Delidded my i7-3770K, loaded temperatures drop by 20°C at 4.7GHz

Discussion in 'CPUs and Overclocking' started by Idontcare, Aug 3, 2012.

  1. Diogenes2

    Diogenes2 Platinum Member

    Joined:
    Jul 26, 2001
    Messages:
    2,151
    Likes Received:
    0

    The package temp shows 2c higher than my hottest IB core, and at least the same as the hottest core on my 3930k, so it wouldn't seem to be that unusual.
     
  2. Yuriman

    Yuriman Diamond Member

    Joined:
    Jun 25, 2004
    Messages:
    5,091
    Likes Received:
    55
    The reason for using TIM over a metal plate (which is what the IHS is) is because no matter how well you lap and polish your mated surfaces, there will still be microscopic gaps where no heat is transferred. That's why we use liquids to fill that space. Adding another solid layer creates more places for microscopic gaps between the die and the cooler.

    @IDC, I understand your want of a shim to recreate stock conditions, but my ideal shim would be exactly the size of the gap between the IHS and the PCB, so that it does not lift the IHS off the die, but also prevents the IHS from rocking when you mount a cooler so as not to crush the core.

    @HondaCop,

    [​IMG]

    ^ Honda police NSX.
     
  3. HondaCop

    HondaCop Member

    Joined:
    Aug 4, 2012
    Messages:
    42
    Likes Received:
    0
    LMAO! Nice!!! We still have Crown Vics and now some Dodge Chargers... I have and prefer the Crown Vic... More spacious and comfortable. :thumbsup:
     
  4. Idontcare

    Idontcare Elite Member

    Joined:
    Oct 10, 1999
    Messages:
    21,130
    Likes Received:
    9
    The IHS does not "rock" back and forth, the IHS is physically locked/held in place by the socket's CPU retention mechanism prior to the HSF ever getting near it.

    There are two flanges on the IHS that are specifically located such that the socket grabs and locks it in place.

    [​IMG]

    In fact it grips IHS so well that I have to initially place the IHS in the wrong location over the die - I have to "back it up" by about 1-2mm - because the socket retention clip grabs it and slides it forward over top of the CPU (gliding on the CPU TIM) until the retention mechanism gets to a full stop with the entire spring loaded compression forces.

    That way when the retention clip is finally latched then the IHS is actually located in the correct place over the CPU, but it is very firmly fixed in placed and does not move at all (I have tried to push it with a screwdriver, it no move ;)).

    The only point to shim the IHS at this juncture is if you want to intentional create a larger gap between the CPU and the IHS. I don't want a gap (not in my final setup at least) but for now I aim to recreate the gap for data analysis purposes.
     
  5. Yuriman

    Yuriman Diamond Member

    Joined:
    Jun 25, 2004
    Messages:
    5,091
    Likes Received:
    55
    @IDC, you'd say then that a waterblock that bolts onto the board without springs is not likely to tilt the IHS even if the mounting pressure is extremely high? I'm concerned that if the mounting pressure is high enough to bend a motherboard if the bolts are half a turn too tight, the socket's bracket might not be sufficient protection. Metal flexes but for practical purposes it does not compress, which makes me feel much more comfortable with a shim.

    Edit:Also bear in mind that I'm using the 4 bolt holes so the pressure might not like up nicely with the mounting bracket. Pressure will be at the corners.
     
    #155 Yuriman, Aug 7, 2012
    Last edited: Aug 7, 2012
  6. Idontcare

    Idontcare Elite Member

    Joined:
    Oct 10, 1999
    Messages:
    21,130
    Likes Received:
    9
    I understand the concerns you have, I had them too until I put things into perspective once I started getting my hands on the situation (literally :D).

    One thing to not lose perspective on here is that the gap is only 140 microns. That is silly tiny, the same width as a human hair.

    Compare this to the width of the IHS itself and I have no doubt in my mind that the IHS is bowed down at the edges (convex shape over the CPU), and the underlying PCB is bowed down in the middle (concave shape) such that between the two the 140um gap is closed under pressure of the retention mechanism.

