Delid fail, help me troubleshoot?

[Tristor]

Howdy Folks,

I decided since I was stuck at my folks house over the holidays without a lot to do I'd go ahead and delid my i5-3570k and migrate to a sealed watercooling loop (in my case the H110). The actual delidding process went okay, but I had to create a "custom" mounting bracket for the waterblock/pump in order to get it to seat fully on top of the IHS (I put the IHS back on top, used CLU between the core and the IHS after lapping the IHS to 2500 grit). I'm concerned that I may have exceeded the mounting pressure limitations of the CPU and/or the socket.

I've backed it off, but no dice still. Any suggestions on how to tell if the core is damaged during a visual inspection? I'm not planning to deal with it tonight, and I'm going to travel back home this week. When I get back to my place I plan to take it entirely apart again and give it a good visual inspection and see what I broke.

The current behavior is that when I power it on I get all the fans to spin up shortly, and then it immediately powers off. It never gets farther than that. Typically this would indicate a short, but I was exceedingly careful with the Liquid Ultra, so I don't believe I would have caused any shorts (I actually applied it outside of the chassis, then seated the CPU in the socket, set the IHS on top, and clamped it down before putting the waterblock/pump on). I'm not sure if I cracked the core (with the delid, the IHS sits directly on it) or I bent pins in the socket, but I don't think a short has occurred anywhere in the system.

Anyway, thoughts/commiseration would be welcome. If I'm unable to identify the cause for the failure, I suppose I'm going to order a new CPU and try again, although I'd be more comfortable doing so if I can prove beyond a doubt that I did not damage the mobo in any way.

 

[DigDog]

take many pics, post pics.

 

[schmuckley]

pics,pls.

 

[PliotronX]

The kind of damage it takes to render a chip inoperable might be outside of the closeup macro abilities of phone cameras but if you can get a real camera with macro capability try posting some shots of the socket, closeup of the four corners of the die, PCB top and bottom. Might be a dumb question but have you removed the old black sealant from the IHS?

 

[Tristor]

Sorry about the delay on my response to this thread. I got back home and luckily at home I have all the necessary equipment to fully test to identify what happened. After getting a lot of light and a magnifying glass on the core, I could see that it did not appear to be physically damaged. Beyond that, after examining the socket I found that it WAS damaged. It had about 5 pins bent.

This photo doesn't show them all, but this is the best I could do with a phone camera:

Anyway, I used the tip of a mechanical pencil to bend them back into place. While doing so I noted that two of the pins had been touching each other. Unfortunately this likely meant that the board wouldn't work (had been shorted). To test I put in a known good CPU (i5-2600k I had laying around), and the board stayed on somewhat longer before shutting off, but otherwise displayed the same behavior. I also started smelling something like burnt silicon so immediately shut the system off [I checked, the i5-2600k was fine afterwards to those that were worried].

To test the CPU more fully, I put it into a Z68 board I had lying around that had a BIOS that supported Ivy Bridge. The system was able to boot into the OS without issues. I did get a BSOD in Windows after awhile, but I think that's due to it being Windows 8 and its sensitivity to changing hardware (there was an i5-2400S in there prior). Later today (I haven't slept yet) I'm going to get up and reinstall the OS and then run some stress testing on the i5-3570k with Prime95/OCCT/IBT/LINX and see what I end up finding out.

It looks to me though like the CPU is likely fine and the board is definitely toast. So on that note, does anybody have recommendations for a replacement Z77 motherboard? My current board is no longer made It's the MicroATX version of the UD3) and the other MicroATX boards seem to get crap reviews, so I guess I'm going to put a full-sized board in here. I'd like to not spend a huge amount because if I have to spend more than $150 on a new board, I might as well just upgrade and go Haswell. If I do, I'll just buy a very inexpensive board and stick this CPU in another box in my folding cluster.

