Help with CPU overclocking on z77 OC formula

[parsec]

Re: Help with CPU overclocking on z77 OC formula

I'm currently reading page 9 of the Hard|Forum link at the bottom of my previous post.
They're debating numerous thermal interface options and it's getting stale.
It appears that the biggest hurdle is trying to mount a water block to the naked die without destroying the pcb.
There is a promising link where there might be some good info at the end of the thread but I can't connect to it:
3770k IHS Removals - CPU temp dropped from 79C to 71C
I'm hoping that page 11 is the end of the thread, I'm getting bored.

I wonder how the PCB without the IHS even survives the the CPU socket clamping mechanism. I've seen people with certain make 1155 CPU sockets getting dents on the edge of the IHS from the contact points of the clamping mechanism. Removing the IHS also removes a small but possibly significant amount of thickness where the clamp contacts the CPU, reducing the pressure between the CPU and socket contacts.

I've read that naked die CPU users will place a shim or spacer around the die, providing some extra contact area for the heatsink contact surface. That also helps to prevent uneven pressure from the heatsink crunching the corners and edges of the die (cringe!)

It's bad enough that we can't see the result of the TIM "fingerprint" on the IHS and heatsink under normal circumstances, but using a naked die and hoping we have even pressure on it... I know to much to be able to handle that psychologically.

[synack]

Re: Help with CPU overclocking on z77 OC formula

I wonder how the PCB without the IHS even survives the the CPU socket clamping mechanism. I've seen people with certain make 1155 CPU sockets getting dents on the edge of the IHS from the contact points of the clamping mechanism. Removing the IHS also removes a small but possibly significant amount of thickness where the clamp contacts the CPU, reducing the pressure between the CPU and socket contacts.

I've read that naked die CPU users will place a shim or spacer around the die, providing some extra contact area for the heatsink contact surface. That also helps to prevent uneven pressure from the heatsink crunching the corners and edges of the die (cringe!)

It's bad enough that we can't see the result of the TIM "fingerprint" on the IHS and heatsink under normal circumstances, but using a naked die and hoping we have even pressure on it... I know to much to be able to handle that psychologically.

With naked die, the retaining clip with the clamping mechanism has to be removed. The die does not protrude past the retaining clip. So unless your heatsink has taken this into consideration the clamp has got to go.
It is pretty close, but if one where to add a shim in there to prevent PCB damage, that would be enough for the heat sink to not touch the die at all.

Of course this is not an exact measurement. Just eyeballing it.

Syn

[profJim]

Re: Help with CPU overclocking on z77 OC formula

3770k IHS Removals - CPU temp dropped from 79C to 71C is a definite MUST READ. I'm currently on page 17 (out of 20) and I'm not bored!!

I completely agree with parsec's concerns with running with the IHS removed.
Common sense tells me that one important function of the IHS is to ensure that ALL of the pads on the bottom of the cpu's pcb are fully seated in the socket.

Check out the posts on page 13 and page 14 of the thread, especially the information in post numbers 302, 304, 333 and 338.
It looks like the IHS is purposely designed to have some vertical "give" to distribute the downward clamping force evenly across the bottom of the cpu.

[parsec]

Re: Help with CPU overclocking on z77 OC formula

All done! This is my post!

Where did that come from? Who be you?

But, good for you, finally done! About time! Nice post, that was a huge help, I'm book marking it for sure!

[parsec]

Re: Help with CPU overclocking on z77 OC formula

I forgot that the clamping mechanism was removed... duh, I thought some of those pics looked odd. Actually, don't they lose the socket back plate too?

Prof. Jim, thanks but we are simply not mad men! I haven't even scratched the surface of all the dangers this dissection... uhm, modification ensues.

No offense synack, more power to you, as they used to say. But OMG, think of all the tolerances and clearances that are out the window when the clamping mechanism and IHS are gone. Then mounting a relatively large CPU water block on a bare die, with one shim being the only protection. Well, the water block better have mounting standoffs, but still... but still!!

I just know what I would do, "... Ok, one more quarter turn of that screw... slowly... crack! NOOOOOOOO!!!"

That is a great thread you posted, thanks! The need for "give" in the IHS makes perfect sense, but when I think about it, does that only apply to IB CPUs? If other CPUs have solder between the die and IHS, and any gap between them is filled in (I assume), what give could there be?

