Has anyone removed the Intel IHS and soldered it directly

>>58787793
AMD doesn't have this problem.

>>58787793
>soldered it directly to a heatsink
For what purpose?

>>58787793
go ahead and do it and show us how it's done.

>>58787927
Metal solder is a very good thermal conductor. It's probably not that much better than Arctic Silver though

>>58787956
What about the thermal compound under the the heat spreader?

Also you cant solder flat pieces of metal like that, it would need to be lapped, and brazed to the heatsink.

>>58787927
Intel has issues with the paste they use drying out, or even the sink being coked up on a corner causing temps to uneven. Soldering the sink on removes the issue and improves heat transfer.
Personally I'd just go die to cooler. But I think soldering would give to resaleablity. No one buys delidded chips.

No, because I'm not an autist who cares about squeezing another .05Ghz out of my overclock.

>>58787986
AMD does it... Learn your shit first faggot.
They don't need to be lapped either. And I don't think you know what brazing is.

>>58787793
>delidding
enjoy your voided warranty

>>58787988
This was an issue on like 3 CPUs, and replacing that compound only dropped overall temps by 3C. Not worth the process involved in removing the heat spreader.

>>58788032
AMD does what?

>>58787831
Lol what? AMD has heat spreaders knobhead

>>58788057
Solders the >>58788065 """"Heat spreader""" To the die.
>>58788065
IHS and Heat spreaders are the same thing dumbfuck.

>>58787793
What's the point when you can remove the IHS and connect the heatsink directly to the CPU.

>>58788090
AMD solders heat spreaders to the die?

Proofs please

>>58788057
Oh. And from what I've seen when I was looking into delidding a while ago. they still have issues with the silicone they use to glue down the plate shifting and causing a corner or side to cock up slightly and create a +6-10c temp difference between the coolest and hottest core.
This doesn't affect a large portion of chips but it has been evident to happen. Thus warranting delidding to square up the spreader. Also gaining a nifty few degrees cooler depending on what paste you use to replace the stock.
I hear that gallium blend thermal metal liquid shit is tops. Getting 6c cooler. But I haven't tried it.

>>58788090
Show me where it is like that on amd cpu's

>>58788118
The fuck? Don't you know how to use google.
The APU's are TIM. The FX series is soldered.

File: solderedihsamd.jpg (101KB, 528x960px) Image search: [Google]

101KB, 528x960px

>>58788201
Cool.

>>58787988

People that do this are like people who have to get those stones to make the audio quality better.

It's just a few degrees, jesus christ people.

>>58788212
>duplicate file.
See you found that too. Guess you know Zen will be soldered as well.
Still don't know what Intel dropped it for TIM. Do they really need to save 0.5 cents per chip?

>>58788229
When your chip is cocked up on one side causing a 10c temp difference, you lose a lot of overclocking overhead that makes it as if it's unstable. It's not autism, it's science.

>>58787793
Yes.
http://overclocking.guide/the-truth-about-cpu-soldering/

>>58788231
>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 (Figure 12).

>>58788231
>save 0.5 cents per chip
If you multiply that number by the amount of procs they sell I bet you it's gonna be higher than AMDs Net income.

There is a reason global giants cut corners in weird places.
Dell saved millions by going from 4 to 2 screws to hold their PSUs in place.

>>58788255
So intel came up with a plausible sounding excuse instead of using a different less rigid solder. As a boilermaker It's small surfaces that resist cracking. Not large, because the tension of the solder stacks over the distance, not magically gets stronger with less material.
(see peening)

>>58788241

R M A
M
A

File: 1466772925832.png (541KB, 825x793px) Image search: [Google]

541KB, 825x793px

>>58788265
And that's why I hate capitalism. Cutting corners on a already profitable product with no gain or even to the detriment of the end user.
Fucking Jews.

>>58787793
>>58787956
It is impossible to solder a silicon die to copper at home, because you need to apply different rare metals like indium and gold in nanometer scales, but easily achievable at great manufacturing scale, what we can see as an example of pretty cheap AMD and other processors with soldered ihs.

