9900KF 5.4, lapped die, direct die, no avx offset, no hyperthreading, evga 390 dark

Joined
Apr 22, 2011
Messages
714
Thought to share since this is hard forum. Haven't ran it longer than just over 5min, don't know if I will cause not a realistic use case for me. Pain in the rear to lap a die, my first time, took .3mm off the top. Not a pro here by any means, just an enthusiast having fun. Scores 1612 on C15, single core was 229.

5.4 Direct Die to Lapped Die .3mm off.jpg
451760_Cinebench_at_5.4-1.jpg
 
Last edited:

Nasgul

Weaksauce
Joined
Jun 11, 2005
Messages
96
Bought a tool to remove the IHS for my 7700K thinking I was going to use some liquid metal, never got around that, still got the tool laying around and also a bunch of sandpaper. Ultimately I bought parts for a loop for my 9900KF but only got it up to 5.1GHz before it crapped out.
Now I moved my Loop to a 10700KF and got up to 5.2GHz.....I'm sure it can go higher, but the voltages are set to "adaptive", so the software has control and I'm not gonna fiddle with it, 5.2GHz is perfect.
.
Looks like too much work for a couple of hundred MHz. Nevertheless, I bet is fun to do when you have the time dedicated to that.
 

cdabc123

2[H]4U
Joined
Jun 21, 2016
Messages
3,134
Why lap a die?? The silicone is normally pretty dang flat and smooth.

(This is from someone who shot water under the ihs over the die to cool the chip)

I'm genuinely curious
 
Joined
Apr 22, 2011
Messages
714
Why lap a die?? The silicone is normally pretty dang flat and smooth.

(This is from someone who shot water under the ihs over the die to cool the chip)

I'm genuinely curious

Better temps. It allowed for a stable 5.4ghz overclock with an ambiant water loop. Previously temps were 70's, now usually 50's, a big difference when it comes to overclocking.

Main purpose is not to make it flatter, though you wanna keep it flat, it's to remove excess silicon material and significantly increase thermal conductivity from the top of the die to the waterblock in this case.

Never heard of direct water cooling a die but I'd like to hear more, how did it turn out, seems like a great way to directly water cool a die, I think you're onto something.

We need a direct die water cooling block, maybe just insert the chip and it has a gasket to seal as the water runs directly over the die, just seal the small leads on top of the chip with liquid tape.
 
Last edited:
Joined
Jan 16, 2013
Messages
2,200
Better temps. It allowed for a stable 5.4ghz overclock with an ambiant water loop. Previously temps were 70's, now usually 50's, a big difference when it comes to overclocking.

Main purpose is not to make it flatter, though you wanna keep it flat, it's to remove excess silicon material and significantly increase thermal conductivity from the top of the die to the waterblock in this case.

Never heard of direct water cooling a die but I'd like to hear more, how did it turn out, seems like a great way to directly water cool a die, I think you're onto something.

We need a direct die water cooling block, maybe just insert the chip and it has a gasket to seal as the water runs directly over the die, just seal the small leads on top of the chip with liquid tape.
You got a 20C drop in temps from lapping the die?
 
Joined
Apr 22, 2011
Messages
714
Yeah I can see the delid doing some serious temp drops, but never heard of 10+ C drop from lapping the die.

Surprisingly there was a greater reduction from lapping .3mm of material off the die, go-figure.

Keep in mind most (including derbauer) only take .2 off, this one went beyond, not many willing to be this [H]ard;)
 
Last edited:
Joined
Jan 16, 2013
Messages
2,200
Surprisingly there was a greater reduction from lapping .3mm of material off the die, go-figure.

Keep in mind most (including derbauer) only take .2 off, this one went beyond, not many willing to be this [H]ard;)
Having a hard time believing this. .1mm gets you 10C+ lower temps?
 

JSHamlet234

Weaksauce
Joined
Apr 9, 2021
Messages
66
I tried to find out the thermal conductivity of silicon, and I got a lot of different answers, but it seems like 100-150 w/mk is a reasonable estimate. That's not super low, but given how small the die is relative to the amount of heat dissipated, I can see how a 0.3mm thick layer of excess silicon could hold back a lot of heat. It probably wouldn't make nearly as much of a difference on an older CPU.
 

JSHamlet234

Weaksauce
Joined
Apr 9, 2021
Messages
66
Just to follow up, I did a little more digging, and the material we are talking about removing is probably magnesium silicon nitride, which conducts much worse than straight silicon. The thermal conductivity of standard MgSiN2 is only 25 w/mk, but there are new manufacturing techniques that can increase this 4-fold, and there are other materials such as β-Si3N4 that can be used in it's place that have conductivity as high as 110w/mk. If Intel is still using the regular stuff, and something tells me they probably are, then I can see this being a very beneficial thing to do.

I sort of want to try this on my delidded 4690K for fun/practice, but it's probably not worth it for that chip. Even though delidding dropped the temps by 15 degrees, it barely improved OC headroom at all (0.02V at the same speed). The same goes for my G3258. I guess if you're running into a thermal wall, it's worth it.
 
Top