- Intel improved the thermal performance of the Core i9-10900K by actually reducing the thickness of the CPU die itself,
improving thermal transfer. But as crazy as it sounds, enthusiasts had already
tried this technique on the previous-generation 9900K with a degree of success. Pun intended. But this kinda tomfoolery
doesn't come without risk. I mean, I'm talking delidding the CPU, lapping down the integrated heat spreader, lapping the silicon die itself to improve its flatness
and thermal conductivity and putting it all back together, sans warranty, of course. For our attempt, we're gonna use ha ha, this! Two tools from nudeCNC. that will hopefully give us
the precision that we need to get the performance we crave without ruining our chip in the process. You know what I never
ruin, though? My segues. To sponsors like Ridge Wallet. Ridge Wallet wants to redefine the wallet with its compact frame
and RFID-blocking plates. Check out how they can help
keep your wallet bulge down, and use offer code LINUS to save 10% and get free worldwide shipping. (upbeat music) We get into it in more detail here, but in a nutshell,
keeping heat under control makes a processor last
longer and/or run faster by allowing you to increase the voltage and overclock it. Now, as for a why a very
small reduction in thickness can help thermal transfer a lot, the reason is that silicon is much, much, much less
thermally conductive than copper. So the more of our distance
that we need to cover to get to our large
heat-dissipating heat sink can be done through copper,
the better off we are. Now, let's have a look at the tools that we're going to be using today. - Yeah, we got a lot of 'em. It's going to take this
whole gamut of tools. We have calipers. We needed
to precisely measure our IHS. Allen keys. That's actually
the right one, you did good. We got Thermal Grizzly Conductonaut. Paper, this is just for figuring out how close the IHS is to the die. Silicone for resealing the IHS to the CPU. Scraper. We gotta scrape
off the old stuff. - Lots of scrapers.
- Lots of scrapers. All the scrapers. Cotton
swabs, cleaning tools. All this Noctua stuff-
- Der8auer delidding tool! Der8auer's a cool guy,
go check out his channel. - Very cool dude.
- But first, before we can get to any of that we need to get a baseline measurement to tell us if what we've
done, at the end of all of it, was actually effective. What we're looking at
here is our temperatures as recorded by the on-CPU sensors. What we're gonna do to
get a baseline reading is we're going to allow it to heat up until it reaches a steady state, which it already did about 10 minutes ago. Then we're gonna take five
minutes of data and average it. So the ones that we are most interested in are both our hottest core as well as our CPU package temp. So both of those are
sitting in the neighborhood of around 81, 82 degrees. - So have you delidded a CPU before? - Yeah. - 'Cause I haven't. - Well, thing is, I haven't
done one of these soldered ones. So Intel has been resorting to more and more elaborate schemes to bring the temperatures
of their CPUs down over the last couple of generations. And with the 10 series, of
course, it was thinning the die. But with the 9 series they
finally went to a soldered IHS, which means that instead of
just using a delidding tool to kinda (slurps) slide the IHS off with a little bit of force, we actually need to soften the solder between the die and the
integrated heat spreader in order to perform that removal. That I haven't done, because the solder performs so much better than
the previous thermal compound that I didn't honestly
really see the point. It's better to clean that
off once you've got it out of the socket, 'cause that way you won't
accidentally lift it up and mangle any pins. - Oh, okay. Fair enough. - There's a free tech tip for ya. Everybody gets one.
- Aw, look at that! I get free ones? - I even pay you to give them to you. If you think about it. I see there's a heat gun. That should be good.
- There is a heat gun. We're gonna give 'er some beans, and then hopefully it'll help- - [Linus] Are we just basically following Roman's guide on this? - Pretty much.
- Good. - [Colin] Yeah. - [Linus] Now we gotta heat
this puppy up I guess, eh? - Yeah, it's set at 150-ish. - Yeah, try 150 to start. - If I apply too much force
without it being melted enough we will break it and
this video will be over. - I saw a guy do it cold, but
he had to really reef on it. - [Linus] What temp are we at now? - [Colin] 190. - That's a full half-turn
that we've done there now. - So it's an M6 screw, it's an M6x1, so one turn is one millimeter. I think we can just give 'er... - Okay. There we go. - [Colin] Hey, look at that. - [Linus] That actually looks
like we got it off clean. Wow, that's comin' off real easy. - [Colin] Well, it's still warm. It's definitely not
perfect, but it's something. - [Linus] And actually, most
of the stuff on the underside of the IHS came off pretty well too. - [David] Is it normal that
my anxiety's really high watching you guys do this? - Yes.
