Why CPU Lapping isn't Very Worthwhile on Air/Water Cooling - Intel Core i9 12900K Lapping

If you want to improve the temperatures of your CPU when overclocking you can either think about delidding or improving the contact between the IHS and the cooling solution. Generally delidding can still help you by 5-10 degrees if the cooling solution you are using is very strong. Liquid metal is better than Indium solder between the IHS and the CPU Die, but the problem is that you can't use it below 0c or even below 5-10c. For extreme cooling such as dry ice, LN2 or phase change indium solder is the best thermal interface material (TIM) there is and it doesn't crack. On top of delidding you can also even sand the die which can improve the result even more and use a custom IHS.
So in my case delidding is out of the question as I do plan to try the CPU on LN2. The only thing I can do is try to improve the surface of the CPU to lower the temperatures. Generally the method is quite old and used for a long time. Back in the day like during 775 you could honestly gain a lot from lapping the CPU. Now it is quite pointless unless you lap the cooler as well because many PC Cooling manufacturers actually make the base of the cooling unit a little bit CONCAVE intentionally, because many of the modern CPUs have a convex IHS. So you can't gain much from lapping as the base of your cooler is not flat and actually it is designed for a little bit convex IHS.
Even if you lap the both surfaces you are likely looking at a gain of around 5 degrees at best, so even then it will still lose to delidding in terms of gains. I did this however quite successfully with Matador SiC sandpapers and lapped the CPU with up to 1200 grit sand papers. The key is to use as flat surface as possible for the process such as a glass panel or a mirror. Many youtubers have used just a wooden table or a mouse pad as lapping surface and that is not flat enough to do the process properly. I will go through the process quite in-depth on the video.
Support me on Patreon: / luumi
#overclocking #lapping #pcdiy

[@joshmerdm]

I agree with lapping both the CPU and the waterblock to see actual results. I tried lapping the waterblock first, only the 4 corners of the IHS made proper contact.

[@michaelthompson9798]

Just found your channel looking on KZbin….. I really enjoyed this video 🥰👍 I personally don’t have the means and equipment to get into extreme overclocking, but I still enjoy watching people having fun doing it 🤩🎉! I See from the below comments many people have differing opinions on how things should be done …But I guess like all of us, we understand what we have based on what we experienced or done ourselves so we’re probably all right and all wrong in different areas of the process. It’s great to see what people found works for them and share their ideas so we can all benefit when it comes to our time to try. Maybe one day I’ll try some of these techniques, but until then, I’m gonna go through Luumi’s back catalogue 🥰👍cheers buddy and keep up the great work 🏆

[@advanced8526]

Good day luumi thanks for all your videos. Hey have you ever tried conformal coating the pcb to waterproof it?

[@dezibeldani]

The edges are more lapped than the center because the pressure applied by the fingers is also at the edges. I made this experience with several 'massive' (unbendable?) metal objects. Because of an unintentional flaw in all of us: - When you move it away of your body, you apply more pressure with the thumb. - When you move it towards your body, you apply more pressure with the other fingers. The temperatures are better because the dye/dies is/are in the centre of the CPU and the centre is now higher than the edges. Together with the pressure of the screwed-on heat sink which helps to clamp down the hightend center, it is better, though.

[@poppyrider5541]

That's wrong. It's to do with the elastic of the paper rucking up on the leading edge. You could make a whole jig with rails and you'd get the same effect. The only way to stop it would be to use a hard stone.

[@dezibeldani]

@@poppyrider5541 You just say 'NO' to my explanation but you don't falsify it. You deliver a different explanation which could add to the effect. But it doesn't exclude other effects. The rucking of the sand paper happens in front of the moving CPU so the other edge on the side of the CPU would be lapped. Glueing the sand paper on the back with douple sided sticky tape would help. My explanation works also with solid files on metal and/or wood. If the pressure is not perfect evenly distributed and the file is not perfect parallel to the surface of the object, the result is a convex surface. One can't get a ruck in a solid file, can one?

[@poppyrider5541]

@@dezibeldani I'm going to disagree. Of course you can sand a dome by not paying attention. If you are trying to make it flat you still wont. It''s not about the pressure. Even with the uneven pressure of the human hand, done on a flat stone you would get far better result. Coz it's not the pressure, it's the paper.

