My Absolutely Horrifying Delidding Story (Long)

[flexy]

Since I had a new case coming today and wasn't too happy with my 4770k's temps I thought I wanted to delid it.

I had already watched many videos about delidding Haswell, so I know about the Vice method, the razor method etc. Since I am rather tech- and mod-savvy I didn't expect there being a problem.

I had CLU here, I even got me different cutters from the hardware store since I read that the common, very thin razor blades could be a problem. I also had tape to insulate the caps and even got me gloves to avoid static discharge etc. In other words I was more than ready to finally get this done.

I dismounted the board, the ol' HSF, cleaned everything, took the 4770k out,

Finally I started cutting as I saw in countless videos, CAREFULLY of course.

The first thing which came to mind is that the CPU when you work with it "in real life" is a lot smaller than what you'd expect, maybe because images and videos make it appear larger than it actually is when you have it in front of you. The second thing was realizing that my eyesight isn't the best any more, really.

Anyway I TRIED cutting into the corners but the black stuff IS ROCK SOLID, LIKE GRANITE. I spent possibly 20mins even just to attempt to cut into one of the corners, NO DICE. ZERO.

I have heard from people how easy it supposedly is "No sawing!"....and the blade would more or less just slid in there automatically. NO. WAY.

So I tried it multiple times, forceful, then sawing a little. NOTHING. This glue stuff is rock solid and I have no idea how someone can just cut this.

THEN....something horrible happened, I sort-a slipped with the knife.

At first I didn't think it would be a problem since it didn't look like that the slip hit or damaged something. (I was already sweating GALLONS, and the PVC gloves didn't really help there either.)

So by chance I turned the CPU over and GOT A SHOCK.

At the edge of the CPU, some of the golden contacts were broken, sort-of "flaked off". I didn't believe my eyes.

Obviously I had anything in mind BUT taking a photo but here is a stock-image where I sort-of marked about what was wrong with the CPU, 3 or 4 golden contacts were missing and there was a white spot where it's supposed to be green.

Also, all around the edges of the CPU I could see more green spots "flaking" in several spots, but this part I marked in pink had definitely pins MISSING as well.

I almost cried!!

I run through the house, out on the balcony where there is bright light and looked at the spot and yelled "this is not supposed to look like that, this is not supposed to look like that", realizing the horror that I just very likely destroyed my CPU which I just got about 3 weeks ago.

I told my wife about it and that I probably just destroyed my CPU because I did something really, really, really stupid. (The max gain I expected from delidding was maybe 200mhz since I have it already stable at 4.4G).

I was devastated since this PC is also what I use for work, and I just repeated how why in the world was I so stupid trying to delid it.

Of course the idea of continuing with delidding was now the farthest on my mind. I didn't even open the pack with the CLU yet.

In my panic I now re-assembled the board, put back CPU, HSF, MX-4 on still lidded CPU....AND THIS THING BOOTED.

Not only did it boot, I also tested stability with OCCT.

Long story short: I re-built my PC and I am up and running again and it's a miracle. I have no idea how this CPU can work still.

Re-tracing my steps, I don't think I nicked the chip when the knife slipped, I would have noticed this. IF (and that means: IF) I caused the damage to the CPU it might maybe have happened when I opened and removed it from the socket. (For some reason this just didn't feel 'right' how it came out).

My other suspicion is that I possibly got the CPU like that. (I bought it used, it was on the board so I never really checked it. I also didn't *closely* check the underside of it today before I started delidding so I don't know whether those contacts which flaked off were already missing before. The more I think about it the more I think the CPU *was* like this before already...)

Conclusion:

Trying to delid with a razor or small utility cutter: INSANE. I do not, not ever recommend this. I have no idea why the black gunk was so hard, there was simply no way to cut into this, not as videos etc. show. I am not stupid (many delidding guides exist etc.) so I am baffled why I couldn't cut into this thing at all. IF it's possibly to cut into the edges then it requires a lot of force, and I mean way too much force for my taste.