>>58788314
The larger the company, the smaller the profit margin is due to increase in labor, and overall complexity of the company.

Unless your apple and just gouge the prices as you see fit.

>>58788333
It's not impossible, but the question is whether or not it's any more effective than Arctic Silver if you don't use the industrial process.

>>58788336
Laws in my country allow you to drive a car with missing wheelnuts providing that it's only one per wheel and no less than three remain. Doesn't mean cars should be sold like that. If you can't afford to operate then close shop.

File: paper.png (302KB, 918x1212px) Image search: [Google]

302KB, 918x1212px

>>58788292
Seems like it's a legitimate excuse to me. Here's the study if you are interested.

>The Material Optimization and Reliability Characterization of an Indium-Solder Thermal Interface Material for CPU Packaging
>http://iweb.tms.org/PbF/JOM-0606-67.pdf
>The center degradation of solder TIM units can be divided into a crack initiation phase and a crack propagation phase. Figure 7 shows typical small-die CSAM images taken at different intervals of unit testing. At low thermal cycles the initiation of small micro-cracks can be seen, but does not impact thermal performance. At higher thermal cycles the cracks have propagated across the entire center region of the die and significantly degrade thermal performance.This failure mechanism is predominant on small die form factors.

The center degradation mode is not observed, as the die size is increased to the large and medium sizes.

>>58788369
There is a difference when it comes to omitting parts, vs just making a part that uses less of something.

>>58788454
Thanks.
From what I can gather the problems arise from the mixing of metals over time. Only affecting the TIM and not the die itself.
Figure 11 and 12 do allude to another factor being heat distortion putting stress on the TIM at smaller die sizes. But I couldn't find any mention of this. I will theorize, that. By having smaller die sizes this allows the base of the chip to warp to a greater degree putting much larger strains on the edges of the TIM. Beginning the cracking. The tests that show the mixing on metals was conducted in thousand hour cycles at 100c. A bit on the extreme end, but probably still a valid working life test.
I can't find anywhere that they tried to alleviate these issues. The document seems just to try to find the optimal thickness for the gold layer.
I think that these issues can and should have been corrected.
One fix being a thicker substrate.

I wonder is AMD had these issues and how or why they got around it. Anyone's chips getting hotter with age?

File: o4Bt3j9.jpg (32KB, 608x342px) Image search: [Google]

32KB, 608x342px

I do want to try this on my PS3.
But I will fuck up anyway.

>>58788776
True. But that doesn't stop company's from doing both is there's profit in doing so. Basically fuck the customer, we've got this slice of the market. Now let's rape it.

>>58787793

That's not exactly how deliding goes. I know you're trying to be edgy, but you're doing it wrong.

>>58787793
Why though?

You mean solder the die to the heatsink? I dont think thats possible in practice nor is soldering the ihs to a heatsink, youre fucked

>>58788032
yes but nobody actually gives a shit about AMD cpus.

>>58788971
Not supporting Jewish tricks is a pretty good reason to care about AMD.
Wanting as much computing power for your dollar is a good reason to care about AMD.
Not wanting your graphics card to become gimped with driver updates is a good reason to care about AMD.
Not wanting to go win 10 is a good reason to care about AMD.
Not getting forced spyware in drivers is a reason to care about AMD.

>>58788815
Just spread that thermal past equaly over those dies and you will be fine

File: 1465798745892.jpg (86KB, 500x375px) Image search: [Google]

86KB, 500x375px

>>58788815
You can't really fuck that one on the right up. There's no parts that a bit of overflowed paste would fuck with. The left however would need a more cautious approach.
But your image gives the right amount of paste. So monkey see monkey do and you'll be right.

>>58788310
functioning as intended

>>58789933
Depends on the window of time to RMA. Or even knowledge level/want to overclock before you notice such a thing is wrong.

>>58787831
You're right. Especially since they literally don't exist in CPU market.

>inb4 muh Ryzen

Not an argument waitfag.

>>58790124
>t. Buttmad TIM chip owner

Xeons are still soldered for what its worth.

>>58788042
>Implying CPUs actually die

File: 1464325902780.png (393KB, 611x607px) Image search: [Google]

393KB, 611x607px

>>58790548
>take off heat spreader
>welcome to die.