[@dezibeldani]

@@poppyrider5541 If "it's the paper," why I get a dome with a file then? And why one get the abrasion on the flat side and not on the edge / small side of the CPU with paper? What type (material) of stone do you use? How thick is it?

[@tacticalcenter8658]

When lapping in professionally, they use flat metal dies and powder abrasives. This way you don't have paper that is not perfectly flat. It can also be automated through machinery in some instances to remove the human factor of finger/thumb pressure. Most sharpening stones are not flat or don't stay flat. Some diamond sharpening stones are flat but its overkill for copper. You could use glass or granite that's been perfectly flattened to .0001 and abrasive powder in a pinch

[@ClarkBark14]

My ihs was very very uneven from a tight socket clamp. Did the washer mod, lapped, applied kryonaut and dropped 20c in cinebench. Started to lap my AIO but it was already very flat. Huge gains from CPU lapping here.

[@TwiztedJugallo]

20C? I don't think so.

[@jonathanhorvat2452]

@@TwiztedJugallo what do you know of the affairs and specifics of others efforts? nothing..

[@kasuraga]

The last CPU i owned to be worth delidding at all was my 7700k at STOCK settings on a 240mm rad it was still on the edge of thermal throttling. Delidded, swapped on a pure copper IHS with liquid metal, temps are nice and chilly now.

[@radicalthunder5740]

And? Want a medal?

[@TrueRegulators]

side to side pressure differences are quite big this gen. have you checked how much the socket flexes on the DARK? on the Apex I get about 0.9mm z height delta from socket side to side

[@maxzett]

Would the condition at around 5:24 already void the warranty? Because technically you can still read the font and it might help if your CPUs IHS is like really uneven Edit: 8:02 looks really satisfying for some reason 8:21 I'd love to see this :D

[@crispysilicon]

Try thinning it as much as possible with the low grit? Nice spread!

[@theHardwareBench]

I was going up to 2000grit but I stop at 1200 now, I always lap the cooler too. 3 to 6 degrees is max but worth doing and much less risk than a delid. On ambient using thermal paste under the IHS (thermal grizzly kryonaut) you are only looking at about another 2 to 3 degrees after lapping. I'm going to do some tests using my 6 year old 8700K on an Apex board and chiller to see what differences I can get. I still have some decent HwBot points from when it was new so will be interesting to see if performance has degraded over time. It's had around 3000 hours of gaming plus another 1000 or so on top as it was part of my daily for 4 years. It benched X265 4K at 5.1ghz, if it can still do it on exactly the same setup I probably won't delid, I'll lap it though whatever the result.

[@christopherjackson2157]

I find I often get temp variances of a couple of degrees between re-mountings using the same hardware, same thermal paste etc Maybe it's my lack of experience lol

[@ramair325]

Using your fingers to lap is not the correct way to do this job, using a holder for the cpu with a jig would be ideal here. also glass is really only OK is not the worst but not the best. ideally you would use a machinist table made of granite that has been made TRUE. Also a hone would likely do a much better job of getting a True surface, how ever at this point is probably moot as it is already good enough. But using your fingers is imho a big mistake for a true surface finish. Was a nice video and is not terrible information for others. I do not mean harm and i thank you for the time to make these videos. i may not need them but i do enjoy them all the same. Best of luck in your OC adventures sir!

[@tacticalcenter8658]

Paper is also not idea. Powder abrasives and a lubricant is more idea. Heck let's build a macjinto do the whole process and spend a lot to do it.

[@ColdVenom159]

@@tacticalcenter8658 this is why you guys just speak on it. people such as luumi and myself actually do it ;)

[@tacticalcenter8658]

@@ColdVenom159 no offense, but you have no idea my background and what I do and have done.

[@ColdVenom159]

@@tacticalcenter8658 no offense, but i really dont care, again as i said this is why you guys are speaking on it while we actually do it. your own channel is more knife focused then anyhting.

