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

Page 43 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

kalamar182

Junior Member
Oct 29, 2013
3
0
0
Welcome to the forums kalamar182!



My 3770K's (three of them) could get to 5GHz stable for 24/7 operation, but I never left them there because the voltages were way too high for my comfort level.

They run comfortably at 4.7-4.8GHz.

Delidding doesn't really seem to bring much in terms of increased overclocking potential, but it does dramatically lower the operating temperatures which will substantially increase the lifespan of your already over-stressed CPU (because it is overclocked/over-volted).


Great... i am follow your steps rigth now, im traying to figure it out how to separate 2mm from the cpu water block and the cpu i have a rasa cpu waterblok. Thatsthe only thing thats stopme. I.m excited to se the my temps. I oc my i7 to 4.8 stable with 93 max temps i really need to get low that temps.

Sorry but can you sendme the link of the complete tutorial??? I am only with my phone and ints not easy seee sow many pages and comments.

Thanks and again....sorry my english
 

kalamar182

Junior Member
Oct 29, 2013
3
0
0
Welcome to the forums kalamar182!



My 3770K's (three of them) could get to 5GHz stable for 24/7 operation, but I never left them there because the voltages were way too high for my comfort level.

They run comfortably at 4.7-4.8GHz.

Delidding doesn't really seem to bring much in terms of increased overclocking potential, but it does dramatically lower the operating temperatures which will substantially increase the lifespan of your already over-stressed CPU (because it is overclocked/over-volted).



OKKKKKKK....
Oc 4.8 / 1.32 vcore / 50 times intel burn test... 67° max temp....
 

Questionario

Junior Member
Nov 8, 2013
4
0
0
Hi IDC,
I read through the whole thread... phew...
Very very nice work explaining everything!
I do have a few questions to you before I start de-lidding my Haswell i5-4670k:
I plan to lap both IHS and Air cooler (Scythe Ninja 3) and delid the HW and use CLU on all 4 sides. I do not plan on taking it apart all the time but every 2 years or so I do like to remove the parts to do some dusting in the PC. after successful delidding I never plan to remove the IHS again.
I dont do this to gain 100-200mhz more, I just plan to do it because I think it will be a fun and rewarding experience (spent weeks on absolutely silencing my last computer with 500rpm fans and special hdd cases I had custom-made), I am fine with the reward of lower temperatures.
1.) do you know if the space between IHS and PCB is the same as on IB?
2.) You mention that you dont recommend any kind of shim between IHS and PCB, not even for protection as the PCB bends and this will most likely create a gap between die and IHS. I do believe you but I would still like to make sure the IHS will not come loose from the die, I know the die and the IHS will eventually bond a little but will that suffice? I do want to be able to take the CPU and be able (dont plan to but...) to send it via mail without it falling apart. What would you recommend? no glue at all, some kind of silicon/silicon glue and then the pressure from the retention bracket as well as the air cooler?
3.) do you consider it far more dangerous to delid haswell with the blade you used than with IB as the die is closer to the edges and more stuff is close to the edges of the IHS that can be damaged by the blade?
4.) any reason why I should lap before or after delidding?
5.) Ive never done any of this before, anything I have to know for lapping? I would just start with 400 (then 1000,2000,3000), put the sandpaper flat on a table and start moving the IHS back and forth with 90degrees rotations as you suggested, not sure yet on the cooler plate as I dont think I would move the whole cooler? would I just mount the sandpaper on a piece of wood and start lapping?

Thanks! :)
 

TreVader

Platinum Member
Oct 28, 2013
2,057
2
0
Hammertimeforrealz.jpg

Wow. Idk how sharp the razor blade is and idk how much your pushing on the hammer but that pic makes it look like you are inches away from at best a large amount of stitches.



Looks like it turned out ok tho?
 

PliotronX

Diamond Member
Oct 17, 1999
8,883
107
106
UOTE=TreVader;35712077]Wow. Idk how sharp the razor blade is and idk how much your pushing on the hammer but that pic makes it look like you are inches away from at best a large amount of stitches.