Maybe my CPU is relatively old (pre-owner told me system was 6 months old), maybe cutting is easier with brand-new ones and the "gunk" sort-of hardened, I don't know. It's possibly that CPUs with age just get "gnarly" so that any removing etc. has the risk of the golden contacts flaking.

Right now I am the happiest person that my system still works.

 

[NTMBK]

If you use it for work, what are you doing overclocking at all? 15% faster for the risk of silent data corruption is a poor deal. If you really need more performance spring for a 4790k, otherwise leave it at stock.

 

[KingFatty]

The problem is your CPU has herpes

 

[Bubbleawsome]

Honestly no one really knows what most of those pins do, and they don't seem important. I'm sure if you cut enough of them it would mess it up, but they seem very redundant.

 

[VirtualLarry]

Wow, OP. Lucky story!

 

[sm625]

It looks like you lost pads AF39,AF40,AE39, and AE40. One of those is VSS, which you could easily do without. But the other 3 are critical signals. lol. They are DDR data lines and data strobe lines. I do not see how your computer can function without them. Perhaps the LGA pins are making contact with what is left of the wire that goes from the pad to the die? If so then do not expect it to work for long. Someone should doublecheck the pads. Here is a datasheet . Refer to sec 8.9 and 9.0

 

[thilanliyan]

I also considered delidding on my crappy clocking 4670k, but decided it wasn't worth the rist of damaging it.

I thought the vice method was the safer one?

 

[james1701]

So, can you send me you CLU. I'll spring for shipping

 

[flexy]

It looks like you lost pads AF39,AF40,AE39, and AE40. One of those is VSS, which you could easily do without. But the other 3 are critical signals. lol. They are DDR data lines and data strobe lines. I do not see how your computer can function without them. Perhaps the LGA pins are making contact with what is left of the wire that goes from the pad to the die? If so then do not expect it to work for long. Someone should doublecheck the pads. Here is a datasheet . Refer to sec 8.9 and 9.0

I didn't mark the *exact* ones, I made this "from memory". I don't even know whether this (in the image) is the right side. This was just an example, the *exact* pins/location I have no idea.

 

[Idontcare]

Conclusion:

Trying to delid with a razor or small utility cutter: INSANE. I do not, not ever recommend this. I have no idea why the black gunk was so hard, there was simply no way to cut into this, not as videos etc. show. I am not stupid (many delidding guides exist etc.) so I am baffled why I couldn't cut into this thing at all. IF it's possibly to cut into the edges then it requires a lot of force, and I mean way too much force for my taste.

Maybe my CPU is relatively old (pre-owner told me system was 6 months old), maybe cutting is easier with brand-new ones and the "gunk" sort-of hardened, I don't know. It's possibly that CPUs with age just get "gnarly" so that any removing etc. has the risk of the golden contacts flaking.

Based on what you report, my conclusion would be that someone already delidded your CPU and reglued the IHS with an epoxy, probably damaging the pins in the process.

Once they realized they weren't going to get a golden OC, they sold it used.

 

[flexy]

Based on what you report, my conclusion would be that someone already delidded your CPU and reglued the IHS with an epoxy, probably damaging the pins in the process.

Once they realized they weren't going to get a golden OC, they sold it used.

Something like that came to mind, but the CPU just from looking at it looked "untouched", at least from the top. But who knows, maybe someone did a pretty good job there glueing it back together?

All I know is that it was literally impossible to cut either in the corners or anywhere. I 100% didn't expect this since noone in any guide or Youtube video mentioned major difficulties to even cut, they use a knife or a blade and don't seem to have a major problem with this. (I didn't expect the glue rock-hard but it definitely was).

 

[flexy]

I also considered delidding on my crappy clocking 4670k, but decided it wasn't worth the rist of damaging it.

I thought the vice method was the safer one?

POSSIBLE, likely - but not 100% convinced either. I had seen some reports of people who discovered their dies shattered after mounting their HSF....and there is the theory that smashing them like you do with the vice method isn't exactly "healthy" either. That being said, should I ever re-consider delidding...VICE it will be, but right now I am done with whatever delidding after this scare.

