able to OC 200-400 MHz more.

 

[EarthDog]

90c is fine, lol. I prefer my builds cool and quiet. Your system probably runs twice as loud as it would have been with delid and repaste.
Also, you would be able to OC 200-400 MHz more.

90C is fine according to Intel. My build is quiet. Static fan speed FTW.

Also, 200 mhz is tops in what you will gain from delidding. Again, most delids shave off 10-20c on average (mosy a lot closer to 10c). Anything more is gravy and rare.

Your ignorance on the subject is preceding you...

 

[las]

90C is fine according to Intel. My build is quiet. Static fan speed FTW.

Also, 200 mhz is tops in what you will gain from delidding. Again, most delids shave off 10-20c on average (mosy a lot closer to 10c). Anything more is gravy and rare.

Your ignorance on the subject is preceding you...

Several people are running 5.3-5.5 on OCN 24/7 on normal cooling and acceptable voltages. You are hold back by the terrible paste job and gap between your IHS and die. Your system is not quiet. A cheap AIO like Corsairs are never quiet.

 

[trog100]

i recon in the end what it comes down to is a %5 gain in clock speed (lets say 4.8 to 5) is a delid worth the hassle.. i dont think so but some may do..

my system is quiet at 4.8 or 5 with no delid.. it runs near 80 C worked hard on all cores at 4.8 gig or near 90 C at 5 gig.. i choose the lower option..

a relid would enable me to use the 5 gig option for the same temps i now get at 4.8.. is it worth it.. ??

trog

 

[HTC]

i recon in the end what it comes down to is a %5 gain in clock speed (lets say 4.8 to 5) is a delid worth the hassle.. i dont think so but some may do..

my system is quiet at 4.8 or 5 with no delid.. it runs near 80 C worked hard on all cores at 4.8 gig or near 90 C at 5 gig.. i choose the lower option..

a relid would enable me to use the 5 gig option for the same temps i now get at 4.8.. is it worth it.. ??

trog

I think it would also enable the use of lower speed fan(s) on the CPU to achieve the same temps. This would provide a quieter experience which, for some, is quite preferable over max overclock.

 

[qubit]

Nope.

You big softie. Myself I'm not a fan of delidding due to the risk of wrecking my expensive new CPU.

If I was buying a CPU now, I'd get one of those that's been delidded and re-timmed (is that a word?) by the retailer that comes with a full warranty. They're usually rated at some overclocked frequency, too. There's also a fair price premium on it, but it's worth it in my opinion.

 

[EarthDog]

Haha, lol! I don't need to delid.

Again I ask (to all), what is the difference between 5 GHz and 5.2 GHz? Or 4.7Ghz from 4.5 GHz? It isn't much. I mean with 16c/16t at 4.5 GHz, my hottest core while gaming was 59C (2 hour session). P95 is in the low 80s. This is with a simple 3x120 rad and Yate Loon fans at 800 RPM....whisper quiet!

 

[dyonoctis]

I'm just curious, why exactly desktop cooler don't use a system similar to laptop to avoid over tightening of the screws, or why intel isn't just making the new socket with a simple "delid guard" ?
A naked cpu seems to be what would make everyone happy, less cost for intel, and better temperature for enthuisast.

 

[EarthDog]

I'm just curious, why exactly desktop cooler don't use a system similar to laptop to avoid over tightening of the screws,

Some do...

or why intel isn't just making the new socket with a simple "delid guard" ?
A naked cpu seems to be what would make everyone happy, less cost for intel, and better temperature for enthuisast.

One word for you, and its the same word to descrive why intel doesn't officially support overclocking... ready for it..........?


.........you sure..........?

Liability.

 

[dorsetknob]

or why intel isn't just making the new socket with a simple "delid guard" ?
A naked cpu seems to be what would make everyone happy, less cost for intel, and better temperature for enthuisast.

to young to Remember the Multitude of threads complain/moaning
"" i Accdently cracked the Die on my CPU ( irca Athlon) when i attach the CPU Cooler""
and the smiles of those running willimite and prescot P4's at those unlucky AMD users
Die Cracking was the Reason CPU's started wearing the Hat

 

[trog100]

a few years back a good intel overclock meant about a %30 performance gain.. now near the end of the line its about %10... in some ways hardly worth the effort..

i still overclock my cpus out habit more than anything esle.. but only in the easy way.. which for me does not include a relid or a delid it really aint worth the hassle..

okay i dont do much video editing but i really cant tell the any difference between my current coffeelake chip and my (yet to sell on ebay) kabylake chip.. benchmarks tell me i have a %50 performance gain but apart from that i would not be able to tell one chip from the other.. he he

as for gaining an extra %5 with a relid.. hmm.. nuff said about that i recon..

the thing most folks need to do is to stop thinking of 90C as insanely hot.. once people do that the problem goes away..

trog

 

[dyonoctis]

Some do...

One word for you, and its the same word to descrive why intel doesn't officially support overclocking... ready for it..........?


.........you sure..........?

Liability.

I see now, from a corporation point of view it's making sense. It's better to spend some money on tankier cpu rather than dealing with panicked customer because they "broke" the cpu.
But it's a shame, those 10- 20 °c can be a big deal when going SFF.

to young to Remember the Multitude of threads complain/moaning
"" i Accdently cracked the Die on my CPU ( irca Athlon) when i attach the CPU Cooler""
and the smiles of those running willimite and prescot P4's at those unlucky AMD users
Die Cracking was the Reason CPU's started wearing the Hat

Ah yes, i wasn't lurking in tech forum in this era, nor do i had any idea what a computer looked like inside. It's looks like even the shipping of computers was a delicate matter...