- Yeah. Okay. That's probably off enough. Now I just gotta remove the silicone. Not to be confused with
silicon. We need the silicon. This CPU is super convenient
to remove the silicone from 'cause there's no surface
mount components up here, so you can just go ham with
a plastic razor blade here. This looks sketchy as heck, but this is actually a
very safe way to do this. We are pretty darn close
to this being stripped. I think we're good to go. - So we need the NLap.
This is the IHS one. - [Linus] Okay. - [Colin] So we need the die. And we're gonna take this
little piece of Blu Tack. So we want the Blu Tack to
sit completely in this pocket. We don't want any Blu Tack around the actual contacting surface of the tool. - And then what? You just like... - Yeah, pretty much. I brought this from my home stash. This is a granite surface plate. It's not super high quality,
but it's flat and consistent. So we're gonna use this as
our base reference surface. And our first step is
we're gonna take some 220- - Did you just say 220-grit? - Yeah. So our first step is we need to create our zero reference, so as we start it won't actually be touching the die. - Oh, I see. So we actually start by rubbing away the
outside plastic here... - Yep. - And then that tells
us through these steps how far we've gone.
- Exactly. - [Linus] Genius. - Now we're getting
into the first bar here, which means we're
approaching our zero plane. Because this bar is larger
than this lower bar, that means that as I'm lapping I'm applying more force
to the lower portion, it's wearing away here first. The beauty of these bars
is that I can now adjust and watch from the top, these bars, and make sure that the lapping is going on very flat across the entire plane, I'm not tilting it onto
one corner or the other. - Ain't that interesting? We're approaching 0.2 millimeters now, we can tell from the red marks, but from our initial go with the 220-grit we've actually still got a little bit that we haven't touched again. Should we go farther? Should we go to 0.3? - More power. Is that a cracked die? - Wait, what? Let's not go any farther. - Okay. - Now we have a problem to solve. Because we wore away part of the die, we need to wear away part of
the outer boundary of the IHS to make sure that they're sitting right against each other. This is real... - (moans) This is too real, man. - Train wreck. Can't look away. That ain't a good sign. - Sure ain't. - Is it possible that because the IHS
isn't contacting the core it's not holding it down? - [Colin] Oh, yeah, got no pressure. - Well, there's some.
- There's some. - But it's not even. Okay. This is among the crazier
things I've ever done, but I'm gonna hold it in the socket. I'll let you know if it's gettin' hot. Ow! Quick! Off, off, off, off, please! Off! Okay. It's hot. - (laughing) Did you just do that? - What? Did I just do what? - [Colin] Are you burned? - What? That really hurt. - Okay. - So let's sand the IHS. - But, like, okay. Mm-hmm. - Right, because there was no
heat when we tried it before. - Yeah-
- That means no contact. - Okay. We're gonna reduce it by the same amount. So went two bars on that guy, we're gonna go two bars on this guy. - Okay. - It's a much more even
wear pattern on this guy, but we haven't even touched the IHS. - [Linus] Wait, yes we have. - Have we?
- Yeah, that's totally it. - Oh! I thought there was a gap
right here. I can feel it. - Oh, it looks like it. And there's no silicone left on it. Do you want some water on that? - Not really, we're not at the point. - [Linus] Okay. - We're not getting any
clogging of the abrasive. - 'Cause we have water. We keep it in LTTStore.com water bottles. - Really? - Yeah, you gotta go check
it out. LTTStore.com. - I'd say that's a pretty even four bars. - [Linus] That actually looks pretty good. - Thank you.
- All right. - Sweet.
- Wow, I think we're good. I think we're ready to go. - [Colin] Okay. (fan whining) - [David] That's not a good noise. (Colin sighs) - What a day. This one confirmed dead. - Toast. (clucks) - The second one- - [Colin] Ah, it's over here. - We had a little bit of a- - I had a little bit of a day.
- We had a moment. We had a moment. Let's call it a moment. Colin removed the IHS without testing the before temperatures. It's fine. It's not dead. We can just put it back together. We haven't sanded anything. - We just don't have a baseline. - We just don't have a
baseline reading, and- - Yep.
- as we discussed, you unfortunately can't
take your baseline reading from one CPU and apply it to another one because part of the whole
point of this exercise is that, due to the
manufacturing tolerance issues around things as, you would think, precise as the die height, you get a lot of variance
from one chip to another. So we went to MemEx and bought yet another 9900K.