    That thing really is locked tight in there. If the gap was 10x larger, on the order of a millimeter, then I'd agree that the concerns have merit. And I'm not telling you to disregard your concerns, you need to do whatever tests you need to do in order to convince yourself as I have (or shim it for the sake of shimming it as you are detailing in your posts above).

    But I am telling you why I no longer have that concern now that I have played around with this delidded CPU, initially I was concerned but I am no longer worried about it. For me it was one of those concerns that turned out to be well-founded in the academic sense but not well-founded in the practical sense.
     
  7. Denithor

    Denithor Diamond Member

    Joined:
    Apr 11, 2004
    Messages:
    6,288
    Likes Received:
    20
    My worry wouldn't be the fact that the gap is tiny, it's that there is a gap at all - I'd definitely be using shims. If you have a gap between IHS and PCB that means the pressure from the locking mechanism and then from the HSF are being applied directly to the cores - not a state I'd want to have in my rig.

    In short - how is this any different from running bare core, except that you still have the IHS in the way of smoother transfer of heat? The danger in running bare is damaging the cores - which you are risking running with a gap between IHS and PCB.
     
  8. Yuriman

    Yuriman Diamond Member

    Joined:
    Jun 25, 2004
    Messages:
    5,091
    Likes Received:
    55
    Awesome, thanks. I'm considering popping my lid in a few days too.
     
  9. yottabit

    yottabit Golden Member

    Joined:
    Jun 5, 2008
    Messages:
    1,190
    Likes Received:
    0
    IDC, I also found some 0.2mm shim stock here:
    http://www.onlinemetals.com/merchant.cfm?pid=16061&step=4&showunits=inches&id=1189&top_cat=1175

    That may get you closer. The trouble with shim stock this thickness is cutting it IMO. If you use metal shears you are likely to get burrs that will throw off the overall thickness but it will be difficult to cut with scissors. Sharp kitchen or linen shears might be best.

    I'm tempted to see if I could waterjet or lasercut a spacer out of shim stock that would cover the whole delidded area, with a cutout for the die :)

    Also, if you're not in the habit of ordering industrial supplies all the time, http://www.mcmaster.com/ has all kinds of supplies including shim stock and specializes in same-day order processing and overnight delivery.

    EDIT: Sorry if I'm a little behind and the shimming is a non-issue now, I haven't had time to read the whole thread yet but rest assured I will!
     
    #159 yottabit, Aug 7, 2012
    Last edited: Aug 7, 2012
  10. Ferzerp

    Ferzerp Diamond Member

    Joined:
    Oct 12, 1999
    Messages:
    6,347
    Likes Received:
    46
    The die can handle it just fine. The alternative is a gap, and that is ruining cooling. He's not looking to shim it for safety. He's looking to shim it to test and see if any TIM performs poorly with the stock gap, or just the intel TIM.
     
  11. HondaCop

    HondaCop Member

    Joined:
    Aug 4, 2012
    Messages:
    42
    Likes Received:
    0
    I've been running my 3770K with the IHS resting on the die for almost a week now and I can safely say it's a non-issue. Performs perfectly.

    - Grammar *censored*, this was sent from my phone.
     
  12. BonzaiDuck

    BonzaiDuck Lifer

    Joined:
    Jun 30, 2004
    Messages:
    12,583
    Likes Received:
    265
    Of course, I've been watching this thread, and chipping in my two-cents' worth. And I will advantage myself of IDontCare's very good work. The good work includes his data collection and documentation.

    But under those circumstances, I will buy my Ivy Bridge chip at the most prudent time. And when I do, the delidding is going to be pretty much a single unrepeated task sequence. Reversing the final result will require the same trouble as the initial delidding. I plan to replace the adhesive with my own -- very likely the "Pit Crew" product I mentioned earlier.

    There should neither be a problem with a gap between the IHS and the die, nor a gap between the IHS and the PCB . . . Nor a problem of too much pressure on the die.
     
  13. oceanside

    oceanside Member

    Joined:
    Oct 10, 2011
    Messages:
    53
    Likes Received:
    0
    Those are some very impressive numbers HC. I think this pretty much puts the wrapper on how much heat density vs. how much Intel's special sauce affects the elevated temps on Intel's new 22nm process.

    Good stuff.
     
