The CPU Intel SHOULD have made!

Linus Tech Tips ·Linus Tech Tips ·2019-05-06 · 1,986 words · ~9 min read
Floatplane YouTube

Transcript

JSON SRT VTT 225
0:00 Intel's x86 architecture was made so popular by the original IBM PC that 40 years later,
0:07 40 years after the original 8086 CPU, we are still using it. So it's pretty important. But
0:14 when Team Blue decided to celebrate that anniversary with what basically amounted to
0:21 an upclocked and upmarket Core i7-8700K, we were a little underwhelmed. I mean,
0:30 I have been asking Intel for years to do something truly special with their really high-end stuff.
0:37 And like, look at this. It's got the same cores, the same IHS, the same green substrate,
0:43 the same non-soldered thermal interface material. So we figured if Intel's not going to do it,
0:50 then we're going to do it. We're going to build the 8086K that should have been,
0:57 and we're going to do it in style.
1:08 Before we could do anything, we needed to make sure that our CPU was performing the way that we would expect,
1:14 because we'll be comparing our end results later on to make sure that we didn't make it worse.
1:19 With that out of the way, it's time to shut it all down, pull the CPU, and delit it. We've delitted
1:24 plenty of processors before, but this time will be slightly different, because now we need to lap it,
1:31 which for the uninitiated means removing the upper layer of nickel plating and leaving behind a smooth,
1:37 flat, and smooth CPU.
1:37 It's a flat copper finish for us to play with.
1:40 Two hours, and two bloodied thumbs later, Anthony eased off the manual work to finalize
1:46 the design that we'll be etching into the top of the IHS. We weren't sure at this stage
1:50 whether or not the resolution would be too fine to etch into the size of a CPU, but once
1:56 we finished with the design, we decided it was time to practice our gold plating with
2:01 a dead CPU. Extra hours of lapping not shown.
2:05 So this is our lapped heat spreader.
2:06 Got this in there.
2:07 Got that exposed copper going on there.
2:09 I guess the smoother this is to start with, the less gold we would have to waste applying
2:15 it thicker and then polishing it down, right?
2:16 Right.
2:17 Okay.
2:18 What we need to do is we need to wrap this around like this, so that there's no metal
2:22 actually showing.
2:23 Okay, so what is this?
2:25 That's to absorb the liquid.
2:27 Okay, and none can be exposed at all?
2:29 It can be exposed, but don't do it too tightly, because the fluid needs to be able to flow
2:34 with the electricity.
2:35 Okay, so this hook is just going to hold our...
2:39 Our piece then?
2:40 No.
2:41 Oh.
2:42 We're going to brush it on.
2:43 We brush it on?
2:44 Yeah.
2:45 So our gold solution is actually clear.
2:48 Dip that in there.
2:50 That gauze will actually clear up, and once it's fully soaked...
2:54 You just brush the gold on.
2:55 Yep.
2:58 They say to do it as if you're petting a cat.
3:01 After a while, we should start to see some yellowing.
3:03 So it paints on a pretty thin layer, eh?
3:06 Yeah, it's basically microscopic.
3:08 Oh, it is starting to yellow a little bit.
3:10 It's really subtle.
3:12 Wow, it's hard to...
3:15 It's hard to tell, because it happens so slowly.
3:18 Brush plating really wasn't what I was expecting when I had said, hey guys, let's make a gold-plated
3:24 CPU, largely because I had actually never heard of it, but it ended up working out great.
3:31 Check out the difference from the start.
3:35 Now applying conformal coating to the PCB or substrate that's going to be nearby anywhere
3:40 you're going to be using liquid metals.
3:41 Pretty standard operating procedure.
3:43 So we've already done that all over the green part here, but the unusual thing we're doing
3:48 today is we're going to plasti-dip a CPU.
3:52 So we're going to go ahead and cover up the dye, and that's just for thermal performance
3:57 reasons.
3:58 We wouldn't actually be harming it by plasti-dipping it.
4:00 Then we're going to go ham.
4:03 Oh, Lordy, I put that on a little thick.
4:09 Oh, boy, oh, boy.
4:13 The adhesion to the conformal coating is not ideal.
4:15 And I wish I'd gone on a little lighter, but what's done is done now.
4:22 So now that we've got a smooth, gold-plated IHS, you're probably thinking to yourself,
4:27 well, hey, almost home free, right?
4:29 Throw it in your guys' laser cutting slash engraving machine.
4:33 I mean, wow, those test runs of the pattern look great.
4:38 And boom, it's off to the races.
4:40 But unfortunately, because we don't have like a solid state laser or what's it called?
4:44 I don't know.
4:45 I don't know.
4:46 What's the other one?
4:47 Fiber optic.
4:48 Fiber optic laser.
4:49 And for that matter, like almost no one around here does because they're really expensive.
4:54 We're going to need a different plan.
4:57 So we're going to do it the old fashioned way.
4:59 It's off to the jewelry store.
5:01 We're going to engrave it.
5:02 That sounds like a plan.
5:03 Anthony's going to engrave it.
5:05 I'll find someone.
5:06 I heard he knows a guy.
5:08 While Anthony gets our practice piece over to the engraver, it's time to gold plate our
5:12 real IHS.
5:13 So he took a turn with the electroplating kit.
5:16 Anthony got back after a bit of polish and some touch ups here and there and a second
5:20 trip to the jeweler.
5:21 I think the end result speaks for itself.
5:24 So something we didn't account for in the original plan was the extra thickness that
5:30 would be added by the plasti dip.