[profJim]

Re: Help with CPU overclocking on z77 OC formula

Looking at the de-lidded IHS and cpu core, it looks like the crappy Intel TIM is filling a (relatively) deep area, thick enough to use a thin thermal pad layer instead of TIM. Due to manufacturing tolerances, most of all of the high end Seasonic power supplies have have a thick thermal pad under the pcb to help cool severval hot components on the underside of the pcb, where the psu case is the heatsink for those parts. Why did they do this with Ivy Bridge? Maybe it has something to do with the ivy bridge cpu tri gate technology. I think that it's time to do some research and see what the difference is in die sizes for Sandy Bridge, Ivy Bridge, and Sandy Bridge-E. Even if IB generates less heat overall, it's in a smaller, more compact package and this smaller package might be the root of the problem.

just my 22nm.....err cents.....

[synack]

Re: Help with CPU overclocking on z77 OC formula

I forgot that the clamping mechanism was removed... duh, I thought some of those pics looked odd. Actually, don't they lose the socket back plate too?

Prof. Jim, thanks but we are simply not mad men! I haven't even scratched the surface of all the dangers this dissection... uhm, modification ensues.

No offense synack, more power to you, as they used to say. But OMG, think of all the tolerances and clearances that are out the window when the clamping mechanism and IHS are gone. Then mounting a relatively large CPU water block on a bare die, with one shim being the only protection. Well, the water block better have mounting standoffs, but still... but still!!

I just know what I would do, "... Ok, one more quarter turn of that screw... slowly... crack! NOOOOOOOO!!!"

That is a great thread you posted, thanks! The need for "give" in the IHS makes perfect sense, but when I think about it, does that only apply to IB CPUs? If other CPUs have solder between the die and IHS, and any gap between them is filled in (I assume), what give could there be?

Which is why I put the IHS back on lol.

Everything is still going strong. No hiccups, temps are steady (if not down one degree or two once the TIM settled in)
I found and compiled up mprime for Linux. Wow now THAT is a torture test. lol. I'm still keeping at <80c on it running at 4700. I got up to 4800 fine, but it would get the very occasional glitch during mprime test runs.
I'm happy with 4700 and the crazy cool temps. I can now stay around 28-30c doing my normal day to day stuff. To me that's incredible.
I'm able to set my fans and water pump to their lowest settings and still run fine.
However I do need to crank them up a bit when I'm gaming.

I promise to get photo's up soon. Work and playing with my new toy is getting in the way hehe.

Syn

[parsec]

Re: Help with CPU overclocking on z77 OC formula

Prof. Jim, I had the same thought about the IB tri-gate design, I wonder if that die cannot withstand the flood of hot solder used in other CPUs. Frankly, I don't see how any CPU survives that treatment. Plus add no-lead solder whose melting point is over 400F... I can understand why TIM is used.

Yes, the IB die is smaller than SB die, given similar processors, for four core CPUs, IB is smaller by 56 square mm. Check the list of die sizes here, scroll down the page:

AnandTech - Dual Core/GT2 Ivy Bridge Die Measured: ~121mm^2

I agree with your 22nm of silicon (HA!), the heat is more concentrated in a smaller area. Apparently those heat transfer pads are not as efficient as standard TIM. They also tend to use them more with macro-scale devices, like the voltage regulator transistors in power supplies, on NAND chips and SSD controllers, and in the large power transistors used in audio amplifiers.

BTW, I love my Seasonic PS, built as you mentioned. The fan never runs, and I mean never runs, it may be broken for all I know! The PS case is a little bit warm, but hardly worth mentioning. I even put a tiny strip of paper over the vent outlet to see if it would flutter if the fan came on. It has never once moved.

[synack]

Re: Help with CPU overclocking on z77 OC formula

No doubt, I was a bit surprised when I first heard they used solder to join the IHS.

I get that they had to come up with a solution...
I still think it's wrong of them to use crappy TIM when clearly the chip is heat sensitive.
I haven't seen crappy TIM like that since the 90's lol.

[parsec]

Re: Help with CPU overclocking on z77 OC formula

Yes synack, it is an enigma.

Just to ask, why do you say it is poor TIM? Believe it or not, the TIM used on the (shudder) stock Intel CPU coolers is some of the best there is. They used the best Shin Etsu (my personal choice) or a Dow Corning product with the same specs. So now Intel changed to something else for use in IB CPUs?

Another (scary) thought just came to mind. The TIM in a IB CPU must last the lifetime of that device. Was the seal around the IHS that you removed a continuous seal, as in air tight? Although sloppy, it looks like it might be, or does the space under the IHS have a vent? Regardless, consider the longevity of an IB CPU using TIM. Will it be fine for three years of usage, or five? I wonder if the Xeon versions of IB CPUs are assembled in the same way. Are other CPUs built this way?