  14. HondaCop

    HondaCop Member

    Joined:
    Aug 4, 2012
    Messages:
    42
    Likes Received:
    0
    Thanks man and I'm still shocked! Just for shits and giggles, I pumped 1.52v to it and was able to boot up to Windows at 5.2Ghz and temps under full loads were mid 70s steady on Prime95. I know that it would probably not be stable and vcore will probably need a few more juice but I think it's doable.

    With this being said, guys, what is the most volts you can safely pump into the 3770K for a 24/7 use while being watercooled? If I can have temps not reach 80s under full load at 5.2Ghz, is that considered safe?

    - Grammar *censored*, this was sent from my phone.
     
  15. BonzaiDuck

    BonzaiDuck Lifer

    Joined:
    Jun 30, 2004
    Messages:
    12,583
    Likes Received:
    265
    "Most volts you can safely pump into the 3770K . . . " This is part of an ongoing discussion that -- if ever ended -- was resurrected as Intel stopped being more specific in its spec summaries. "More or less," "Here and there," the upper limit had variously been 1.37V through Nehalem; 1.35V on Sandy Bridge; and (maybe, tentatively) 1.30V on the Ivy.

    But these are just arbitrary thresholds in a a speculational statistical distribution which is "mathematically continuous" as opposed to discrete. What does it mean if you exceed such "thresholds" by 5%? By 10%? What does it mean if your peak (but momentary) VCORE ascends to 1.38V before the processor gets fully loaded and droops to 1.32?

    Like everything else, "you take your chances," and chatter about probability distributions, odds of electro-migration over different time spans, and everything else is "all about chances."

    Somebody remarked to me (here) the other day that INtel had updated its specs on the old Q6600 processor -- as much as two or three years after its release and as data became available. And I THOUGHT they told me Intel had revised a "maximum safe" limit on VCORE -- UPWARD.

    ADDENDUM: Just to add more detail to the landscape of "intel" about Intel -- a UK overclocking site seems affiliated with some outfit that sells over-clocked rigs. They insinuate that "they" have been in contact with Intel engineers, trying to get a grasp of the same answer to the same question. And I saw a remark which I'd guess is very reasonable: Intel promulgates spec information or "advice" to assure near-100% certainty of "NO RMAs."

    So . . . there's what you "know," what you "believe," what you "suspect," what you can "prove" or what you can "guess." You can prove a lot of things in Time, if you wait long enough . . .
     
    #165 BonzaiDuck, Aug 8, 2012
    Last edited: Aug 8, 2012
  16. HondaCop

    HondaCop Member

    Joined:
    Aug 4, 2012
    Messages:
    42
    Likes Received:
    0
    Thanks for the response... So long story short, it's anyone's guess? lol
     
  17. BonzaiDuck

    BonzaiDuck Lifer

    Joined:
    Jun 30, 2004
    Messages:
    12,583
    Likes Received:
    265
    "long story . . " -- my affliction. Not exactly "anyone's guess," but you'd consider what the processor is likely to run at under stock conditions, and know that there is some "equally harmless" range above that, and some "less harmless" or "more harmful" range above that.

    I'm currently in an ongoing exchange with a local friend, who worries that any over-volting and over-clocking will leave me with smoke pouring out of my computer case. We visit this forum, tweak our BIOS and run our machines the way we do because "we can."

    We'd like to keep them running that way for as long as they might run without the tweaking. Some might care more than others.

    And . . . "We take our chances . . . . " We take them variously, depending on personal priorities and other factors . . .

    AND ANOTHER ADDENDUM: There may be a myth afloat that the colder you can make the CPU, the less harmful higher volts will be. On the contrary, you would hope that making it colder will reduce the need for extra volts. But with this technology, that only holds true to some small extent . . . which . . . is what we were looking for with your lower temperatures and minimum voltage necessary to sustain 4.7 or 4.8. Or -- "MY" interest in those scenarios betrays my "conservative" attitude toward voltage.
     
    #167 BonzaiDuck, Aug 8, 2012
    Last edited: Aug 8, 2012
  18. sm625

    sm625 Diamond Member

    Joined:
    May 6, 2011
    Messages:
    8,175
    Likes Received:
    134
    To be clear, this is the area you're talking about shimming, correct?