5:32 So a quick extra step we had to add was just sanding down the bottom of the IHS a little
5:37 bit to account for that clearance because it's the glue that we're going to apply around
5:42 the edges when we relit it, that's going to hold it on.
5:45 So that's sort of important.
5:46 So now all that's left is to peel this off, exposing our die.
5:53 Liquid metal the die.
5:54 No, no, no, no, no, no, no, no.
6:00 And that's more like it.
6:01 It's so chilly in here that it's not staying very liquidy.
6:05 I'm going to wipe some up.
6:08 I just want to err on the side of a lighter rather than a heavier application here.
6:15 I think that's pretty close actually.
6:22 Let's put our CPU back together.
6:27 Oh.
6:28 That was spot on.
6:29 Nailed it.
6:30 What?
6:31 That should be tight enough.
6:32 Okay.
6:36 Now we leave that for about an hour.
6:37 All right.
6:38 We'll be back.
6:39 This is it.
6:40 This is we think the limited edition CPU Intel should have made.
6:53 Black PCB, gold spreader, custom engraving.
6:58 And should we make sure it still works?
7:00 Probably a good idea.
7:01 That is so cool.
7:03 If you're into that sort of thing anyway.
7:04 So we want to enable XMP and disable multi-core enhancement.
7:06 Oh, it looks like...
7:07 Yeah.
7:08 So just F10.
7:09 Cool.
7:10 27 degrees.
7:11 Looks good.
7:12 Oh, cool.
7:14 So the IHS is lapped.
7:15 We've got liquid metal on it.
7:16 Theoretically, it should be better than before in terms of its temperatures, but we also
7:17 engraved it.
7:18 And we don't know if that's going to cause micro ridges around the engraving or anything
7:19 like that.
7:20 There's a very good chance it could.
7:21 And we don't know if that's going to cause micro ridges around the engraving or anything
7:22 like that.
7:23 There's a very good chance it could.
7:24 And we don't know if that's going to cause micro ridges around the engraving or anything
7:25 like that.
7:26 There's a very good chance it could.
7:35 And we applied a new layer of metal on top of it that obviously wasn't lapped.
7:40 So we don't know exactly how flat it is.
7:43 So we're hoping for the best, but we won't know for sure until we actually test it.
7:47 We're using an air cooler, so we don't have to wait around for the coolant to reach equilibrium.
7:51 Max.
7:52 So 64 to 75 degrees.
7:56 67 to 71.
7:59 Does that sound reasonable?
8:00 Yes.
8:01 Look at that.
8:03 We had a spike and then it settled in.
8:05 Yeah, because that was in turbo.
8:07 Now it's no longer turboing.
8:08 If we look up here.
8:09 Oh, okay.
8:10 But was that the same way you tested it last time?
8:13 Did you take your temps with turbo or without?
8:16 Without.
8:17 Without?
8:18 So sustained.
8:19 Oh, so then we're killing it.
8:20 Oh.
8:21 Yeah.
8:22 We're at like 62 degrees.
8:23 Yeah.
8:24 We're not finished our before and after testing yet though.
8:26 Now this CPU's thermals were not good enough before we liquid metalled it to really do
8:30 any overclocking.
8:32 So all we could do was turn on multi-core enhancement and then, where's this guy?
8:37 And then max out these power limits here.
8:39 So 4095 here, 127 here, and 4095 here.
8:46 So now we can have a look at what kind of a performance improvement we can get now that
8:51 our true limited edition, very limited edition, one of a kind.
8:55 And CPU can be fully unfettered.
9:00 So let's hit it again.
9:01 We're running at 4.3 right now, 4.3, 68 degrees on the hottest core.
9:10 Wow.
9:12 Dang.
9:15 That max turbo though.
9:16 We could just run that all day.
9:18 As it turned out, we didn't gain much in gaming, but when it came to blender, we saw a tangible
9:24 improvement in performance thanks to our better thermals, which were much better than stock.
9:28 When we removed the turbo, we got a much better performance.
9:29 When we removed the power limit, we hit the same temperature, but our gold plated CPU
9:35 remained at five gigahertz on all cores with a Noctua NHU 12S.
9:40 With a beefier cooler, we could do five gigahertz all day long.
9:45 So even after running for 10 minutes, our thermal results are still looking great, which
9:48 means the last before and after is a good old fashioned Cinebench run while we wait
9:52 for that to go.
9:54 I should probably say like, in fairness to Intel, there's a lot of stuff that you don't
9:59 have to change in the manufacturing pipeline in order to do something like gold plated heat
10:04 spreader.
10:05 Like the kind of validation a company like this does.
10:07 I mean, let alone turning the substrate black, they'd have to do all kinds of materials testing
10:12 and science and stuff in order to do that.
10:14 So I'm not giving them too hard of a time, but I also do think they could do better.
10:19 And that's the purpose of this video is, hey, come on guys.
10:23 If we can do it, then I think a multi-billion dollar company can probably do it too.
10:28 So our final result is...
10:29 1390.
10:30 1386.
10:31 1386.
10:32 So we have lost no performance.
10:37 We have better thermal results and presumably we'd be able to overclock this damn thing
10:41 now.
10:42 Probably at this point.
10:43 Freaking A.
10:44 So thanks for watching guys.
10:45 If you liked this video, you can hit that button.
10:47 I mean, if you disliked it, I guess there's the other one too.
10:49 If you really liked it though, get subscribed or check out where to buy the stuff that we
10:52 featured at the link in the video description.
10:56 Also in the video description is our merch store, which has cool shirts like this one,
10:58 and our community.
10:59 I'll see you in the next video.
10:59 Bye.