[@kasuraga]

I used to draw an X with a sharpie on CPU's I lap once I hit the copper so I know when I hit true flat. I used to use 2000 grit sandpaper then switch to computer paper for a final polish

[@Zurvv]

What is your room temp? You CPU idle temps are lower than what it is in my home :) Is there a chiller on your loop? (or just an open window) :)

[@someguy9466]

My air cooler fan doesn't ramp up as aggressively after sanding my cpu and cooler. Steady state temps are only 1C lower though.

[@metalguitarman01]

so i couldnt help but notice the dark board there and i was wondering if you can say anything about it like you know release date perhaps?

[@battlephenom8508]

I found liquid metal on the IHS paired with a 360mm rad worked really good in my case. I don't even have a very good chip (SP 83) and I got 5.2ghz all P-core 4.1ghz E-core and 5.4ghz turbo on 1.39 volts with good temperatures. With that I was able to get over 30K in R23. Honestly, I could probably push it a little further too.

[@stonyspeedxd131]

What temps and did u overclock it to get those scores?

[@battlephenom8508]

@@stonyspeedxd131 In R23 on multiple runs not a single core hit 90c. The core temps hung around the mid 80's at the highest temps.

[@radicalthunder5740]

@@battlephenom8508 pfft

[@timtoomuch]

I know this video is old but you should get more of a gain using a contact frame getting rid of the stock mechanism

[@veniaminshows]

LUUMI FTW! Under zero water-chiller cooling I use gallium+indium liquid metal conductivity 120W/mK. Works up to max -30C before melting.

[@el_manosanta]

The best!

[@CR055FIRE]

am i wrong to consider just getting a direct die mount

[@socialistsuccubus822]

it depends on how bad your individual CPU is. I have a Ryzen 7 here that is definitely warped

[@vannak139]

So, I think lapping towards true flat is probably wrong on 12th gen, but still good on a cooler. I lapped my CPU and AIO flat, and only saw like 4c immediate drops. I took that apart and relapped it, focusing on sanding down the edges, while leaving a flat, raised, rounded square in the middle. This covers 90% of the short width, and 70% of the long width. I really sanded down all of those 90deg corners with the IHS walls. I got around 10c drops, am very happy. My main though is that if the cooler is contacting the IHS plate too closer to the IHS walls, under thermal expansion those walls expand directly into the cooler, and the cooler's pressure is going to act to compress those walls, which aren't going to budge. Also, thermal expansion, plus the CPU bracket, will make the plate concave on its own, I think. However, if the cooler is contacting the IHS plate further away from the outer wall of the IHS, then the cooler isn't trying to compress the IHS walls as much, and can act more on the flexible IHS plate and maintain better contact. IDK though, just my thoughts.

[@radicalthunder5740]

Neeeerd

[@jgvtc559]

So why not create a jig that only exposes said rim for lapping?

[@thefurrykiwi4489]

I ground down an fx4100 and its max temp dropped by 10C same stress test, same clock, same room temp. Were as for the 12700f it only dropped by 3C

[@theodanielwollff]

Is your WC setup below ambient? 76C for 350 watts is insane on water.

[@DSDSDS1235]

Does going to finer grid like 2000/3000 after following what you did achieve anything beyond a nice mirror finish? I am trying to lap my 3700x basically for fun, and I plan on doing 400 (remove nickel) -> 600 -> 800 -> 1200.

[@arturaspf2566]

I think its enough 400 -> 1000 -> 2000 -> 2500 :)

[@UnknownPerson-nl7te]

15:30 it clearly shows the IHS is a bit concave making the temps worse

[@TrueRegulators]

should you not do the lapping in an 8-pattern, so all edges are getting even pressure?

[@ubbgn]

yes!

[@tacticalcenter8658]

There are multiple ways to lap, this is just one way.

[@LuumiJuhani]

You can do whatever way you want to.

[@Apollo-Computers]

The figure 8 just helps the sanding go quicker. If you back and forth one direction, it creates tiny grooves, change direction it's no longer in grooves so it removes material better. I don't really know how to explain it that well. Just remember figure 8 is quicker.

[@ubbgn]

@@Apollo-Computers if u want an even surface, 8-pattern is the way to go!