Looks like it turned out ok tho?[/QUOTE]

Unsafe and those boxcutter blades are thick and can bias towards the PCB andnick it. DDouble sided blades that are used in shavers are paper thin and what I have used for Intelchips since 775 with the protruding lip. Never a nick.
 

Questionario

Junior Member
Nov 8, 2013
4
0
0
Hi IDC,
I read through the whole thread... phew...
Very very nice work explaining everything!
I do have a few questions to you before I start de-lidding my Haswell i5-4670k:
I plan to lap both IHS and Air cooler (Scythe Ninja 3) and delid the HW and use CLU on all 4 sides. I do not plan on taking it apart all the time but every 2 years or so I do like to remove the parts to do some dusting in the PC. after successful delidding I never plan to remove the IHS again.
I dont do this to gain 100-200mhz more, I just plan to do it because I think it will be a fun and rewarding experience (spent weeks on absolutely silencing my last computer with 500rpm fans and special hdd cases I had custom-made), I am fine with the reward of lower temperatures.
1.) do you know if the space between IHS and PCB is the same as on IB?
2.) You mention that you dont recommend any kind of shim between IHS and PCB, not even for protection as the PCB bends and this will most likely create a gap between die and IHS. I do believe you but I would still like to make sure the IHS will not come loose from the die, I know the die and the IHS will eventually bond a little but will that suffice? I do want to be able to take the CPU and be able (dont plan to but...) to send it via mail without it falling apart. What would you recommend? no glue at all, some kind of silicon/silicon glue and then the pressure from the retention bracket as well as the air cooler?
3.) do you consider it far more dangerous to delid haswell with the blade you used than with IB as the die is closer to the edges and more stuff is close to the edges of the IHS that can be damaged by the blade?
4.) any reason why I should lap before or after delidding?
5.) Ive never done any of this before, anything I have to know for lapping? I would just start with 400 (then 1000,2000,3000), put the sandpaper flat on a table and start moving the IHS back and forth with 90degrees rotations as you suggested, not sure yet on the cooler plate as I dont think I would move the whole cooler? would I just mount the sandpaper on a piece of wood and start lapping?

Thanks! :)

Looks like I caught a bad time to post my questions for IDC, hope he didn't leave for good.
Does anyone else have the knowledge to answer my questions?
I plan to first lap the HS, then delid, remove the intel thermal compound, clean with 99%iso then use CLU and use a black automotive silicone glue (for high temperatures) to glue the lid back on, I assume with the pressure on the IHS when I mount the CPU back on, the glue should be quite thin, thin enough not to create a gap (i hope).
For delidding I plan to use an actual razor blade as they are much thinner, lets see if I can apply enough pressure with it or if I have to use that kitchen cleaner knife that IDC used.
in my local paint store I only found paper with grit 220,400,1000,2000,3000. Is it worth it to look through more stores to find a 600grit and 1500grit? (I was able to find 2500grit but I assume that wouldnt be necessary)

Also I am not sure what the "standard" procedure for measuring temperature before and after is so it gets somehow comparable.

Thankful for all advice! :)


Q
 

Questionario

Junior Member
Nov 8, 2013
4
0
0
hmm,
Idontcare seems to have stopped using this forum :p
as a result of that the thread died?
if anyone reads this and has an answer to my questions... :)
 

Tristor

Senior member
Jul 25, 2007
314
0
71
hmm,
Idontcare seems to have stopped using this forum :p
as a result of that the thread died?
if anyone reads this and has an answer to my questions... :)


Maybe I can help.

1) No, the gap is slightly different between IB and HW. I think HW is less gapped than IB in a fairly significant way. IB benefits more from delid than HW because of this, but both benefit strongly from the delid.

2) Keep the plastic clamshell that the CPU comes in, if you need to mail it, put it in the clamshell. It helps protect the CPU from damage and should be done delid or not, but it will help hold it together if it's delidded. Also, do not add any shims or adhesive. If you put the lid back on, lap it first and then just apply CLU/CLP/IC7 to the die and then set the lid gently on top. Once you've done this, socket the CPU immediately and use gentle even pressure in the center on the lid when you close the retaining mechanism of the socket to keep the lid properly centered and to achieve even pressure across the die.

