Getting ready to delid Haswell..

[JimmiG]

I think I'm gonna do it. I'm sick of my CPU running at a scorching 87C+ while my huge, expensive heatsink is cool to the touch.

I'll use the hammer+vice method because I've read too many horror stories about people cutting into the PCB with the razor method.

This method seems to be pretty "safe" (if hitting your CPU repeatedly until it breaks into hopefully no more than two pieces can be considered safe..). The biggest risk seems to be that the CPU flies off and hits a wall or something because you hit too hard and didn't have anything to catch the CPU.

What's the safest way to clean off the epoxy from the PCB? Apparently the memory controller lanes are really close to the surface of the PCB, so you don't want to scratch them. Would a credit card or similar work?

After de-lidding I'll use CLU between the IHS and die. Is a small brush the best way to apply this? Do I need to apply it to the underside of the heat spreader as well, or is it enough to apply it to the die?

What about the VRM components to the side of the die? Some say you should cover them with something to protect them from the (conductive) CLU, but most seem to ignore this step. Is it really necessary, and if so, what's best to use? I feel that covering might maybe cause them to overheat?

 

[NTMBK]

Those about to delid, we salute you!

 

[Aikouka]

What's the safest way to clean off the epoxy from the PCB? Apparently the memory controller lanes are really close to the surface of the PCB, so you don't want to scratch them. Would a credit card or similar work?

In IDC's thread, he used a dremel with a polish adapter that was dipped in isopropyl alcohol.

 

[Elfear]

I'm about to delid my chip as well and hopefully this question will help the OP so it's not a blatant thread-jack but what do you use to seal the IHS back to the PCB when it's all cleaned up?

 

[JimmiG]

I'm about to delid my chip as well and hopefully this question will help the OP so it's not a blatant thread-jack but what do you use to seal the IHS back to the PCB when it's all cleaned up?

I think most just put the IHS on top of the PCB and clamp it down with the retention bracket This will also ensure that the gap between the IHS and die is as small as possible.

 

[Mushkins]

I think most just put the IHS on top of the PCB and clamp it down with the retention bracket This will also ensure that the gap between the IHS and die is as small as possible.

If you're putting the IHS back on, you typically replace the stock thermal compound with a much thinner coating of something considerably better. Without anything you'd end up with uneven contact and worse- an insulating air pocket thats probably going to end with a fried chip.

Alternatively, some people elect to go bare die and attach the HSF directly to the exposed dies after applying new thermal compound. You'll get better temps, but you run the risk of cracking the die and ruining the processor if the HSF isnt mounted properly or is too tight.

 

[Elfear]

If you're putting the IHS back on, you typically replace the stock thermal compound with a much thinner coating of something considerably better. Without anything you'd end up with uneven contact and worse- an insulating air pocket thats probably going to end with a fried chip.

Alternatively, some people elect to go bare die and attach the HSF directly to the exposed dies after applying new thermal compound. You'll get better temps, but you run the risk of cracking the die and ruining the processor if the HSF isnt mounted properly or is too tight.

I believe the OP was referring to the sealant (or lack thereof) between the edges of the IHS and the PCB after delidding rather than the interface between the chip and the IHS.

OP - Thanks for the info. I guess you'd just have to be really careful if you sold and shipped the chip one day.

 

[tw33k]

I use Sekisui #5760 thermal tape to attach the IHS to the PCB. It's so thin that it has no impact on temps and it isn't difficult to remove the IHS again if needed

 

[JimmiG]

Yeah I was talking about the sealant.

Between the CPU die and IHS, you can use anything really, but from what I understand, MX-4 etc. are not optimized for direct to die application so will quickly degrade.

Coollaboratory Liquid Ultra or Pro seems to be the TIM of choice when delidding. The drawback is that they're conductive and also quite runny, so you have to be really careful, especially with Haswell because of its surface mount components (capacitors?) on the PCB. You basically "paint" it on with a brush and then check to make sure there are no actual droplets left.

 

[Idontcare]

Coollaboratory Liquid Ultra or Pro seems to be the TIM of choice when delidding. The drawback is that they're conductive and also quite runny, so you have to be really careful, especially with Haswell because of its surface mount components (capacitors?) on the PCB. You basically "paint" it on with a brush and then check to make sure there are no actual droplets left.

Use CLU, there really is no better alternative.

But use it correctly. When you brush on the CLU it should be so thin that it looks and acts like you painted a layer of paint on the die. There should be no drops, nothing liquid-like whatsoever.

You only have issues with a conductive TIM like CLU if you can't apply it correctly.

The same goes for delidding, you only have issues delidding if you personally can't delid the IHS correctly.

So these are all part and parcel to the same overall goal - delid your processor and replace the TIM without destroying the CPU itself.

You have to do each step correctly, and that includes correctly selecting the TIM you are going to use (use CLU) and correctly applying it (thin non-runny layer that looks like silver paint).

Small drop of CLU in the center:

Brushed on and thinned, the excess removed with the brush, leaving a surface that looks like it has a thin layer of silver paint on it:

When it is that thin the CLU doesn't migrate, ever. It won't matter if the CPU is on its side, upside down, spun in circles, etc. The surface tension is such that it adheres to the die and the IHS without migrating.

 

[Soulkeeper]

That black adhesive is silicone, similar to the black RTV you get at the local hardware store.
I used the black RTV to put a IHS back on a llano.
Just note that solvents are unlikely to melt or clean that stuff off, you're more likely to melt the conformal on the cpu package itself.

