Teardown of a $250,000 1991 vintage IBM 9121 TCM Processor module from a System/390 ES9000 Enterprise Server mainframe computer,.
Amazing state of the art 63 layer ceramic hybrid module construction with 2772 pins!
Part 2 X-RAY! https://www.youtube.com/watch?v=xa0mieJHM94
And Dave demonstrates an amazing BGA desoldering technique that can only be done on these ceramic hybrid PCB's.
TCM PDF: https://eevblog.com/files/ibm-system390-air-cooled-alumina-thermal-conduction-module.pdf
Another teardown with an amazing cutaway view! https://www.youtube.com/watch?v=s7lVfOi7su4
Forum: https://www.eevblog.com/forum/blog/eevblog-1341-amazing-$250-000-ibm-processor-teardown/
Subscribe on Library: https://lbry.tv/ @eevblog:7
EEVblog Web Site: http://www.eevblog.com
The 2nd EEVblog Channel: http://www.youtube.com/EEVblog2
EEVdiscover: https://www.youtube.com/eevdiscover
Support the EEVblog through Patreon! http://www.patreon.com/eevblog
AliExpress Affiliate: http://s.click.aliexpress.com/e/c2LRpe8g
Buy anything through that link and Dave gets a commission at no cost to you.
Donate With Bitcoin & Other Crypto Currencies!
https://www.eevblog.com/crypto-currency/
T-Shirts: http://teespring.com/stores/eevblog
#IBM #Vintage #Teardown
Amazing state of the art 63 layer ceramic hybrid module construction with 2772 pins!
Part 2 X-RAY! https://www.youtube.com/watch?v=xa0mieJHM94
And Dave demonstrates an amazing BGA desoldering technique that can only be done on these ceramic hybrid PCB's.
TCM PDF: https://eevblog.com/files/ibm-system390-air-cooled-alumina-thermal-conduction-module.pdf
Another teardown with an amazing cutaway view! https://www.youtube.com/watch?v=s7lVfOi7su4
Forum: https://www.eevblog.com/forum/blog/eevblog-1341-amazing-$250-000-ibm-processor-teardown/
Subscribe on Library: https://lbry.tv/ @eevblog:7
EEVblog Web Site: http://www.eevblog.com
The 2nd EEVblog Channel: http://www.youtube.com/EEVblog2
EEVdiscover: https://www.youtube.com/eevdiscover
Support the EEVblog through Patreon! http://www.patreon.com/eevblog
AliExpress Affiliate: http://s.click.aliexpress.com/e/c2LRpe8g
Buy anything through that link and Dave gets a commission at no cost to you.
Donate With Bitcoin & Other Crypto Currencies!
https://www.eevblog.com/crypto-currency/
T-Shirts: http://teespring.com/stores/eevblog
#IBM #Vintage #Teardown
Hi. That's not a processor. That's a processor. This bad boy is 600 watts.
power dissipation, 2772 pins. Is that a Socket 2772? I guess it is. It's got 121 dies installed. It's 63 layers.
It's liquid cooled. It's automatic. It's systematic. It's hard.
hydromatic ultramanic. What could be grease lightning. Thank you very much Jim Renery for sending this into the mailbag. I had to do a separate video for this.
It's so cool. This is an Ibm 9121 processor module which comes out of a System 390 slash Es 9000 whatever you want to call it parallel. Uh, Enterprise Server from 1991. this thing is a beast and we're going to tear it down.
So let's just toss this little pathetic thing away. Check it out. There's the day code there now a second week, 1991. I can't make heads of the tales of the part numbers on this, but I do believe is an Ibm 9129 processor out of a processor frame For a System 390 Es 9000 server.
it went under different names. It's kind of weird if anyone knows the exact like story behind that. Some people like they actually branded it System 390, other times they branded Es9000 or something apparently. And anyway, this is a state-of-the-art mainframe processor from the early 90s.
Oh, and it's just look at it. Ah, Thing of beauty, This is joy forever. Behold, the Wonka Mobile Thing of Beauty is a joy forever. We've got a gigantic ceramic substrate in here.