EDIT: To answer PilotronX's question, yes I removed all of the black silicone sealant from the IHS and PCB of the CPU post delid. My process for delidding was as follows:

1) Delid via the vise method (went smoothly, only took one solid tap)
2) Use a razor blade to scrape the majority of the black sealant off the PCB and IHS
3) Soak IHS and PCB in a 91% IPA bath for 2 hours
4) Using a buffing wheel on a dremel [low speed] soaked in 91% IPA (and repeatedly soaking as I went to keep it wet) I buffed off the remaining black sealant from both the PCB and IHS until it was all gone except in the corners of the IHS.
5) I used the razor blade, a needle, and the dremel to work in the corners until I'd basically gotten the IHS 99.999% clear of black sealant.
6) I lapped the IHS to 2500 grit mirror finish, verified flat by both a razor blade and the sharpie method.
7) Using a digital caliper I had made pre-delid measurements of both the motherboard and the CPU. I then made post-delid measurements. I had removed a total of 0.12mm of material lapping and had lost 0.091mm of material due to the change in gap. for a little over a 0.2mm difference in height (not that significant) since I put the IHS back on top.
8) I used M3 2inch bolts + #4 SAE and #6 SAE washers + the 8 thumb-nuts included with the Xigmatek backplate I used (since the H110 backplate is crappy plastic) and created a standoff difference of 14.5mm from the board (which is required due to the way the bracket on the H110 is designed, which is different than the H100).
9) I used the caliper to verify that the bolts were protuding identical lengths above the first thumb-nut on all 4 sides of the socket (indicating equivalent pressure on the backplate) and then did the same on the second thumb-nut on all 4 sides (indicating equivalent mounting pressure of the cooler at full-stop of the bolts).

That was my methodology for delidding and mounting the cooler to the top of the proc. Since I kept the IHS, the IHS sat on top of the die (I had applied CLU to the die) and I locked it into place using the socket's retention mechanism.

Hope that helps visualize, since I have minimal photos. I did all of this at my parents house with only what I had ordered to be shipped there online before I left home and with what tools and hardware I took with me or bought locally.

I'm still holding out before declaring the CPU good until after I've concluded my testing later today, but so far in my rather extensive overclocking/modding/PC building career I've yet to kill a CPU while also doing a lot of zany stuff. I hope this won't be the first one I've killed, and after realize the pins on the mobo were bent I feel a lot better since I thought everything went so well. Since the mobo was presumed good at the time I put everything back together, I could only consider my mounting scheme as being the culprit since it's always a risky plan. Seems I did pretty good with my measurements after I did more testing here. I could have actually increased pressure by another 0.3mm and it would have been okay, but this makes a firm impression on the top of the IHS so I'm going to leave it this way with the new board and/or CPU.

 

[BonzaiDuck]

. . .It looks to me though like the CPU is likely fine and the board is definitely toast. . . . .

Somehow, I think the expected value of losses in this regard of motherboards is higher for enthusiasts generally.

I occasionally debate buying extra "insurance" with duplicate part orders for my occasional builds -- to be used later in a second build if everything is copacetic.

It never seemed worth it. What we do here -- it's like the Meyer Lansky prototype said in an historical crime drama: "This . . . is the bidnis . . . we're in!"

All I did was to buy a sale-priced PSU from the manufacturer, which went south the second time I flipped on the toggle switch. I wasn't even over-clocking -- just being cautious! I had to test all the parts. It definitely fried the mid-range Gigabyte mobo somewhere; the CPU was fine; I think the RAM was fine.

If I were going to accumulate parts slowly over months for a building project, I'd quickly find that my choices would be limited to those unlikely to need timely confirmation of no defect.

 

[ali1988]

Out of curiosity, what causes the pins to bend like this. Is it accidentally hitting them with the corner of the CPU during installation? Or is it somehow possible for them to bend the wrong way just from installing a CPU regularly. Should be impossible to damage after installation from excessive mounting pressure or whatever else.