3) Yes, I absolutely would not use a blade to delid anything, but certainly not HW. HW has surface mount components on the PCB that are very sensitive to damage and will cause the CPU to be nonfunctional with even the lightest nicks. Use the vise method to delid and take extreme care in removing the black adhesive (buffing with a dremel is preferred to a razor blade).

4) It's pointless to delid if you are not also going to lap your IHS or do direct-to-die watercooling. Lapping the IHS is mandatory if you're putting it back on top post delid. Consider for a moment, lapping is a significantly less risky modification to the CPU than delidding is, if you've already sunk the risk into delidding, doing lapping in addition does not add any additional risks and it's functionally required to get the best temps, truly even mounting pressure, and reduce core to core delta.

5) See my post here for some photo tutorials and a video showing lapping methods:
http://forums.anandtech.com/showpost.php?p=36022929&postcount=38

You may wish to read that thread since it provides some insights into the process from my own experiences.

As to using adhesive to reapply the lid. Do NOT do this. If this is your plan, you might as well not delid at all. The adhesive is what causes the gap, the gap is the primary reason for post-delid temperature drops. Possibly shitty TIM on Haswell aside, there's seriously no point in doing this if you're going to introduce an expanding sealant back into the equation. You really do not need any sort of shim or adhesive for this to work...

And if you're delidding you should definitely do so FIRST, before lapping the HS and IHS, as then you aren't handling the sensitive electronic components while lapping and have no risk of introducing copper slurry into the electronics. Basically delidding is more risky than lapping. Once you've delidded, lapping is no risk. Pre-delid, lapping has some risk.


Hope this helps.
 

BonzaiDuck

Lifer
Jun 30, 2004
15,699
1,448
126
hmm,
Idontcare seems to have stopped using this forum :p
as a result of that the thread died?
if anyone reads this and has an answer to my questions... :)

We're beginning to wonder ourselves why IDC hasn't returned since curfew.

You don't need to go beyond 600-grit wet-or-dri sandpaper for lapping.

I had thought myself about "silly-cone" sealant/adhesive, and somewhere earlier in this or another thread, we discussed the active chemicals in the sealant before it dries. I'd research first what it could do to the PCB, but I have no reason to think it would do anything. You might instead want to look at an adhesive like "Pit-Crew," but make sure you seat the IHS and apply some pressure to it before it sets. It is still rubbery, but sets up with more firmness than silicon sealant.

The reason I also had been interested in "adhesive re-attachment" is the minor risk of moving the IHS around when you put it in the socket, maybe a possibility of bent pins and so forth.

Others with firsthand experience may provide better guidance on this issue.
 

crashtech

Lifer
Jan 4, 2013
10,521
2,111
146
The problem with re-adhering the IHS is it then becomes very difficult to check on the integrity of the aftermarket TIM which has been placed between the die and IHS by the user. I note that many of these products dry out over time and must be re-applied. The factory Intel TIM, while not the best conductor of heat, certainly has been designed to remain stable over time.
 

james1701

Golden Member
Sep 14, 2007
1,873
59
91
Oh wow, I just seen where he hasn't been here since November 2nd of last year.

Anyone know where he is positing now, or if he's ok?
 

Questionario

Junior Member
Nov 8, 2013
4
0
0
nope, he doesnt respond to personal messages either (which I believe are emailed to you) so I hope he just decided not to waste his time online anymore as that would be the best reason...

about the adhesive, I have some black silicone based adhesive which is used for automobiles and can withstand a lot of heat, I have tried it on a PCB and it doesnt seem to damage it (from just plain looking at it after removing it again), it also seems to be able to spread super thing if put under pressure and still hold together well (it can be much thinner than what I can see at the HW now, from the side I cannot even tell its there, it is black and can be spread so thin that it becomes see-through but still black/dark).

I think once I have time I will try it with the adhesive just to hold the IHS in place =)
PS: I dont think it would work for glueing it before putting it into the PC unless you find a way to safely apply quite a bit of pressure for a longer amount of time to the IHS without damaging anything.