- Number three. - We're doing this for you. This one has now been lapped
2 millimeters on the die- - 0.2 millimeters. - Excuse me, 0.2, yeah. (snorts) - Big difference. - The IHS has been lapped 0.4 millimeters. We've got a bit of a bigger
gap, but I told Colin I think it's probably okay. - By gap we mean between
the IHS and the PCB. - Yeah. So what we're gonna do now is we're gonna liquid metal this puppy. We're not gonna try and
do like a rando test without it properly glued on, so Colin's gonna glue it on with silicone. And we are finally ready
to find out if this works. Fortunately no cracks. - Don't wanna use too much. Less is more. According to Linus. Oof, this stuff reeks. (Colin groans) So now we should have, if I did my job right, perfect contact between the liquid metal. Now we wait. We'll do that magic wipe
thing to three hours from now. - Moment of truth. What are you thinking? - Well, I think my silicon job, or silicone job, was a little bit messy. We might have to clean up a little bit. - So we had a little bit
of silicone ooze out, but all in all I would actually say yes, Colin did a good job of
gluing this back together. You can see we've got a
little bit of a gap there, which is exactly what we would expect. And we've got what looks like an identically-sized
gap on the other side. As long as shaving down the
die didn't kill the CPU, I have a pretty high degree
of confidence right now. - Big if. - Think of it this way, Colin.
- Hmm? - [Linus] You can't make an omelet without breakin' some eggs. The goal wasn't to make an
omelet, but...(chuckles) - Contact.
- Okay. There's POST codes! That's a good sign! It's a lot better than 00 POST code, which means zero CPU detected. Whoa, hey! I saw... Yes! (Colin chuckles) You have to say now- - If it's gonna be better?
- if it's gonna be better. - I think it's gonna be
within margin of error. - I didn't see how you
applied the liquid metal, but I'm gonna say you were
sure to consult with Alex, you did a great job and
it's gonna be better. - I did it exactly how you did it. - Okay.
- So as little as possible. Just wet it down. - Okay. - So I ran the last bench for 12 minutes, so that gives it two minutes
to kinda acclimatize. - And what was the average last time? - I think it was 92, 91? - Okay, so if we're under 90, that basically tells us- - We did it.
- The system worked. Yeah, we did it, Reddit. In case it wasn't obvious
by the point in the video, even if this works it's not something that we would recommend. - [Colin] Not even a little. - Okay, so our per-core temps all look- - 34.
- very normal. - [Colin] Yeah. But they're
actually really even. They're within three degrees. - [Linus] Well, that looks very promising. - [Colin] Very promising. - I'm not gonna celebrate yet. - Yet. We'll, I'm celebrating
because it's working. - It's not going up any more. I think it might've worked. It doesn't take that long for
an air cooler like this to... What is it? Reach heat soak or whatever? What would you call it? Equilibrium. Oh, we should check our clocks. Oh, 4.8. Okay. So we're turboin',
boys, we're turboin'! - [Colin] Yep. I mean, 88, 87, that's a four-degree drop with liquid metal and a lap. So I wonder what that would
equate to in overclocking. - Very little. Don't be a downer. (laughs) - I'm not a downer, I'm a realist. Well. Data. - Wow. So red is good. - Red is good. - This is great! So this is a five-degree
gap here, so as expected, we are three, two, four degrees better over the time of our run. - [Colin] Yep. - [Linus] Huh! - [Colin] And it did soak up a little bit. I mean, that's like a degree, so up here we're probably closer to three degrees. - But it worked! - It did. It did work. - And theoretically, the, ah...(laughs) Sorry, I'm just...(laughing) And theoretically, the flatter you lap it and the better your contacts,
the more it will work. And if you went another 0.1 millimeter you might even get another-
- Degree. - fraction of a degree, or a degree. Again, not recommending it. But... Darn it, Roman, you were right. And I guess you've already
got your validation because Intel went and did exactly this on their next-generation CPU. But this was fun anyway. - Fun. - Speaking of fun, you're gonna have fun if you check out our sponsor, Micro Center! Micro Center is open to supply your work or learn-from-home technology needs. The Maingear Element 3
laptop is available now for $1,999 on Amazon or
$1,899 on Micro Center. It's got an Intel Core
i7-10875H processor, it's got an Nvidia GeForce RTX2070 Super, 32 gigs of DDR4 memory, a terabyte solid-state drive and a 17.3-inch, narrow
bezel, 240-hertz display. Check out this and other
Micro Center offers over at the link in the video description. If you guys enjoyed this
video, maybe you would enjoy... Oh, how 'bout checking out
our liquid metal adventure where we put liquid metal in a phone. That phone eventually died. - [David] Spoiler. - I'm fine. Look. - Do you need first aid? - I'm not burned. Yet. Actually, this one's a little red. - We do need to fill out a report later. - Why? - Because-
- I'm not burned. Look, look, see? Look. - But even if you weren't
injured it would be a near-miss. - I mean, I could've taken
my hand off it at any time. - We're gonna fill out a report later. (Linus sighs)