 

[BonzaiDuck]

Out of curiosity, what causes the pins to bend like this. Is it accidentally hitting them with the corner of the CPU during installation? Or is it somehow possible for them to bend the wrong way just from installing a CPU regularly. Should be impossible to damage after installation from excessive mounting pressure or whatever else.

This just points up the various risks of the mod. So far, we'd seen more than one:

-- Occasionally, some person here or there noticed a CLU-modded de-lid project whereby the CLU was somehow misapplied or caused "hot spots" on the die
-- Now this -- bending of pins
-- There's also been the reticence of using IC Diamond -- speculating that minute scratch-penetrations of the film on the die cause copper migration from the IHS or the directly-applied heatsink base (properly lapped to eliminate the nickel, of course . . )

I don't remember seeing cases where someone reported shorting or CPU malfunction due to shorting from the CLU or similar TIM. But we worried about it . . . . didn't we?

 

[rgallant]

Out of curiosity, what causes the pins to bend like this. Is it accidentally hitting them with the corner of the CPU during installation? Or is it somehow possible for them to bend the wrong way just from installing a CPU regularly. Should be impossible to damage after installation from excessive mounting pressure or whatever else.

-got to wonder if it is not peeps dropping said chip on the pins if it's over current ,poor connection on some power pins over loading ones making a good connection, they might heat up and warp ,
-seems like some cheap 1156 sockets years ago burnt up with that issue

 

[BonzaiDuck]

-got to wonder if it is not peeps dropping said chip on the pins if it's over current ,poor connection on some power pins over loading ones making a good connection, they might heat up and warp ,
-seems like some cheap 1156 sockets years ago burnt up with that issue

Innerestin' . . . .

 

[DigDog]

sorry if i butt in, i dont want to create a whole thread, my delid took ages with the vice method, then i went with razorblades and did it in 10 minutes ~ish.

lost 17C from the temps

in my case (4670k), the paste on the cpu was well placed, really nice print, and load @ stock was 77C; after putting a microscopic bit of coollab pro, 60C.

re: vice method, i tried, but it really wouldn't budge. i guess not all epoxy applications are made equal, and considering how easy the blade method was, i'd recommend it.

 

[Tristor]

sorry if i butt in, i dont want to create a whole thread, my delid took ages with the vice method, then i went with razorblades and did it in 10 minutes ~ish.

lost 17C from the temps

in my case (4670k), the paste on the cpu was well placed, really nice print, and load @ stock was 77C; after putting a microscopic bit of coollab pro, 60C.

re: vice method, i tried, but it really wouldn't budge. i guess not all epoxy applications are made equal, and considering how easy the blade method was, i'd recommend it.

I'm glad your delid went well. It looks like since I haven't found a Z77 motherboard to buy I'm going to be ordering a Z87 Gryphon + the "armor" kit + an i5-4670k today. I'm considering attempting the delid on the i5-4670k as well when I get it in, considering that the delidding aspect went very well in my previous attempt, just the mobo failure to throw a wrench in the works.

I'm a bit surprised though to hear you found the vise method to be more difficult. It's literally just one good whack with a hammer onto a block of wood nudged up against the PCB. I think the reason many people have so much trouble with the vise method is they're trying to be TOO careful and not giving a solid enough whack. If you do it right, you'll feel it give on the follow-through of the hammer swing so you can pull it short, just like pulling a punch after contact.

Anyway, I guess I'll let you guys know if I decide to do delid on proc #2.

 

[sm625]

Out of curiosity, what causes the pins to bend like this.

Only one thing can cause bent pins like that: human error.

 

[BonzaiDuck]

Only one thing can cause bent pins like that: human error.

I was trying to imagine. An equal amount of de-lidding advice does not suggest re-applying the IHS with something to replace the black rubber adhesive. Instead, they simply tell you to apply the IHS to the CPU's PCB on top of the TIM and die, then drop it into the socket and apply the levered clamp.

