Separate names with a comma.
Discussion in 'Overclocking & Cooling' started by erek, Apr 24, 2018.
any proponents of lapping the cpu die for better thermals after deliddin'?
It wont offer better thermals if it doesn't work lol.. lapping the substrate just mean you will be literally cutting parts of the Silicon.. that's why even smallest scratch just end in damaged CPUs....
I did it to my 7700K during my de-lid, re-lid. I also did the lid itself. Everything worked out fine and CPU overclocks extremely well!
Should not be a problem if you use extremely high grit material.
I don't think you're going to make the die any flatter or smoother by hand than it came from the angstrom-tolerance flat and smooth silicon wafer from which it was cut.
The smoothest surface is the mirror finish that the die already has the metal bits can be flatter and closest to mirror you get the better. So unless your aim is to make a keychain don't lap the die.
This won't end well.
You could have a very high paying job in a number of different industries if you had the ability to eyeball and lap silicon flatter than with the resistance etching and CMP techniques they use today.
If you do, make sure you sand it all the way down to the green substrate. Gotta get all that pesky silicon out of the way...
There is a protective coating on the silicon that people with too much time and money can realistically remove to eliminate 0.05mm of thermal insulation, but that's about it...
Now this is odd. I have never hard of a non flat die. While it may be possible to do this without destroying the cpu, I doubt any benefit could be realized from it. However lapping many coolers bases and even a couple of cpu lids has brought tangible benefits in the past. I have a 3770 delid that I flattened the cooler base and the top of the lid. The interior of the lid was fine. The top of the lid was high in the middle slightly. Got that razor flat. Actually watched a video of Kyle explaining different cooler seats because of non flat bases and he was showing sanding them down flat. Worked out fine. That cpu is still going at 4.5ghz.
If someone told me they lapped their CPU's exposed die, I'd respond that they have an unhealthy obsession.
That said, if it makes you feel like Evel Knievel jumping Snake River Canyon, more power to ya.
It's gonna make you feel like Evel Knievel destroying his expensive processor.
"lapping processor died"
I wonder how many people noticed that the pic isn't just the die(s) ground down, but the entire IHS as well. That was a whole CPU that was just held down against a belt sander or something lolol
Probably used a CNC machine. I could do it at my shop.
How did you delid ? It looks all scratched...
The die was lapped
Further updates on this thread: hehe
What kind of cooler are you using? Did you lap the cooler as well?
Everything gets lapped. IHS discarded tho
did you lap the heatsink for the "stock" test?
Should lap the bottom side too. Make sure its even all around.
[H]ard as f**k
Yeah they would lap the outer part. the "lid" But not the actual silicon.... This guy just ground down the lid and everything.
It wasn't lapping the die, people would lap the IHS, delid the processor, remove the gunk they put inbetween the silicone and the IHS, put their own gunk in that was better, then relid and attach the lapped heat sink. This often produced significant results. But they did not typically lap the actual die.
You can find all kinds of links, but lapping the die was not common. Typically what people ended up doing is exactly what people today are doing, delidding, removing the crap the manufacturer put on, replacing with something better, lapping the IHS or directly mounting the heatsink onto the die. You can still find a number of articles from back in the day about this process. One of the reasons lapping the actual die became a thing is there were certain processors that just had a ton of gunk on the die that seemed to be put on halfhazardly. People would sand the gunk off. This was especially popular with the Celeron 300A, one of the best OC'd chips of all time.
Just wanted to chime in a re-iterate that Billy the Dimwitt is 100% right. Actual CPU die lapping was a real thing back in the day on the "Flip Chip" based CPU manufacturing process both Intel and AMD used starting back in the slot-1 and slot-a days..
Also as an FYI: I have lapped the CPU die on a couple 7600K's I have (even noted this much higher up in this thread). There is nothing wrong with lapping the die; the closer you get to the actual heat source (rather than hot silicon) the better your temps (obviously). How much it matters is debatable, but the 7600K's I have done this to run 24/7 and at 5GHz without issue.
Doesn't sound worth the risk, except if you love doing it I guess
My 7600k runs very well at 5.2Ghz 24/7, and its "only" delidded (with lapped IHS)
Lower temps means obviously more headroom for voltage, but a crappy chip is a crappy chip
Wouldn't do all that work without knowing it's a good chip
If it's a good chip I don't think I'd risk ruining it; shrugs
Is there an actual reference of how much you could even remove without grinding into something important?
Yes it wasn't always soldered very evenly so people would lap the the top of the Celeron and remove the gunk (solder). I am not sure what your point here is, except that they weren't typically lapping the die. My point is that die lapping was never popular, it was never "a thing". It was something people rarely did because it typically did not produce enough results to be worth the risk, where simply removing whatever gunk the manufacturer used on the die was far more worth it and produced far better results.
yup a couple oddballs doesn't make it "a thing"...
I saw that the writing on the die transferred to the copper of the alpha pal so i lapped until the writing was all gone from the die .
Relapped the pal = 1.150 GIGAhertz !
Wanted to add one thing before I start.. I have not even tried going over 5.0GHz on the 7600k's I have, but I will do so when I have time (in the middle of moving half way across the country)... And by 24 hours stable, I am talking running max heat on Prime95, not just the system operational.
As to how far to go before its "too far" on the die, on the original flip chip based processors, you would see a copper layer under the silicon right before you would hit actual die components. I dont know if the current gen chips are set up the same way or not. On my 7600k's, I did not go very far; I mainly wanted to remove any residual byproducts vs. much of anything else. I do have another 7600k here I could maybe use as a guinea pig to see how far is too far...
I lap all the way to my desk. IMO this gives the best thermalz.
I do have to say that using a thermal paste that is using gallium (Liquid metal ones on the DIE) would be best for thermals, except if you want to cool bare DIE, which is difficult for the best of enthusiasts (have to modify the board, Kyle tried several times until he got it right)
If you can see copper you might run into problems with the best of pastes as those are conductive (liquid metal ones), the ones that follow are capacitive not good for "exposed" copper either
Gotta be careful not actually exposing anything that conducts electricity
If you use the IHS (hopefully lapped) then using a paste that isn't LM would probably give worse temps then not lapping the die and using LM
Wasn't very happy with my IHS, it's not a lot I gained, the cooler base should be convex after all
But it was something (hottest core dropped a few degrees)
. . .
an excellent point! Evel Knievel never succesfully jumped the Snake River Canyon but
he did successfully break every bone in his body except for the tiny trio in the inner ear