>>58790124
t. goldbergstein

File: tahna.webm (392KB, 540x540px) Image search: [Google]

392KB, 540x540px

>>58789547
This is how you apply paste.

>>58790557
If I pull that that IHS off, will you die?

>>58787793
I was entirely prepared to do this with a 4770k because of it's shit internal paste but then intel released the 4790k which is essentially the same chip with better cooling and maybe some VT-d thing added that I can't remember at about the same price.

>>58792348
>that webm
what the hell is that from haha

This is a very bad idea. You need access to industrial soldering machines to even get uniform precision below the 20µm-1mm range to apply various layers, several chemicals that - at least in the EU - are hard to get a hold of, rare (and expensive) metals in machine compliant shapes and sizes and a cleanroom. On top you need to have tools for precise measurements to see if your solder went through alright and if not there's pretty much no way back and you ruined the CPU. Make a tiny error and you get fractures or air eclosures and can throw the CPU away as well.

I soldered about fourty GPUs during slow days at work when the early Bitcoin craze began but I can tell you it's not worth it. The research effort was significant and the time and financial investment so high that you are better off just sticking with what you got out of the box. If you already had to ask it's likely you don't have the tools, connections or knowledge to pull it off.

Could see this work as a service and I'm sure with the right connections in hardware cities like Shenzhen or tech based universities that allow you to use their shit (means none of them, officially) you could get a supply chain going but given that there's no warranty on any such modifcation and you have to deal with customs fees I don't suppose demand is going to be high enough for anyone to bother with it, particularly as the equipment you need is too expensive to just buy in order to dabble in CPU soldering. There are professional soldering services available but the price is too high, it's around 8k€ for a custom one-time job of this nature.

What you could do is remove the heatspreader and use (non-conducting, means non-metal) thermal compound to directly attach the die to the heatsink but you need to be vary of heat expansion related wear and probably end up with air enclosures that eventually cause damage. In that case your best bet is probably just end up smearing the same shit Intel uses onto your die. qed

>>58792348
this and only this

File: 1483950203215.gif (468KB, 500x282px) Image search: [Google]

468KB, 500x282px

>>58790557
You must've been born in 00s or were a console fag all this time if you think that.

You'd be better off milling .005" off the top of it, since there's copper underneath. Perfectly flat, all-copper finish sitting flush against a heat sink? Awesome.

At least that used to be true.

File: 1291773949331.jpg (82KB, 610x460px) Image search: [Google]

82KB, 610x460px

>>58793112
Not sure. But 1990, and I was referring to the die under the heatsink. Think it went over your head.

File: 1467679553180.jpg (156KB, 738x614px) Image search: [Google]

156KB, 738x614px

>>58793043
(You) and only (You)

File: understood.gif (2MB, 800x450px) Image search: [Google]

2MB, 800x450px

>>58793175
I understood you.

>>58793112
https://www.youtube.com/watch?v=IdAmkx8eAos

>>58788265
>Dell saved millions by going from 4 to 2 screws to hold their PSUs in place.
I'd rather they increased the price by a few cents.

>>58793396
u mad brah?

>>58788314
>blames jews
>the real culprit is capitalism
>is too brainwashed to see

File: 1485747509058.jpg (63KB, 453x588px) Image search: [Google]

63KB, 453x588px

>>58793396
>can try to not be such a gigantic fucking faggot holy shit maybe one day anon-kun
hahaha what

File: a.png (494KB, 738x614px) Image search: [Google]

494KB, 738x614px

>>58795902
maybe one day anon-kun can try to not be such a gigantic fucking faggot, holy shit.

>>58792348

>Hellmann's mayo

Kek

>>58787793
But why? Do you really need to overclock it? Because that's the single reason you'd need to delid. And even if you're overclocking, most people do just fine without deliding that.


So just... why?

File: EdyCaDB.jpg (2KB, 400x480px) Image search: [Google]

2KB, 400x480px

>>58795850
I wonder who could be behind this post...

>>58795902
Uncultured faggot swine