I wonder if we'll be seeing issues with "old" IB CPUs, overheating due to deterioration of the TIM? How does Intel know it will last for many years? I have a PC with a Pentium D 945 CPU that was six years old this year, and still works fine, along with the original ECS board. Will I be saying that about my i5-3570K CPU in the future?

OMGosh, I looked up my old '945, which is no longer supported by Intel, or is EOIS, or End Of Interactive Support. There was a note on the page that if you owned a Dell or Gateway PC, if you opened the case, you "may have" voided the warranty.

[synack]

Re: Help with CPU overclocking on z77 OC formula

Yes synack, it is an enigma.

Just to ask, why do you say it is poor TIM? Believe it or not, the TIM used on the (shudder) stock Intel CPU coolers is some of the best there is. They used the best Shin Etsu (my personal choice) or a Dow Corning product with the same specs. So now Intel changed to something else for use in IB CPUs?

Another (scary) thought just came to mind. The TIM in a IB CPU must last the lifetime of that device. Was the seal around the IHS that you removed a continuous seal, as in air tight? Although sloppy, it looks like it might be, or does the space under the IHS have a vent? Regardless, consider the longevity of an IB CPU using TIM. Will it be fine for three years of usage, or five? I wonder if the Xeon versions of IB CPUs are assembled in the same way. Are other CPUs built this way?

I wonder if we'll be seeing issues with "old" IB CPUs, overheating due to deterioration of the TIM? How does Intel know it will last for many years? I have a PC with a Pentium D 945 CPU that was six years old this year, and still works fine, along with the original ECS board. Will I be saying that about my i5-3570K CPU in the future?

OMGosh, I looked up my old '945, which is no longer supported by Intel, or is EOIS, or End Of Interactive Support. There was a note on the page that if you owned a Dell or Gateway PC, if you opened the case, you "may have" voided the warranty.

The TIM is so dry it's like plaster of paris. I didn't wipe, I had to scrape the TIM off. It literally came off in chunks. The TIM looked "cracked" on removal, unfortunately like Schrodinger's cat there's no way for me to tell if the TIM was like that with the IHS on or if the act of removing it caused the cracking. In my books TIM cracking and flaking off is baaaaad.

Also the results speak for themselves!
before TIM replacement: temps were higher and varied wildly across the cores. 6C+
after replacement: temps dropped to normal and are now within 1C across the cores.
I used the Gelid OC extreme TIM that came with my mobo. it's considered "middle of the road" of the top TIMs. (it's not the best of the best)

The black silicone they used to "seal" the IHS was not all the way around for me, and mine looked just like a few other pics out there. the "bottom" where the serial number is, is open.
Due to the extreme heating and cooling, I expected a gap as we would not want the air inside to build up pressure.

LOLz at the case thing. I worked in the service dept for a discount PC manufacturer a long time ago. They used to pull crap like that to hide substandard hardware lol. Not suggesting Dell or Gateway did... but it does make you wonder.

Syn

[parsec]

Re: Help with CPU overclocking on z77 OC formula

Ahh yes, Schrodinger's TIM, you don't know if it's working or not unless you remove the IHS... wait, yes we do, it's not working. That's a good one synack!

I've also thought that TIM with the consistency of cookie dough was a bad thing. Then I noticed so many good TIMs are like that. I read an article about TIM, that said the problem with thin, wet or oily TIM, is that it gets squeezed out of the IHS - CPU cooler interface gap, officially called Pump Out. That is not just on the initial mounting of the CPU cooler, but over time, weeks and months later. Also, even oil and other semi-solvents evaporate over time, leaving TIM dust, officially called Bake Out, instead of what looks like old pizza cheese. So the cracked, clumpy, pizza cheese is actually better...

I wonder if the higher temps the TIM on the CPU die is subjected to causes it to bake out faster. As you said, the IB IHS is not air tight, for good reason, which again contributes to bake out. Hmm, now what unusual TIM does not suffer from this (solder!!) I officially have a case of PC nerd nervosa over IB die TIM.

Regarding the "do not open" PCs, I'd be afraid to look inside, to find all their weird, proprietary connectors, etc. Isn't it Dell that uses a non-standard fan header on their boards and fans? Sure, make me buy your goofy 80mm fans for $20, that you pay 50 cents for... "Dude, you've got a Hell PC!"