[@Scouse29K]

Luumi are you certain you didn't get more gains than you thought? Your minimum temps where a few degrees lower than your test after lapping the chip

[@LuumiJuhani]

Not much anyways for water, as I said the waterblock is not flat so I could gain more if I lapped it.

[@ghostofdream]

how is lapping not useful for water i dropped 5 c avg on all cores and had all cores within 1 degrees of eachother

[@LuumiJuhani]

Depends on CPU and generation, with all of these latest generations I've had very minimal gains from lapping only the CPU IHS. Both surfaces would have to be lapped so the CPU and the cooler.

[@tacticalcenter8658]

@@LuumiJuhani yup, it all varies. Heck you could get a 12900k that's way more messed up than most of them due to manufacturing process variables that lapping might be necessary for. It all depends.

[@MrWhizeGuy]

My 3900x was concave so I gained alot, my friend tried the same thing tho I told him it's not worth it since his CPU wasn't concave. I gained around 6 - 8C, he didn't gain anything.

[@sldkjh]

I expect some gain from lapping a 5600X that I have. The block that I have has a mirror finish so I think it's flat, I will try to see if it really is. At the moment the CPU has a delta of almost exactly 6 degrees Celsius compared to the water when idle. If I lap the 5600X I will see if I can make a video about the lapping and the results and post them here. Wish me luck!

[@tacticalcenter8658]

@@sldkjh mirror finish does not mean flat. No cooler is made flat unless its for ln2.

[@UnknownPerson-nl7te]

Advice : Do this after the 3 year warranty to avoid violating warranty conditions *LOL*

[@MenCanNotBeWomen]

lapping voids warranty...same as deliding..

[@tacticalcenter8658]

I don't buy CPUs that are under warranty :)

[@redclaw72666]

Warranty is for normies

[@belmicah]

NOTE: You must be careful not to lap your fingertips...don't ask me why I know that

[@karamalinumberone7478]

nice video thank you very much i done it last week on my 9800GTX+ cooler Wich was a really rughf aluminuim surface its gived me 20 degree gain from 100c to 80c the aluminum surface on that cooler was so shit I guess LOL

[@MrHeHim]

It used to be very worth it years ago when there wasn't as high performance thermal paste / liquid metal. But when I got a 3700x and a NH-D15 i first tried it with nearly the best conventional thermal paste, took readings of the inlet surface temps etc then lapped the CPU and retested. I was expecting the normal 8-10c drop but only saw about 4-5c drop. Two things i think are happening, the actual CPU die is much smaller now so respectively the IHS is much bigger in a way thus being less of a bottleneck and the thermal paste was doing a good job of moving heat though the gaps in the uneven surface. Then used liquid metal, to which had just a ~2c drop in temps. Long->Short; Lapp the cpu/heatsink if you don't plan to use liquid metal 😉

[@radicalthunder5740]

Liquid metal isn't worth it

[@ayo0069]

igor’s lab documented socket 1700 mounting pressure being so high it is warping the cpu while mounting. your cpu is not flat and it will never be flat if lapped outside a socket.

[@LuumiJuhani]

Would need separate socket to do the lapping on.

[@ayo0069]

@@LuumiJuhani also heat/cool the spare socket to the temp you will be operating under.

[@lillen141]

Deliding is more risk then laping. Deliding is just worth it if you gonna overclock ALOT. Laping is much better for everyday use.

[@mikerzisu9508]

With the water and debris I don't know how the cpu survives this process

[@davem6027]

as long as there isn't any power, you can take a bath with your pc. just make sure its dry all the way before plugging it in. its not the water that damages it, its the short that the water causes to happen

[@kristoferstoll587]

Why don't you try this on a 12900K that isn't a SUPER BINNED chip? Like 5.5GHz and 300W+ and your temps are in the 70's?!?!? Like that is NOT normal! That is a extremely rare 12900k you have there. Why don't you try this on a 12900k that runs around 90C on water with a OC of 5.3GHz all core like most normal 12900k's?

[@btmedic04]

z690 dark?

[@GeorgeWPush]

debauer said delidding isnt worth it for alder lake.

[@-eMpTy-]

Only worth it if your CPU is really concave or convex. You should also consider a real lapping plate. With sandpaper you're just sanding, not lapping. Big difference.