Delidding haswell with the razor blade method should be much easier than AMD (since the SMD caps are ontop under the IHS). Taking your time is key imo. I felt like if I was too nervous/impatient i'd be likely to screw up.

One method I picked up was marking a line accross the razor blade so I'd know when I went too deep. Just be carefull around the corners.

 

[JimmiG]

Looks like there's some spill around the die. If that was a Haswell, it would nearly touch those VRM's...

 

[Idontcare]

Looks like there's some spill around the die. If that was a Haswell, it would nearly touch those VRM's...

That's not spill, I had no reason to be careful with my brushing of the CLU and I let the brush touch the PCB.

Obviously had there been electrically active elements in that area then I would have been more careful about temporarily masking off the area with tape which I would have then removed after applying the CLU.

 

[JimmiG]

That's not spill, I had no reason to be careful with my brushing of the CLU and I let the brush touch the PCB.

Obviously had there been electrically active elements in that area then I would have been more careful about temporarily masking off the area with tape which I would have then removed after applying the CLU.

Yeah that might be a good idea. Ivy Bridge is much more forgiving since there are no components on the PCB other than the die itself. You'd have to really mess up and get the stuff on the pads or on the actual mobo...

Wouldn't there be a risk of ripping the capacitors out when you remove the tape?

 

[Idontcare]

Wouldn't there be a risk of ripping the capacitors out when you remove the tape?

Zero risk, those capacitors are soldered onto the PCB. Your tape would have to bond to the capacitors more strongly than that existing metal bond strength if the tape was to rip out the capacitors.

 

[sm625]

I wouldnt say "zero" risk. I've pulled small 0603 and 0402 parts off a board when peeling back tape. Especially the lead-free crap, it doesnt hold nearly as well.

 

[Idontcare]

I wouldnt say "zero" risk. I've pulled small 0603 and 0402 parts off a board when peeling back tape. Especially the lead-free crap, it doesnt hold nearly as well.

What kind of tape are you envisioning using for masking off the caps?

I'm assuming an enthusiast is going to use standard garden variety scotch tape used for paper.

Nothing is zero risk if one is intent on being silly about it, I'm giving the benefit of the doubt here in assuming the guy isn't an idiot and isn't going to use some super-adhesive duct-tape like material for a simple task like this.

 

[zephyrprime]

Coollaboratory Liquid Ultra or Pro seems to be the TIM of choice when delidding. The drawback is that they're conductive and also quite runny, so you have to be really careful, especially with Haswell because of its surface mount components (capacitors?) on the PCB. You basically "paint" it on with a brush and then check to make sure there are no actual droplets left.

Just paint the smd components with some paint of nail polish remover. That will be enough to insulate them.

 

[JimmiG]

Done.

All naked. Is this considered pr0n?
The remains of the epoxy seal look much worse than they are in this pic. The layer left was extremely thin (in fact the bulges from the traces you can see are thicker).

Reminds me of the old Socket A / 370 days. Life for the overclocker was much simpler back then :hmm:

Applying CLP, officially the most difficult thermal compound to apply (and to photograph).
I decided to apply the CLP with the CPU in the socket to avoid accidentally tilting the CPU with the CLP on top. Just be very, very careful so nothing drips on the mobo...
I cannot stress enough jus how little you need. I thought I used very little, but once I started spreading it, it just kept spreading and spreading...

The brush I used to apply the CLP. Worked great. Make sure you have at least 3-4 of them so you can switch to a clean one from time to time.

CPU assembled again.

Initial tests:

Previously, it would peak at around 87-88C with dips to around 82-83C, now 74C with dips to the mid 60's. 4.3 GHz / 1.218V.

So far the drop seems to be at least 12 - 14C. Not the huge 25+C drop some have reported, but I am using an air cooler (Phanteks PH-TC14PE) rather than liquid cooling. Apparently the CLP takes up to 48 hours to cure so we'll see how it changes in the next few days.

 

[BallaTheFeared]

What MHz/Voltage?

 

[biostud]

Intel should sell the K chips de-lidded.

 

[JimmiG]

What MHz/Voltage?

4.3 GHz, 1.218V. Uncore is 4.1 GHz I believe and 1.1V. My CPU requires quite a bit of voltage even for relatively low overclocks. If I was able to do 4.5 GHz+ @ 1.2V or less, temps would not have been such a big problem.

Here's Intel's new LinPack with AVX2 (HT needs to be disabled for this one).

Pre-Delid:

Note that even though the max Aida64 has recorded is 98C, the CPU has actually throttled at one point, meaning the temp must have spiked at over 100C at some point.

Post Delid, same settings, roughly the same room temperature (maybe 1C warmer):

Looks like I'm ready for some AVX2 action

 

[BallaTheFeared]

What's your stock vid?

 

[JimmiG]

What's your stock vid?

Don't even remember. Think 1.15V or something like that.

 

[Kenmitch]

Don't even remember. Think 1.15V or something like that.

Seems high. Was that with the .1v avx spiking?

Are your core temps more consistant now?

I've decided againt delidding mine. I get 4.4ghz just under 1.2v's without throttling in Linpack 11 with the H100i. Just need to replace the fans with quieter ones. Using quiet profile is pretty quiet but temps will hit 80's with IBT or 90's.with Linpack 11.