@Crashtech: I guess I could always delid again if I need to check on it?

Thanks everyone for your responses! =)
 
Last edited:

Howard

Lifer
Oct 14, 1999
47,989
10
81
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.
A thin layer of silicone rubber of the appropriate durometer would suffice. The temperature resistance is adequate and it can be squeezed like a spring. If additional deflection is needed, one can instead use a foamed silicone rubber instead of solid.
 

chris89

Member
Dec 28, 2010
36
0
66
It will actually run cool without that part on the cpu. If you do a bit of testing, it yields in the neighborhood of 20C at times. Since the diode is directly connected to the heatsink so thermal conductivity is much greater this way. It's actually more like a laptop cpu, as the diode is directly connected to the heatsink, as to yield the best possible thermal conductivity transfer.

The diode is very fragile so the clamping force must be less, which will yield a much better temperature.

I need to do this to my i7 930, and To be Xeon X5675... :D

Thumbs up to the OP BRO!!!! My i7 930 atm is 130W TDP at 4.2Ghz which is 1.4Ghz overclock. However I like the idea of 4.4Ghz yielding 1.6ghz overclock per core. Doing this to my CPU would be epic! My temps are very nice since I lapped the cpu down to pure flat copper, and did the same to the cooler. CPU temp reports only 65-75C at 4.2Ghz and 1.475v vcore and qpi.

However once I pull 4.4Ghz the vcore must be increased to crazy levels like 1.6v vcore and qpi which gets very hot.

What are your temps specifically on all cores and the cpu reported temp?

Another thing is I'd like to compare the i7 930 at 4.4ghz to yours at 4.7ghz. When i'm running at 4.4Ghz my memory read/write/copy is right at the 30GB/s mark. I ran AIDA64 CPU Queen and pulled off a crazy 50,631 points but idk if it is that crazy. I wonder if yours is faster? If so I'll be very jealious and will surely pick up the Intel Xeon X5680 since it's 50% faster by stock than my overclocked i7 930. Overclocked to 5Ghz is potentially 80% faster than the i7 930... :D

th_50631CPUQUEEN_zps4391e9c5.jpg
 
Last edited:

Zardnok

Senior member
Sep 21, 2004
670
0
76
It will actually run cool without that part on the cpu. If you do a bit of testing, it yields in the neighborhood of 20C at times. Since the diode is directly connected to the heatsink so thermal conductivity is much greater this way. It's actually more like a laptop cpu, as the diode is directly connected to the heatsink, as to yield the best possible thermal conductivity transfer.

The diode is very fragile so the clamping force must be less, which will yield a much better temperature.

I need to do this to my i7 930, and To be Xeon X5675... :D

Thumbs up to the OP BRO!!!! My i7 930 atm is 130W TDP at 4.2Ghz which is 1.4Ghz overclock. However I like the idea of 4.4Ghz yielding 1.6ghz overclock per core. Doing this to my CPU would be epic! My temps are very nice since I lapped the cpu down to pure flat copper, and did the same to the cooler. CPU temp reports only 65-75C at 4.2Ghz and 1.475v vcore and qpi.

However once I pull 4.4Ghz the vcore must be increased to crazy levels like 1.6v vcore and qpi which gets very hot.

What are your temps specifically on all cores and the cpu reported temp?

Another thing is I'd like to compare the i7 930 at 4.4ghz to yours at 4.7ghz. When i'm running at 4.4Ghz my memory read/write/copy is right at the 30GB/s mark. I ran AIDA64 CPU Queen and pulled off a crazy 50,631 points but idk if it is that crazy. I wonder if yours is faster? If so I'll be very jealious and will surely pick up the Intel Xeon X5680 since it's 50% faster by stock than my overclocked i7 930. Overclocked to 5Ghz is potentially 80% faster than the i7 930... :D

th_50631CPUQUEEN_zps4391e9c5.jpg

WARNING!!! Do not delid your i7 930. The heatspreader is soldered to the die and you will damage something if you try. Delidding is only for Ivy-Bridge and beyond processors. Previous processors were soldered and did not need the fix.
 

sxr7171

Diamond Member
Jun 21, 2002
5,079
40
91
to play devils advocate, having /\/\/\/\/\/\/\/\ as in with an unlapped surface vs __________ with lapped would increase the surface area, allowing for greater thermal dissipation....if you could get the thermal compound into those grooves...no?