Uh, as I said, 2772 pins on this bad boy in a gigantic metal frame that weighs 2.2 kilos. Just the just the processor module on each frame. Uh, like a physical frame. or try and find a photo or actually had uh, two of these processor modules I believe back then like it was a big deal.
No I didn't just crush the pins because it has like little tabs on the bottom. It's got like individually engraved numbers on it so I don't know they tested and would have tested like and maybe characterized each individual one. back then and I found an old uh ad in a magazine back in the day that the Ibm 9121 processor system uh went for up to one and a half million dollars. I don't think that was this actual just this module.
It was probably only like a couple of hundred thousand dollars for this module. So cheap as chips. Uh, this is 1991 too. Uh, none of this modern 2020 Fiat currency rubbish.
Now this has around 20 Mips or 20 million instructions per second. And as far as processors in the early 90s go, we're talking like an 804 86 processor back then. I believe that came out in 89. but basically that was still like the top of the range processor.
the highest speed version that I believe the 50 megahertz of the 804 rounded about double that it was about. you know, 40 million instructions are per second. So this thing wasn't exactly stated out in terms of Mips processing power compared to desktop pcs of the day. But you've got to remember, this is a massive mainframe processor system designed for huge amounts of data processing and other stuff. and this could access up to nine gig of memory per processor. That doesn't sound like a lot these days, right? Nine giga memory. But back in 1991? Whoa. This is heavy.
And of course, heavy it is. As I said, this weighs 2.2 kilos just for the processor module itself. And this package is actually called a Tcm or a thermal conduction Module because it actually came in two types. obviously.
Uh, these pins the Uh 2 72 pins on it. They connected down to a matching socket in this thing, but these processors actually had a huge heatsink on top of them that weighed like five kilos or something like that. And I believe this came in like both air-cooled and water-cooled versions. And here's a picture of what one of those would look like.
So there were, as I said, two of these modules per uh, physical frame they called it Uh, which held the two processes and had all the wiring and other Uh support stuff and the power supplies and the cooling infrastructure and things like that. So yeah, this thing could dissipate up to 600 watts, there's a 121 chips, or up to 121 chips, and they can dissipate up to 10 watts each for a maximum module power dissipation of 600 watts. So yeah, this little piss ant thing. So let's take this puppy apart First we get this frame off and check this out.
Just notice this little uh tab on here. I reckon that is for a thermocouple to measure the temperature of this thing. Most likely I'd be stunned if they weren't measuring the temperature of this thing. Look at all these screws to hold it down.
I believe it is, uh, oil filled inside like the entire substrate filled with oil and we'll see a whole bunch of heat sinks as well. Oh yep, can crack it. This could take a while. Yes, I do have an electric screwdriver but not one that supports this tip.
I just, uh, unscrewed the last screw. damn I didn't have the camera running and I heard this and like big o-ring seal in there. I don't know. oh can you? I'm not sure if you can hear it.
Hang on Now I'm just deciding which is the best orientation to lift this up from. Oh I can see in the back. I'm having a peek and I know that this will be the most impressive if I lift it up like this. Actually I have to cut along there.
so sorry for all you purists out there, but we're going to avoid the avoid the warranty on this bad boy. A million voices just cry out in cry out in anger. All right here we go. Are we ready Rolling camera? Because I only get one take at this.
I think wow Look at that. Whoa. that is gorgeous. That is processor porn right there.
Wow. wow Wow. Wow. Look at the individual copper heatsink slugs on each little chip inside there.
and yes, there is a like a thin little layer of oil, but uh, it hasn't oozed out or anything. You can see the o-ring seal around the edge there. Wow, that's just brilliant. Yes, they all have shifted. They all should be nice, evenly spaced, but absolutely beautiful. And you can see that they all go into individual little uh, machined slots in here, which by the way, have little springs in them. Check that out. There's actually little springs inside each one of those.
Not all of them are populated. There's some that actually don't have any, uh, heatsink slugs at all. But yeah, there's little little springs down in there to keep the pressure to keep the tension down on the uh, the die and you'll notice there that some of them just don't have dies installed in them. I don't know if that was like just an optional thing, but it looks like like the pattern is there.