[@ColdVenom159]

lapping is sanding lol. youre making a non flat surface flat by sanding away extra material hence lapping. you can use any surface like a glass panel or something along those lines long as its completely flat.

[@-eMpTy-]

@@ColdVenom159 kzbin.info/www/bejne/qYLaY2OofdujiNk

[@ColdVenom159]

@@-eMpTy- I guess you better go tell kingpin hes doing it wrong as well. i think you better go back to playing your videogames man. you might wanna look at the uploaders videos. hes not even in the field of computersl. guys gut completely random videos.

[@-eMpTy-]

@@ColdVenom159 He's not "doing it wrong" but he isn't lapping the CPU, he's sanding it.

[@ColdVenom159]

@@-eMpTy- are you dense? lapping is sanding. lapping is a term machinists use to say sanding. Example in the freight industry forklifts are called motors not fotklifts.

[@eilizezila4650]

the goal of cpu and waterblock lapping is to STOP using thermal grease. we use thermal grease because the surface area on both cpu and waterblock are bad.the thermal grease fill the air buble, but thermal grease isolate heat so you want to use up to 8000 grit sand paper, you want to have a mirror surface on both side. at this point, the cpu stick naturally on the waterblock, without thermal grease. but , it's really painfull to have this result you just need to look at the end of your video to understand why. it's difficult to have something absolutely plane when you sand something by hand

[@DyllonMcDonald68]

What about AMD 5000's.

[@kommandokodiak6025]

but now we know why this was wrong, get the tensioner system from derbauer and an untouched 12900k and repeat the testing

[@gamingevolved760]

Hi Luumi my I9 9900k is laapt and i get almost 10° less on my water cooler🤩🤗🥰is Working now 5.1 GHz on 1.33v on 420mm AIO.I Love Your movies👍🤩🤗🥰Try making more movies,from Your nolledge people learn lot 🥰🤗🤩Take Care Luumi🤩🤗🥰👍

[@MenCanNotBeWomen]

My 10850k runs at 4.8 peaks 5.2 really good for games.

[@maxzett]

Cries in Ryzen 3600 at a maximum of like 4,2 with PBO cause I've got such a bad chip that manually OCing isn't stable beyond 4,1 AC thus just using it in default with PBO that at least SC boosts to 4,215 actually gives me better results

[@hamzashaikh9310]

What do you think of using liquid metal with liquid nitrogen overclocking for gaming fps benchmarking. All these cinebench shit generate heat in a decentralized manner spread over many cores but gaming generates less heat but far more concentrated on like just 2 cores. And therefore those hotspots need to be cooled well as even under liquid nitrogen their temperature is high enough that liquid metal can help.

[@LuumiJuhani]

Well liquid metal has bad performance when going below zero.

[@hamzashaikh9310]

Have you done gaming tests to verify that the core with the heaviest load which is the most important core for gaming is actually getting below zero ?

[@Apollo-Computers]

@@hamzashaikh9310 liquid metal solidifies below ambient(not sure exact temp but like -20 or so), and when that happens it contracts and no longer filling in those imperfections of the contact surfaces. Thus bad for dry ice and ln2.

[@serithin4703]

you like to woid the warranty :P

[@Chris-ih4hj]

You’re supposed to do it until 3000

[@GamingKing545]

i lap all of my cpu's

[@JoneKone]

grind down properly not improperly take instructions from ppl who make lenses. Improper lapping/grinding just makes it tilt sideways.

[@vsRaiden]

It is very worthwhile lapping the cpu, if you lap your CPU mounted/socket-ed. The fact that you make it perfectly flat and shiny will not help you if the moment you installed it will be bent. And it will bend even if you use a custom ILM. + of course, you need to know if you cooler/cold plate is perfectly flat as well.

[@saravanantr2895]

he stops talking and starts doing it at 3.57

[@SixDasher]

Always do a figure 8 to avoid grooves from sanding in the same direction. Bad lapping technique.

[@LuumiJuhani]

Depends, with GPUs you can't do figure 8.

[@CYGAN19911]

you did it only with cpu ....?thats why its only 2 less i think