Just delidded my i7-8700K

IEC

Super Moderator
Super Moderator
Jun 10, 2004
13,680
728
136
#1
Pre-op:

delidPRE.PNG

Post-op (-19°C...):

delidPOST.PNG

i7-8700K @ 4.7GHz core and cache, no AVX offset, -50mV core voltage offset

Process:
1) Delid with der8auer's DelidMate 2 tool - took seconds
2) Removed old TIM with 97% isopropyl
3) Shaved down/removed sealant (minimize gap) and wiped everything down with isopropyl again
4) Used tiny drop of Thermal Grizzly Conductonaut liquid metal TIM and spread on naked die with provided q-tip (extremely thin layer)
5) Used tiny drop of liquid metal on inside surface of IHS (again, extremely thin layer)
6) Super glue on 4 corners of IHS
7) Relid using provided DM2 alignment tool and clamp, do something else for 30+ minutes
8) Reinstall CPU, marvel at the temperature difference

tl;dr Intel's IHS assembly is trash. Get rid of the gap and re-TIM and see a world of difference.
 
Mar 10, 2004
28,522
238
126
#2
Intel probably has to have wider tolerances due to the number of CPUs they have to turn out.
 

IEC

Super Moderator
Super Moderator
Jun 10, 2004
13,680
728
136
#3
Per Silicon Lottery:
Peak core temperatures under a heavy overclocked load typically decrease anywhere from 15°C to 25°C when delidded for processors with PTIM (Intel 3rd through 8th generation)
My results are apparently quite typical.
 
Mar 10, 2004
28,522
238
126
#4
What % of processors are delidded?
 

Engr62

Senior member
May 31, 2001
789
4
81
#5
What cooler are you using?
 

firewolfsm

Golden Member
Oct 16, 2005
1,828
4
81
#6
You should flatten/polish the top of the IHS. That should give another ~5C drop.
 

IEC

Super Moderator
Super Moderator
Jun 10, 2004
13,680
728
136
#7
What cooler are you using?
Swiftech CPU waterblock with 360mm rad. Hydronaut for the TIM between waterblock and IHS.
 

aigomorla

Cases and Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
17,294
93
126
#8
overclocking boards should just give us a different retension mount so we dont need to reapply the IHS and can just run the die naked.

I bet you would of gotten even better gains, and much lower deltas from idle to load if you didnt have the chunk of metal between the die.
 
Jun 12, 2018
126
19
51
#9
overclocking boards should just give us a different retension mount so we dont need to reapply the IHS and can just run the die naked.

I bet you would of gotten even better gains, and much lower deltas from idle to load if you didnt have the chunk of metal between the die.
Definitely, as decent as the thermal transfer is, nothing there is better!
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
18,000
1,809
136
#10
I had a delidded Xeon (no lid), and I got some flat copper plate to make up the difference with paste on both sides. Worked pretty well.
 
Jun 12, 2018
126
19
51
#12
But then you'll crack the die...
I mean with proper mounting pressure you shouldn’t..... but.... yes the risk is higher...

Either way, going straight from die to cooler will have better thermals than having the IHS but it does give some tolerance to the setup...

If I ever get to a point when I can get a little more hardcore into thermal testing I would love to try direct die cooling, for now I have to protect my investments as I can’t afford to replace them if I should mess up...
 

aigomorla

Cases and Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
17,294
93
126
#13
But then you'll crack the die...
not if you had a shim around it to protect it like this:


sigh too bad msi only released it once, and never again..
They had a good product.

Of course idiots still cracked the die by using bowed blocks and bowed heat sinks.
You would have to lap the block and heatsink so it was dead flat.. and not bowed like how most come now.
 
Jun 12, 2018
126
19
51
#14
not if you had a shim around it to protect it like this:


sigh too bad msi only released it once, and never again..
They had a good product.

Of course idiots still cracked the die by using bowed blocks and bowed heat sinks.
You would have to lap the block and heatsink so it was dead flat.. and not bowed like how most come now.
Yeah was about to say that was a full on spider die lol
 

IEC

Super Moderator
Super Moderator
Jun 10, 2004
13,680
728
136
#15
Testing 5100MHz, no AVX offset @ 69°C load:

 
Mar 7, 2019
60
15
16
#16
Testing 5100MHz, no AVX offset @ 69°C load:

That is impressive! did you have to increase voltage? what about LLC adjustments to account for vdroop? I had an 8700K delidded, but I am a wuss because I had Silicon Lottery do it for me! but I could only do 5GHz stock voltage with AVX offset of 2.
 

IEC

Super Moderator
Super Moderator
Jun 10, 2004
13,680
728
136
#17
That is impressive! did you have to increase voltage? what about LLC adjustments to account for vdroop? I had an 8700K delidded, but I am a wuss because I had Silicon Lottery do it for me! but I could only do 5GHz stock voltage with AVX offset of 2.
It's a 50mV offset with LLC2. I was briefly able to test 5.2GHz AVX and even complete a Cinebench R20 run at around 1.36V effective, but it was not fully stable. So I suspect if I up the LLC to LLC1 and add a tiny bit more voltage I may be able to hit 5.2GHz. But first wanted to clear 24+ hours of testing at 5.1GHz so I have known good settings before pushing further.
 
Mar 23, 2019
33
6
16
www.hisevilness.com
#18
Did the same with the 8086k now it is running 5.3Ghz @ 1.41vcore under load it hits 85c llc7, wonder if you can post your vcore numbers for 5.0 to 5.* and llc I am just curious.
 

guskline

Diamond Member
Apr 17, 2006
5,340
65
126
#19
IEC, I run my torture test at least 5 minutes if not 10 on AIDA64 to assure that I'm hitting max temps.

Really impressive drop in your temps.
 

IEC

Super Moderator
Super Moderator
Jun 10, 2004
13,680
728
136
#20
Did the same with the 8086k now it is running 5.3Ghz @ 1.41vcore under load it hits 85c llc7, wonder if you can post your vcore numbers for 5.0 to 5.* and llc I am just curious.
5.1 requires around 1.33V AVX stable and 5.2 required nearly 1.38V+ with overshoot up to 1.42V. I wasn't comfortable with that so I backed off to 5.1 due to the exponentially higher requirements for 5.2 (on my sample). I use LLC2 (ASRock) as per my above posts.
 
Mar 23, 2019
33
6
16
www.hisevilness.com
#21
5.1 requires around 1.33V AVX stable and 5.2 required nearly 1.38V+ with overshoot up to 1.42V. I wasn't comfortable with that so I backed off to 5.1 due to the exponentially higher requirements for 5.2 (on my sample). I use LLC2 (ASRock) as per my above posts.
Very interesting those are good numbers for an 8700k the overshoot is a bit weird though on load my Vcore dips slightly.
 

B-Riz

Golden Member
Feb 15, 2011
1,178
270
136
#22
I just picked up an 8700K to play with until Zen2 drops, on the fence about de-lidding if I am going to probably sell in a few months.

Going to run it on a big air cooler, should I spend the $40 at Silicon Lottery? I have little desire to do it myself...
 
Mar 23, 2019
33
6
16
www.hisevilness.com
#23
I just picked up an 8700K to play with until Zen2 drops, on the fence about de-lidding if I am going to probably sell in a few months.

Going to run it on a big air cooler, should I spend the $40 at Silicon Lottery? I have little desire to do it myself...
Delidding is worth it, costs about the same to buy a delid tool, LM and good casket maker to seal.
 

Similar threads



ASK THE COMMUNITY

TRENDING THREADS