[synack]

Re: Help with CPU overclocking on z77 OC formula

I agree, Thick paste is a good thing.

Pump out isn't always a bad thing however. really, all you want is enough to cover the tiny spaces between the two materials.
In a perfect TIM world, all excess would be squeezed out, and only left with filling the gaps that would otherwise be occupied by air.
This is how I perceive when TIM is "settling in." (cycles of hot and cold to induce movement which causes the gaps to be better filled, with the air and extra TIM squeezed out)

I just don't know how else to reason the temp issue. my CPU temps were wonky from the day I installed this CPU until i replaced the TIM. The OEM TIM shouldn't have had a chance to dry out. Even still this PC is only a few weeks old. Should TIM be hard after 2.5 weeks?
I know the stuff I put between the water block and IHS was still awesome (probably not even broken in yet lol) and it had about the same "running" life as the TIM under the IHS.

LOL at the "Hell PC"
I have to say though, in a corporate world, Dell PC's are aweeeesomer. Gold tech support from Dell is magical. Call up the "special phone number" give them the PC specifics, tell them what's wrong, 1-2 day parts shipping with a return paid box for the defective part, if they even want it back. (sometimes they don't if they cant salvage it and/or it's not worth the shipping cost)

I should rename the subject to: "Schrodinger's TIM Approximation"

[parsec]

Re: Help with CPU overclocking on z77 OC formula

As a contrast, in my other hobby, audiophile-mania, the way that large output transistors, the main ones in an amplifier, are interfaced to their heatsink is a little different. In the old days, they used mica "washers", which just look like thin clear plastic, which is needed to insulate the metal surface of the transistor from the heatsink, with TIM on both surfaces of the washers. I've mounted large power transistors like that, and the TIM is very thin and oily, at one time called silicone grease. Nothing like CPU TIM, literally a very light grease consistency. Now they use "sil-pads", a silicone impregnated, soft material cut to shape, like a thin piece of cloth, with no TIM, the pad is the TIM. You can buy sil-pads in sheets even, to custom cut, I often wonder how good they'd be on CPUs.

IMO, the theory of what TIM is for, filling in very small imperfections in the two surfaces in contact with each other, and the reality, are very different. In theory, I think the layer of TIM we find between the IHS and heatsink, should be as thin as a piece of paper, or less, much less. In reality, it is much more than that. Would you agree?

IMO, the reality is TIM is filling in relatively large gaps between the IHS and heatsink, mainly due to the non-flat nature of the IHS. Yes, we can lap the IHS (and void the warranty), but I will ignore that for this discussion. Next is the potential non-flat nature of the heat sink contact surface, usually not as bad as the IHS, but variable. Frankly, I'd say the (forgive me) stock Intel CPU cooler has a really flat surface, about the only good thing about it.

Given all the potential variations and tolerances that can exist (and are worse with push-pin mounting), plus application error, the result is a TIM thickness (variable across the two surfaces) I would estimate to be 0.1 - 0.2mm thick. Ideally, wouldn't the TIM be nothing more than a hazy coating on the IHS, that allows you to read the text on the CPU? No more than that on the heatsink contact surface too?

Of course we try to minimize the amount of TIM, using the "pea size" application method. We then worry it was to little. Or we check and see bare spots on the IHS or heatsink. Bare spots are likely caused most of the time by the non-flat surfaces, and other variations (pressure, or just poor mounting technique.)

Sorry, just my TIM reality-opinion-rant

I have no doubt you get great service from Dell in the enterprise environment, and in reality in most cases it is a good thing that opening the PC case is forbidden, due to the infamous, ubiquitous ID10T error factor.

[profJim]

Re: Help with CPU overclocking on z77 OC formula

From User error - Wikipedia

User Error and related phrases such as PEBKAC ("Problem Exists Between Keyboard And Chair" and the complementary PEBCAK ("Problem Exists Between Chair and Keyboard")), PICNIC ("Problem In Chair Not In Computer"), PIBCAK ("Problem Is Between Chair And Keyboard") or ID-10T error ("Idiot error") are also used as slang in technical circles with similar potentially derogatory meaning. This usage implies a lack of computer savvy, asserting that problems arising when using a device are the fault of the user. Critics of the term argue that the problems are caused instead by a device designed in such a way that it induces errors.

The term can also be used for non-computer-related mistakes.

parsec, is this a bad time of the month for you?
It's OK, we don't mind occasional rants.
Funny but unrelated quote: "It's hard to soar with eagles when you work with turkeys."