 

[njdevilsfan87]

I had the same issue delidding a 3770K. The stuff was rock hard. I did manage to get it into the corners, but because of how difficult it was, I was not confident going the rest of the way, so I finished with a vice. Even that wasn't easy because my CPU went flying through the air while time slowed down as I watched the bare exposed CPU.

It was worth it in the end though. Massive decrease in temperatures. My little brother now uses that chip.

 

[flexy]

It looks like you lost pads AF39,AF40,AE39, and AE40. One of those is VSS, which you could easily do without. But the other 3 are critical signals. lol. They are DDR data lines and data strobe lines. I do not see how your computer can function without them. Perhaps the LGA pins are making contact with what is left of the wire that goes from the pad to the die? If so then do not expect it to work for long. Someone should doublecheck the pads. Here is a datasheet . Refer to sec 8.9 and 9.0

Thanks, great information!

As said, the exact pins I don't know.

I did various tests now, including memtest, OCCT, graphical benchmarks etc...I can't see anything wrong. So whatever flaked off must be redundant and non-criticial, at least for my system. (Likely also that pins flaked off which would only play a role with more memory or under certain other configurations. I don't know.)

Right now I am tending to say it was not me causing this, for example also the flaking edges which were basically on several sides of the CPU, not only there where the pins are missing. Even if my knife slipped and it would've had hit the CPU (which I don't think it actually did) it would not have caused flaking on *several* locations on the CPU. It's also unlikely this all happened from removing the CPU from the socket. If so I would've probably found flakes, copper etc. on the socket....

 

[flexy]

It was worth it in the end though. Massive decrease in temperatures. My little brother now uses that chip.

Of course that was my reasoning too, I am OC at 4.4G now and under STRESS TESTING like OCCT etc. I see up to 90 degrees. (At 1.222V)

BUT....the point is, running stress tests like ASUS Real Bench and other more reasonable test I "only" see mid 70s...and gaming, including BF4 the CPU barely gets 60C. So in "real life" I will never see my CPU hit 90s.

 

[flexy]

It looks like you lost pads AF39,AF40,AE39, and AE40. One of those is VSS, which you could easily do without. But the other 3 are critical signals. lol. They are DDR data lines and data strobe lines. I do not see how your computer can function without them. Perhaps the LGA pins are making contact with what is left of the wire that goes from the pad to the die? If so then do not expect it to work for long. Someone should doublecheck the pads. Here is a datasheet . Refer to sec 8.9 and 9.0

SM625,

I did some "best guessing" now (still from memory) so it's not 100% accurate. According to the datasheet, my missing pins are likely in range G to M, 39 and 40. (Can be any of those. 2 to 4 pins total missing)

(The defect is at the side where the right notch is, looking at it from the underside like in the image. Mind you, the first image in the data-sheet shows from above, not the pin side from below)

So I checked a number of pins in the likely range (so basically pins F to U, 39s and 40s).

What I found is that most pins in that area are "testability signals" that seem only to be used by (external?) debugging and monitoring tools.
I found a few more candidates in this range, but they were either RSVD or not used at all. There was also one or two VSS in that range.

It seems that indeed in the affected range there is no critical signals...I guess I was incredibly lucky.

 

[hawtdawg]

I did the razor method on my 3770K a couple of years ago. Don't know why anyone would try it that way now that people know about the vice method.

 

[escrow4]

. . . stuff was rock hard. I did manage to get it into the corners, but because of how difficult it was, I was not confident going the rest of the way, so I finished with a vice.

I swear read that alone I nearly laughed my way into a coma. :awe:

 

[BonzaiDuck]

Just like everyone else, when I need a quick answer for something, I post a thread versus running long searches for lengthy threads that might provide the answer.

I wasn't clear that the OP's problem with CPU contact pads had anything to do with his de-lidding frustrations or efforts thus far.

I only observe that I posted a thread on this very same issue about a month or two earlier. I purchased a retail 2700K in the blue box -- factory-sealed with the official Intel tape.

For the first time ever, I was inspecting both a motherboard for bad/bent pins (an AsRock Z77 Extreme 4 -- which I RMA'd for a refund), and the processor. I was using the strongest magnifier I could find. I discovered -- like the OP -- a contact pad with part of the gold-plate missing.