This is not true. TIM is less conductive than metal itself. However it's much better than air which is poor conductor of heat.
 

MongGrel

Lifer
Dec 3, 2013
38,751
3,068
121
Still will never do it though been following it for years.

Seems almost a moot point these days really.

But I shouldn't be raining on parades I guess.

GL
 

StoatWarbler

Junior Member
Aug 28, 2008
3
0
0
Is there a reason you haven't tried direct die contact?

Bear in mind that a lot of current heatsinks has a slighty convex bottom. Couple that with the contact pressures these days and you'd probably crack the die unless the HS is lapped.

(On the subject of lapping - there is such a thing as "too smooth". Those mirror-like surfaces you often see are actually slightly worse for thermal conductivity than slightly roughened ones, once the thermal paste is in there.
 

ehume

Golden Member
Nov 6, 2009
1,511
73
91
I don't know if you've seen this yet. It's a video from Umberto Fiorentino, over in Italy. Very Clear. Very thorough.
 

Idontcare

Elite Member
Oct 10, 1999
21,118
58
91
I don't know if you've seen this yet. It's a video from Umberto Fiorentino, over in Italy. Very Clear. Very thorough.

Wow! That is awesome! Sure beats the old way of doing it with a blade rolling the dice that you are going to cut up the PCB somewhere.

I like how they sealed it all up in the end so you can't tell visually that it had ever been delidded.

I hope that isn't done for warranty fraud reasons, otherwise I like the cleaned up look and no doubt the chip runs 20C cooler because of it.

So sad that we enthusiasts can spend the extra $10 (retail) to make it happen for our $300 chips but Intel can't spend the extra $3-$4 (am guessing that would be their wholesale cost) to make it happen to the chips before they sell them to us.

All the "K" chips should be soldered IMO.
 
Feb 25, 2011
16,777
1,466
126
Wow! That is awesome! Sure beats the old way of doing it with a blade rolling the dice that you are going to cut up the PCB somewhere.

I like how they sealed it all up in the end so you can't tell visually that it had ever been delidded.

I hope that isn't done for warranty fraud reasons, otherwise I like the cleaned up look and no doubt the chip runs 20C cooler because of it.

So sad that we enthusiasts can spend the extra $10 (retail) to make it happen for our $300 chips but Intel can't spend the extra $3-$4 (am guessing that would be their wholesale cost) to make it happen to the chips before they sell them to us.

All the "K" chips should be soldered IMO.

Correct me if I'm wrong, but I'd kind of assumed that the chips are packaged before they're sorted into K and non-K versions and labelled as such.

Do they get binned as bare silicon?
 

jpiniero

Lifer
Oct 1, 2010
14,510
5,159
136
I imagine doing the K models with solder and the rest with something else would mess with their thermal validation schemes.
 

ehume

Golden Member
Nov 6, 2009
1,511
73
91
Wow! That is awesome! Sure beats the old way of doing it with a blade rolling the dice that you are going to cut up the PCB somewhere.

I like how they sealed it all up in the end so you can't tell visually that it had ever been delidded.

I hope that isn't done for warranty fraud reasons, otherwise I like the cleaned up look and no doubt the chip runs 20C cooler because of it.

So sad that we enthusiasts can spend the extra $10 (retail) to make it happen for our $300 chips but Intel can't spend the extra $3-$4 (am guessing that would be their wholesale cost) to make it happen to the chips before they sell them to us.

All the "K" chips should be soldered IMO.

I wonder if they wouldn't mind squeezing maybe another $50 from us by soldering the chip and calling it an 'i7 4770x' or an 'i7 4970x' -- with '-x' chips soldered, with the '-k' chips with TIM?