I can't delid my i5-4670k, with a razor at least...

[BallaTheFeared]

Maybe I should get a vice?

I've tried twice now, once when I got it, and once again this morning...


Both times I was unable to get my blade under the IHS. There is basically no separation for me to squeeze a blade in between; I keep pressing on the copper.

I tried for probably 20 minutes this time. I tried all the corners, as well as the slide flanges that are used by the socket to secure the chip.

I feel kind of stupid, what's the problem? I delidded a i5-2500k that was soldered easier than this...

 

[Kenmitch]

Are you trying straight in or at a slight angle? How thick is the blade?

Haven't tried on mine yet. Nerd vs wallet phase still.

Vise method looks scary to me. I'd probably hit it too hard.

 

[Jaydip]

What's your temp Balla?

 

[wilds]

That's weird. Careful, we wouldn't want you to kill your golden i5

 

[BallaTheFeared]

Are you trying straight in or at a slight angle? How thick is the blade?

Haven't tried on mine yet. Nerd vs wallet phase still.

Vise method looks scary to me. I'd probably hit it too hard.

Tried just about ever angle I could image that wouldn't gouge the PCB :hmm:

Using a standard utility knife blade, perhaps too big? Seen videos of the same type, with small kids capable of pushing right though glue... Sadness.

What's your temp Balla?

Depends on what I'm doing ^_^

Ran some Cinebench, HandBrake and a quick memory test to give you an idea.

Pretty much worse case for me since I mostly just game and internet.

 

[Jaydip]

I thought that temp is great for Haswell?

 

[BallaTheFeared]

Yeah it's not bad, but you know "It could be better"...

 

[Magic Carpet]

Yeah it's not bad, but you know "It could be better"...

Can't tame your temptations? :awe:

I'd order another chip, just in case.

 

[JimmiG]

Maybe your excellent temps without de-lid and the lack of separation are related? From what I've read, the big problem with Haswell isn't the thermal paste itself, but rather the large gap between the die and IHS, caused by the thick layer of epoxy or whatever they use to fasten the IHS.

If you're getting 78C at 4.9 GHz I really don't see what the point is. You're more likely to ruin the chip and have to get a new one that will likely be far worse.. Besides, it's obvious that thermal transfer is already excellent so you probably won't get the -20+C reduction that those with worse pre-delid temps are reporting.

 

[Kenmitch]

I wouldn't bother with those temps....I know the Nerd in us can be strong at times.

The thing that sucks about Haswell the most is the voltage control. I hate running at a fixed voltage but it seems like it's the only way to get a decent overclock out of them.

What happened to the additional turbo voltage setting as with previous generation boards....Would be nice to have the option to use a -0.1v or so offset to lower the spike under AVX loads.

I didn't play around much with auto and using a negative or small offset yet. Trying to get a feel of my chip still. Seems like once I think I've found it Gigabyte releases a new uEFI for my board....I just start the adventure all over each time I update the uEFI.

 

[BallaTheFeared]

Dang you guys reminded me of an old SNL skit..


http://www.funnyordie.com/videos/70bf1f8630/debbie-downer-lindsay-lohan-snl-sketch-full-of-outtakes

 

[wilds]

Can you upload a photo of the cpu? I can't see why you can't get a razor under there. Are you starting at a corner?

 

[Idontcare]

Maybe your excellent temps without de-lid and the lack of separation are related? From what I've read, the big problem with Haswell isn't the thermal paste itself, but rather the large gap between the die and IHS, caused by the thick layer of epoxy or whatever they use to fasten the IHS.

If you're getting 78C at 4.9 GHz I really don't see what the point is. You're more likely to ruin the chip and have to get a new one that will likely be far worse.. Besides, it's obvious that thermal transfer is already excellent so you probably won't get the -20+C reduction that those with worse pre-delid temps are reporting.

I bet they are related.

The lack of clearance between the PCB and the IHS where the blade needs to go will directly translate into just that much less gap between the silicon die and the underside of the IHS.

Even if Balla succeeds in delidding his CPU without destroying it, the benefits he'll get from delidding will be quite minimal because he already has a small(er) gap to start with.

Balla - if you look at my first delidding thread you'll see that I too had an issue getting the blade under the IHS. I had to use a small tack hammer to gently ease the blade into the interface. My second and third 3770k's weren't like that, the gap was much bigger.

 

[BallaTheFeared]

I'm out of LUP so I won't be able to get it out now to take a picture.


What kind of temp drop did you see with your first chip, IDC?

 

[Anth Seebel]

I dropped ~15C @ load when I did my 4770K using CL Liquid Pro which was about the same drop as my 3770Ks. Be careful of those resistors on the pcb , I covered them with a couple coats of nail enamel.

Thought I knicked the die when I did mine (focusing on the resistors too much), fortunately everything seems to work fine. But yeah, keep that in mind too.

But Kenmitch makes a good point, unless you really want better temps (for higher OC or for summer) then not worth the risk, seems like you got a healthy OC and temps anyway.