    [​IMG]

    And the shim thickness should be somewhere in the neighborhood of 0.2 mm?

    Can you go over where the 0.06mm and the 0.140mm numbers came from? I thought the adhesive was 0.060mm so I figured all you want is a 0.060mm shim. Where does the 0.140mm come from?
     
  19. BonzaiDuck

    BonzaiDuck Lifer

    Joined:
    Jun 30, 2004
    Messages:
    12,583
    Likes Received:
    265
    I don't know what I like better -- a metal shim, or simply replacing the adhesive that has similar "elastic" properties to the original. Of course, many here would not want to leave the processor in a state where later access to the die would require the same initial tedium with a razor-blade.
     
  20. sm625

    sm625 Diamond Member

    Joined:
    May 6, 2011
    Messages:
    8,175
    Likes Received:
    134
    Yeah, the shim I placed looks like copper, but one could easily use a double sided foam tape which would compress down to about the right thickness. If you buy that cheap family dollar foam tape it wont stick very well so there wont be any problem removing the IHS later.
     
  21. Idontcare

    Idontcare Elite Member

    Joined:
    Oct 10, 1999
    Messages:
    21,130
    Likes Received:
    9
    Thanks for the helpful links and tips yottabit :thumbsup:

    I wasn't too worried about the edge being too rough after the cut because I aim to cut the width of the shim to be much wider than the width of the IHS landing footprint so the edge effects won't impact the shim height.

    The burrs and edge roughness will have plenty of room on both sides of the IHS after cutting out my shim templates.

    [​IMG]

    That said, presumably you've seen my dremel video ;) you know I'm not going to create a crappy looking shim, just not my style :p
     
  22. AnonymouseUser

    AnonymouseUser Diamond Member

    Joined:
    May 14, 2003
    Messages:
    9,939
    Likes Received:
    100
    I just wanted to point out that the original IHS adhesive did not form an airtight seal (notice the small gap at the bottom). This may be something to consider when resealing the IHS.

    [​IMG]
     
  23. Idontcare

    Idontcare Elite Member

    Joined:
    Oct 10, 1999
    Messages:
    21,130
    Likes Received:
    9
    Yeah you don't want to make an airtight seal, the entrapped air will increase its internal pressure as it heats up from the hot CPU and hot IHS.

    That pressure will not be enough to impact the IHS or CPU (they are structurally rigid enough to handle it) but the sealing adhesive and PCB are not, something will give so that the pressure is relieved, most likely the adhesive, so you may as well give it a controlled pressure relief point from the outset if you plan on resealing the IHS onto the PCB.

    That opening on the stock CPU is the reason why I do not wet-sand the IHS when I lap my CPU's, I only dry-sand. The last thing you want is for some water to get inside the IHS/CPU cavity. Its not an electrical concern, nothing will short out inside the CPU if some water gets inside, but the corrosion potential is huge and easily avoidable by just dry-sanding when you lap the CPU.
     
  24. Idontcare

    Idontcare Elite Member

    Joined:
    Oct 10, 1999
    Messages:
    21,130
    Likes Received:
    9
    That is correct.

    As for the where the numbers come from, see this post and this post.

    The metal shim gives you control over the gap height between the IHS and the CPU.

    Anything that is compressible, be it adhesive or foam, gives you no control over the final gap height - meaning it could still be zero and the IHS resting on the CPU die for all you know when the final compression of the HSF is brought to bear.

    So its a matter of thinking about why you care to do whatever it is you are doing. If you are shimming because you care to ensure the IHS is not resting on the CPU silicon then you don't want to use something that is compressible if you have the option of using something that is not.

    If you are re-adhering the IHS to the PCB simply because you don't want the IHS to fall back off the PCB with later handling of the CPU then that is a different story, rubber cement FTW.
     
  25. BonzaiDuck

    BonzaiDuck Lifer

    Joined:
    Jun 30, 2004
    Messages:
    12,583
    Likes Received:
    265
    I was thinking that replacing the original adhesive with a new adhesive having similar properties would allow for close contact with the die, while providing some protection against excessive pressure of the IHS spring-loaded assembly. I wasn't thinking that Intel did much different than what I describe, although maybe they did.