So I learned something after posting a thread "Does Intel ever ship defective product?" -- something like that. An Intel employee -- engineer -- posted responses.

Before they put the processor in the retail box, seal and ship it, they run each and every processor through three batteries of software tests. After that, somebody makes a visual inspection -- more likely to miss something.

It wouldn't have reached the packaging and shipping stage if it had failed those tests.

So I'm not surprised the OP's processor worked fine. I'm also not surprised he went into a panic about the processor: my own experience sure caused me some concern!

It's useful to study the pinout diagram for any particular processor model. There are "dead" or unused contact pads (or motherboard pins, also). There are "reserved" contacts and pins, and I think there are other categories which have absolutely no bearing on processor operation.

Of course, just studying those diagrams will strain your eyes, and it's important to realize that the processor pinouts are a mirror-image of the motherboard pins.

 

[PG]

Things like this are the reason I wouldn't even consider lapping or delidding anymore.
I have to admit I lapped a core 2 duo once back in the day, but I wouldn't do it again.
The loss in resale value and the possibility of destroying the cpu outweigh any potential benefits in my opinion.
Of course people can do as they wish.

 

[BonzaiDuck]

Things like this are the reason I wouldn't even consider lapping or delidding anymore.
I have to admit I lapped a core 2 duo once back in the day, but I wouldn't do it again.
The loss in resale value and the possibility of destroying the cpu outweigh any potential benefits in my opinion.
Of course people can do as they wish.

And . . . I do . . . But I began to consider those very worthy points more recently. If you choose to lap off the nickel-plate, it might be wise to assemble, test and then dis-assemble before lapping.

Some of this dovetails with one's usage and resale intentions. My father was a car-mechanic before he got into the white-collar insurance business. He quickly embraced the trade-in program for cars at the dealership, and we had a new car on average every two years.

I've always attempted to squeeze ten years or more from automobile ownership before either selling a used car or simply having it towed to the junkyard.

If I didn't have family members whose upgrades didn't have to be latest-greatest, I'd think more than twice about lapping with EBAy prospects . . .

 

[monkeydelmagico]

I considered delidding a 3570k recently. A cursory inspection and some probing with a razor blade around the edges quickly convinced me not to mess with it.

 

[flexy]

Bozaiduck, it gets better:

I came across a thread here on AT from some years ago where there were problems with 1156 sockets made by Foxconn.

Check this out:

http://www.anandtech.com/show/2859

In this case they found that poorly manufactured sockets didn't seem to make contact (!) with, like, 70-80 pins, check that image. Not 2 or 3 - 70 or 80!! So (at least back at the time when the Foxconn problem existed) although it certainly wasn't exactly "benefiting" the CPU, those CPUs worked too even with 80+/- pins not even making contact.

On the other hand, it looks to me as if the damage had happened on the other side of the CPU I likely would've been in trouble. (As the other guy pointed out, data strobe signals etc. on that side, I sure wouldn't want to "lose" those).

So or so..after that scare..I am really wondering WHAT makes those pins flake off? Did my pre-owner of the CPU (or the pre-owner before him, assuming he existed) employ some crazy cooling that caused extreme temps in combination with crazy high voltages/currents? I just can't see why pins would "flake off" like this in normal use. ("Normal" here meaning to operate at stock or stay in healthy voltage and overclocking ranges). It's just odd seeing this.

 

[chimera15]

I tried to "delid"? take the heat spreader off right? off an old 775 processor and the chip was glued to the spreader essentially, so ruined the whole chip, the processor separated off the pcb when I tried to lift it off. Never thought of trying that again... people actually do that? I guess it depends on the cpu huh.

But yeah I know from fixing bent socket pins that the first thing they tell you is likely as not the pin if it's super bent or broken it's fairly likely the processor doesn't even care or use it anyway.

 

[rpjkw11]

Each time I've delidded a CPU, I've used the hammer & wood block method successfully. But each time I sweat bullets and swear I won't press my luck again.