I can imagine the possibility that the CPU and IHS would tilt before the lever is locked in place, and that might cause the pins to bend at the pivot-point. So one would need to hold the IHS/CPU steady and flat while locking it in place.

 

[Tristor]

Only one thing can cause bent pins like that: human error.

In this case, I think they were bent during the 3 day gap while I was working on the CPU (it took a long time to lap the IHS, plus I had to wait on some parts). I seriously doubt that I bent any pins while putting the CPU back into the socket all things considered, and given the followup to this it's even more doubtful. Since I didn't have the CPU in it and I didn't have the black cover that comes with the mobo (since I was ~800 miles from home), the pins were exposed other than a paper towel I laid on top of the mobo to keep debris from falling in from the air. I have no idea what exactly occurred, but it makes the most reasonable sense that it happened while the pins were exposed for 3 days and not due to direct error in installation.

At any rate, it's dead Jim. The new parts are ordered. We'll see how the next one goes. For the record, this is the first component I've completely killed unintentionally through warranty-voiding activities. I guess we all have a few

 

[DigDog]

~~~
I'm a bit surprised though to hear you found the vise method to be more difficult. It's literally just one good whack with a hammer onto a block of wood nudged up against the PCB. ~~~

One would think that's true, given the many videos around; but i literally cut a pcb-shaped hole in my piece of wood from whacking one end with the hammer and the cpu's green bit trying to burrow into it.
At that point i gave up; never did it before, had no idea how much punishment the pcb can actually take.

Also, its worth mentioning that the tim application on my sample was pretty good. Perfect print, was dry but not "rock hard" as others have. I guess the "TIM lottery" is a thing, now.

So, before you give it "one whack", test it, it might be not necessary to remove the IHS.

Also, *even* if you are going with the vice method, i recommend easing the process by using a bit of the blade method first.

I have read that one can damage the cpu with the blade method, from both scratching the pcb and from ... cutting capacitors?
Anyway, from what i've seen, with the blade, you can be careful, and slow, and you *will* get the IHS off. With the vice method .. hmm.. i dont know.

 

[Tristor]

Well, I'm delidding the new proc. All the parts should be delivered in the next few hours, I probably won't be able to get around to doing the delid until this weekend, but we'll see.

 

[BonzaiDuck]

Well, I'm delidding the new proc. All the parts should be delivered in the next few hours, I probably won't be able to get around to doing the delid until this weekend, but we'll see.

I have this strong, irrational urge to build a new computer so that I can de-lid the core and see how it works out. But we have five computers in the house -- my Sandy the only one that is seriously overclocked, nobody feeling their machines are slugs.

And there's no room for one more. What would I do? Junk my standby E8400 system, which I use for bidnis and could otherwise be the guinea-pig system?

Nobody would want to BUY an E8400 mobo/processor -- maybe one person I know. Works fine for what I use it for, and that gives it full utilitarian value. Still -- I could junk my mom's system instead and give her the E8400 with a spare case I have.

Not any time too soon . . . .

 

[Tristor]

I don't know man, but I honestly have to say using the vise method the delid process isn't that terrible. I just got back from a run down to Harbor Freight. I picked up a set of digital calipers, a table vise, and a rotary tool, so I'm going to go ahead and do the delid and the lap tonight. Should have it working by morning and I'm taking plenty of pictures so I might write a guide (assuming I don't miserably fail this time around).

Edit: Haswell has some more stuff on the PCB next to the CPU core that Ivy Bridge doesn't. It didn't really make the delid any worse (although I should note that Haswell had significantly less gap than Ivy Bridge, so my benefit may not be as much). It does make the process of applying CLU a bit more treacherous though. I'm considering using liquid electrical tape on top of the VRMs(?) next to the die on the PCB before applying CLU to minimize the risk of causing a short internally to the CPU. Any thoughts on this?

 

[Tristor]

*sigh*