 

[EarthDog]

If you are going SFF, it really isn't an ideal environment to overclock in the first place.

 

[qubit]

to young to Remember the Multitude of threads complain/moaning
"" i Accdently cracked the Die on my CPU ( irca Athlon) when i attach the CPU Cooler""
and the smiles of those running willimite and prescot P4's at those unlucky AMD users
Die Cracking was the Reason CPU's started wearing the Hat

The die on my ancient Athlon 3200+ is ever so slightly chipped on one corner, but I was lucky and the thing carried on working perfectly, including a decent overclock. I was real careful after that.

 

[RejZoR]

My AXP 2400+ was also all chipped on the edges as I was fiddling with coolers and thermal pastes a lot. It was working fine till I sold it.

 

[Vario]

Bad TIM under the IHS has an exponential affect on temperature. You are adding another bottleneck preventing maximum heat transfer. If the paste under the IHS can't effectively transfer all the heat to the IHS then in turn your paste on top will be less effective and so will your CPU cooler. Intel's stock paste is designed with stock settings in mind. Once you start overclocking the stock paste is very ineffective at transfering heat over the rated TDP of the processor.

There's really no need to guess at this either, there have already been plenty of people who have replaced Intel's stock paste with higher quality paste and gotten 12-18c temperature drops. I personally got a 25c drop with liquid metal and 17c drop with GC-1.



Do you have anything to support this? "I believe" isn't really objective data or a professional review.

I believe that my CPU's above average performance is due to a lucky minimum gap. I have the same TIM as everyone else. And my VID is nothing special. I am also using an air cooler and arctic silver 5, so nothing exotic to cool it either.

As far as the theory relating to the gap, that is a pretty popular idea. You can use google and find people claiming that the chief benefit of moving from the original paste + RTV glue to delidding and using gallium on the die and superglue to reattach the lid is that the gap is narrower so the lid is closer to the die.

If you ever repaste a laptop, you will find that you need to use a copper shim to replace the original thermal pad to close the gap between the die and the heatsink. Running a thick layer of thermal paste does not work very well, you need a shim to transfer. It is a similar idea.

The problem is the black RTV adhesive that Intel uses to adhere the lid to the circuit board. It is thick, possibly to reduce damage to the die from overclamping. The thick glue causes the lid to be further from the die, so the thermal paste is excessively thick or even gapped and does not perform well at transferring between the die and the lid.

Haha, lol! I don't need to delid.

Again I ask (to all), what is the difference between 5 GHz and 5.2 GHz? Or 4.7Ghz from 4.5 GHz? It isn't much. I mean with 16c/16t at 4.5 GHz, my hottest core while gaming was 59C (2 hour session). P95 is in the low 80s. This is with a simple 3x120 rad and Yate Loon fans at 800 RPM....whisper quiet!

There isn't any. Just imaginary E-Peen.

 

[MrGenius]

Also, 90c is fine for these cpus. They do not throttle for another 10c.

We've had this conversation before...

They don't "throttle". That's a totally inaccurate and misleading word to use to describe what really happens. Since it implies a reduction in speed. As in backing off the throttle in your car to reduce engine/vehicle speed. Nothing like that actually happens. They run, at whatever speed they've been set to run at, until Tj Max is exceeded and then immediately shutdown. The processor shutting itself off is not reducing its speed. You can't reduce speed to 0MHz. 0MHz is not a speed. 0MHz is the absence of speed.

Intel doesn't help matters by referring to it as "throttling". Further spreading the myth and misconceptions. But at least they include the part about the shutdown that actually happens instead.

Tjunction Max is the maximum temperature the cores can reach before thermal throttling activates. Thermal throttling happens when the processor exceeds the maximum temperature. The processor shuts itself off in order to prevent permanent damage.

https://www.intel.com/content/www/us/en/support/articles/000005597/processors.html

 

[trog100]

We've had this conversation before...

They don't "throttle". That's a totally inaccurate and misleading word to use to describe what really happens. Since it implies a reduction in speed. As in backing off the throttle in your car to reduce engine/vehicle speed. Nothing like that actually happens. They run until Tj Max is exceeded and immediately shutdown. The processor shutting itself off is not reducing its speed. You can't reduce speed to 0MHz. 0MHz is not a speed. 0MHz is the absence of speed.

Intel doesn't help matters by referring to it as "throttling". Further spreading the myth and misconceptions. But at least they include the part about the shutdown that actually happens instead.

https://www.intel.com/content/www/us/en/support/articles/000005597/processors.html

i dont think so.. they run (full speed) until they hit the magic 100C.. when they hit the magic 100C they start to throttle down just like a car would if the gas pedal is eased off just a tad..

they throttle back just enough to maintain that magic 100C.. they do it core by core as well.. no more no less.. they ease off the throttle just enough to maintain that max of 100 mph.. whoops i meant that max of 100 C.. mostly they dont have to throttle down that much.. normally about 2 or 3 hundred mhz does the job..

now if the heatsink falls off or some other daft thing happens and they cant throttle back enough to maintain that magic 100C they will shut down.. but it does require extreme circumstances for them to shut down..

in the real world full of poorly maintained PCs.. fully crudded up and operating in less than ideal environments there must be many an intel cpu running up near or at that magic 100C..

trog