CPU die lapping

[cleric_warlock]

I've been watching several die lapping videos by LTT and drbauer and I've been wondering how exactly you know how much material you can safely remove from the die without actually damaging any of the integrated circuits in the die itself. Do you just look for guidance from the extreme overclockers who have sacrificed some cpus to determine how far you can go or is there some spec somewhere that will tell you how much of the die isn't occupied by any integrated circuits?

[AnonymousGuy]

In this video it looks like they're using a fixture to guide them to the correct depth:

 

 

When the fixture starts getting sanded on then they stop.  There really isn't a "safe" amount to take off because that was already done at the fab in a process called backgrinding....which uses a diamond slurry deal.  Sandpaper is extremely crude.

[Somerandomtechyboi]

I prefer trial by fire

Murder some cheapo celeron/athlon before lapping the cpu you wanna lap, ofc the dummy cpu gotta be from the same generation

 

if someones already lapped their die and can prove how much you can take off before the cpu dies then you can skip the trial by fire

 

 

As for the question of is it worth it?

Depends on how much effort you wanna put in, most ppl will say not worth it cause theyre average joes that wouldnt bother doing stuff like this anyways either out of fear, lazyness, or both

 

If i wasnt already gonna delid my cpu i wouldnt bother either, but if im already gonna delid my cpu i might aswell go and die lap cause i may aswell maximize temp drop

[OhYou_]

you start by googling how thick the silicone substrate is for your specific cpu family, next you take a starting measurement, then you sand it and take more measurements to see how much you removed and how much is left. 

 

[SorryBella]

3 hours ago, Somerandomtechyboi said:

I prefer trial by fire

Still easier said than done and its only getting you to learn even pressure application. wood blocks and sandpapers are cheaper on that. You really need to nail the thickness as @OhYou_ stated, otherwise youre going to be out with dead daily driver. I wouldnt play doctor with it if it is your main desktop, especially on a pretty rare 12900KS. Getting contact frame in is like 60% of the battle anyway.

[cleric_warlock]

3 minutes ago, SorryClaire said:

Still easier said than done and its only getting you to learn even pressure application. wood blocks and sandpapers are cheaper on that. You really need to nail the thickness as @OhYou_ stated, otherwise youre going to be out with dead daily driver. I wouldnt play doctor with it if it is your main desktop, especially on a pretty rare 12900KS. Getting contact frame in is like 60% of the battle anyway.

I'm not actually considering die lapping for any of my current devices, the build that i would use that for would use direct die water cooling and is probably a few years down the road. Just want to learn more about the process.

[AnonymousGuy]

3 hours ago, Somerandomtechyboi said:

Depends on how much effort you wanna put in, most ppl will say not worth it cause theyre average joes that wouldnt bother doing stuff like this anyways either out of fear, lazyness, or both

It's not about how hard you are it's about recognizing when something is pointless to do.  Congrats you decrease your peak temp by 2 degrees.  What did that buy you?  You're not going to be able to step up the voltage any and even if you could, every chip for the last several years has been speed limited before it's temperature limited.   We're not even getting into the high likelihood that you're going to kill the silicon either immediately or down the road after it's thermal cycled a few times.

 

The only place this might be interesting is if you're doing it to a top end thermally limited chip like the new w9's.  But good luck blowing $5k to probably fuck it up.