I would only do it if you are prepared to kill the chip. I used a siingle sided scrapper razor which is thinner than a stanley knife but thicker then a flexible razor. The scrapper razor is good cos it doesnt bend much (so you dont chip pcb). You need to keep blade angled away from the pcb and just rock it gently into the corner. The blade will slowly go through, dont rush it takes time.

 

[Idontcare]

I'm out of LUP so I won't be able to get it out now to take a picture.


What kind of temp drop did you see with your first chip, IDC?

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

But I didn't have great temps to begin with, so the temperature drop for me is as expected. But I had to use a hammer to get the blade between the PCB and the IHS, (see post #2 in the thread).

You have great temps and a silly tight seal, which makes me conclude the two are likely related in your case.

 

[BallaTheFeared]

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

But I didn't have great temps to begin with, so the temperature drop for me is as expected. But I had to use a hammer to get the blade between the PCB and the IHS, (see post #2 in the thread).

You have great temps and a silly tight seal, which makes me conclude the two are likely related in your case.

So your thought is good temps already, unlikely even with the best paste I'd see much of a difference?

I've had the chip for two months now, and this is the first time I've tried going 4.9GHz 24/7, for awhile I was at 4.6, then I went to 4.8...


Your previous posts about low temps > lower voltage > longer life are at play here, plus degradation at over 1.3v now I'm not fully comfortable with it knowing I might be able to get the temps down even more with my Dark Knight 2.


Plus everyone is doing it

 

[Aikouka]

I noticed a similar thing when I was going to delid mine, Balla. The gap between the IHS and the silicon at the corners was just so small that I wasn't sure how I would manage to get a razor in there! Although, I'm not really sure how good my temps are to begin with. On water, I get ~50C across all cores at stock speeds running Cinebench 11.5. What originally pushed me to consider delidding was that I would get really close to 50C when just at 20-30% processor usage, which is a typical amount to see while playing a game.

 

[Idontcare]

So your thought is good temps already, unlikely even with the best paste I'd see much of a difference?

I've had the chip for two months now, and this is the first time I've tried going 4.9GHz 24/7, for awhile I was at 4.6, then I went to 4.8...


Your previous posts about low temps > lower voltage > longer life are at play here, plus degradation at over 1.3v now I'm not fully comfortable with it knowing I might be able to get the temps down even more with my Dark Knight 2.


Plus everyone is doing it

Every degree cooler helps, but only if you don't kill it in the process of getting there.

I am sure you'd see at least, minimum, a 5C decrease in temps if you delidded and used CLU for the TIM.

But do you really need it? I question that. Are you going to use it for 5yrs?

 

[sm625]

Are you guys saying that there may be a way to select a processor based on its IHS to pcb gap? If so then wouldnt the total height of the chip be measurable and variable? Can we simply select the chip with the lowest possbile height to get the best OC results?

Also, what would happen if you stuck the chip in a vise and tightened it? (So as to reduce the gap between IHS and die...)

 

[Aikouka]

Are you guys saying that there may be a way to select a processor based on its IHS to pcb gap? If so then wouldnt the total height of the chip be measurable and variable? Can we simply select the chip with the lowest possbile height to get the best OC results?

CPUs aren't exactly like the nuts and bolts section at Home Depot as in you can't just go and toss them around looking for the one you want.

Also, what would happen if you stuck the chip in a vise and tightened it? (So as to reduce the gap between IHS and die...)

The excess epoxy still exists in between the IHS and the silicon. Also, I'm not too keen on putting pressure on that part of the CPU. The vise method of delidding only applies vise pressure to the IHS.

 

[Essence_of_War]

Balla,

I don't suppose you have access to calipers do you? I'd be interested to see if you could compare with others exactly how much tighter your seal is.

 

[BallaTheFeared]

Are you going to use it for 5yrs?

No, I update every two years generally.

Balla,

I don't suppose you have access to calipers do you? I'd be interested to see if you could compare with others exactly how much tighter your seal is.

Nope

Visually it looks like it's very tight, I wasn't able to get in between the two using just forearm strength. There was unquestionably far more gap with my i5-2500k that was quite easy to slide in between.

 

[Idontcare]

Are you guys saying that there may be a way to select a processor based on its IHS to pcb gap? If so then wouldnt the total height of the chip be measurable and variable? Can we simply select the chip with the lowest possbile height to get the best OC results?

Also, what would happen if you stuck the chip in a vise and tightened it? (So as to reduce the gap between IHS and die...)

I measured my three 3770k's with a micrometer for total thickness (PCB+adhesive+IHS) before delidding, then correlated the thickness after delidding (PCB+IHS) and the temperature reduction I observed and it was essentially linearly correlated.

I meant to post that graph at some point in the past but things got lost in the move, so it would be quite a project just to locate the data and the graph at this point.

But the bottom line is that you are correct IMO. If one had a supply of chips and a micrometer, measuring the thickness of the PCB+IHS at stock would be a great proxy for setting one's expectations of baseline OC performance without delidding.

 

[john3850]

If chip that was going to be delid was warmed up to 150f would the extra heat help to soften up intels black glue paste.
I would also round off or break off the point on the razor then polish it smooth on a buffer wheel to be safe.