Well, I got everything completed and put back together. Unfortunately the system doesn't POST. It does stay on (all the fans continue spinning) after power on, but I have no display output and none of the LEDs on the board light up. It's an Asus Z87 Gryphon. First time I've ever bought an Asus board, I typically get Gigabyte UD3 or UD5 or EVGA Classified. I mainly got it for the backplate (I purchased the "armor kit") since I thought that may help prevent motherboard flex with high mounting pressure.

I'm going to continue troubleshooting. Much to my chagrin I got ahead of myself and did not test the CPU/Board as received prior to going straight into my process of delidding and lapping. All things considered, this one came out even better than the last (which works) so the likelihood of it being a failure caused by the CPU is very slim.

We'll see...

EDIT: I've pretty much tried all troubleshooting options and I've got nothing. It never POSTS, I get no lights on the board like I'm supposed to, not even trouble lights. I get nothing on a PC Speaker I found and plugged in. No beep codes. After reading more online it appears I am not the only person to experience this with this particular board and that this isn't entirely uncommon with the Z87 Asus boards. Apparently they have some issues in their non-ROG boards. I guess I'm RMAing the board to Amazon, hopefully I can get them to cross-ship overnight so I can identify very quickly for sure (hoping the next one works).

If the CPU were an issue it should be lighting the "CPU LED" on the board as well as emitting a beep code, and more severe CPU issues would typically cause the box to power off. The behaviors I'm seeing are identical to the behaviors others have reported with non-modified CPUs and bad instances of the Z87 Gryphon, so I'm pretty sure it's the board. Seems I can't catch a break.

EDIT2: I've submitted the RMA. First time I've ever done a return through Amazon and they're pretty rocking. They automatically cross-ship via their online form. My replacement mobo should be here Saturday, so I guess I'll hopefully having a working system then. They even picked up the tab for return shipping. Newegg is great, but I always have to call them on the phone if I want them to cross-ship and eat the shipping.

 

[Headfoot]

Good thing you got that RMA done through Amazon. Suffice it to say, I've never had worse RMA service in my entire life than through ASUS. I literally can't say enough bad things about their customer service. Absolutely terrible. Amazon is awesome on the other hand, always treat me well.

 

[Tristor]

Well, I received the replacement motherboard. Unfortunately it exhibits absolutely identical behavior. Either I've gotten more than one DOA board in a row (possible), or the CPU is a non-functioning state. While I find the CPU being non-functioning very unlikely based on a close physical examination with a magnifying glass, and the vanishing small failure rates of CPUs out of the package.

At this point I can't really make a determination without a tester Haswell CPU to verify if it's the board or the CPU. I'm going to reach out to my coworkers this week to see if somebody will let me borrow one. Also, if any forum members are in South Texas and have a Haswell CPU I could borrow for max a couple of a days, I'd very much appreciate you sending me a PM.

 

[PhIlLy ChEeSe]

Well, I received the replacement motherboard. Unfortunately it exhibits absolutely identical behavior. Either I've gotten more than one DOA board in a row (possible), or the CPU is a non-functioning state. While I find the CPU being non-functioning very unlikely based on a close physical examination with a magnifying glass, and the vanishing small failure rates of CPUs out of the package.

At this point I can't really make a determination without a tester Haswell CPU to verify if it's the board or the CPU. I'm going to reach out to my coworkers this week to see if somebody will let me borrow one. Also, if any forum members are in South Texas and have a Haswell CPU I could borrow for max a couple of a days, I'd very much appreciate you sending me a PM.

One prick of the PCB and the CPU is dust, so your magnifying glass is useless as you wont see the damage. The fact you still have a problem shows you are the issue at hand.............Which means the original board is still good mostly likely.

 

[DigDog]

wait .. you delidded *before* you tested the CPU? not even a token run ?

 

[PliotronX]

wait .. you delidded *before* you tested the CPU? not even a token run ?

I do that all the time. It is exceedingly rare for CPU's to be DOA.