I'll show you a closer up later. but uh yeah, it looks like the the die pattern is there. But uh. anyway, these are like, um, flip chips.
These are chip scale packages. so they're It's like from 1991. Whoa. So you can see why this is called a Tcm or thermal conduction module because it's all about the thermals.
It's all about getting the heat out of each individual die on there each individual chip. And as I said, each chip can dissipate up to 10 watts. So one of these are copper solid copper slugs here. Even though they don't actually have direct contact.
Uh, to this except on the sides. I guess their fit is if I can take. Oh yeah yeah yeah, their fit is very, very, Oh yeah. Hang on here We go.
Here's one of these copper slugs and they just fit brilliantly and I can see the oil, the bubble oozing out and look at that. Oh, I could play with that all day. Oh, that's just beautiful. Absolutely beautiful.
So yeah, they don't have. well, the mineral oil is going to be, uh, heat conductive as well. So you know it's like it's going to have an extra bit of thermal resistance in there. But anyway, that's how each one of those individual dyes can dissipate up to 10 watts with a maximum module dissipation of 600 watts.
It's just. oh, it's fantastic. There's some of the pattern on an unpopulated chip. Look at that.
Wow, that's really something and I know you want to know how many viewers and what via hole sizes we've got on here. We're talking 78 500 of them and they're 100 microns a pop. Now I'm actually going to get all of these heatsink slugs off here and I'm going to put them back over here so this could, uh, take a while. And they do have like a little, uh, knobby bit on the top, little nib on the top and that goes down into the spring just to center the spring on this thing.
So anyway, um, I'm going to stick them in and you know which ones actually which holes are populated because they've got springs in them. Um, the ones that don't have springs. I won't put a copper slug in there because it might be hard to get out. Geez, a lot of a lot of suction on there.
You've really got to get them off at an angle and uh, oh, and will those springs yet? the springs will self-realign so that's nice. Obviously, the manufacturing process is not as messy as the disassembly process. This is ridiculous. What am I doing? So I'm not sure what sort of mineral oil this is. So what sort of oil seems? some sort of mineral type oil in the seismic industry? We used to use Isopar or Isopar M I think it was. um, so you can go look that one up and we had a license to use that with these gigantic tanks of it. Um, just filled with because it used to. Uh, we we had one tank that was filled with Isopar.
That, but that tank was wasn't just a tank, it was like part of the manufacturing. Ah, these slippery little suckers. part of the manufacturing process where we would put the outer skin onto a seismic streamer and uh, so the outer outer poly put the kettle on. Skin was uh oh, gotta get the right way up.
It was, um, yeah. like extruded kind of for want of a better term. Oh, there's a little little something or other there. A little.
um yeah. It was like extruded out of this machine. That was this big drum that was filled. I'm talking bigger than a human size drum.
Like seven foot, eight foot tall or something. And it was, uh, there would be extruded out of that and it was just filled with isopower and we'd have Isopar all over the floor and it was just. ah, it was. yeah.
Fun stuff. Okay, that one doesn't have a spring, so I'm gonna give that one a miss. It's only like four or five that don't have springs. Oh oh yeah, yeah, I've broken off.
Are they like little bypass caps in there on the ceramics? How would they like just come off like that? There's a couple of them we'll have a look at those under the microscope, but surely like I'm not putting much force on these at all. so I'm not sure why they'd just they wouldn't just fall off like that unless there's some sort of like they're not actually soldered down. They're just there's another one. There's another one right there.
Look at that cannot explain how unusually therapeutic this is. Yeah, yeah, I think I'm just breaking off all those little brown what look like caps, but I don't know unless I get them under the microscope. And no, that's all correct. I got them all in the same.
I got them all in the correct order. one two, three, four, five, six unpopulated chippies. and they didn't put the springs in and oh, these are oh I don't wanna. but ah, this is just so much fun.
Oh boy, I wish this was feel a vision. Wow. Anyway, yeah, we've had some of these little chippy things come off like half a dozen of them, so I'm not sure what the deal is. I'm going to, uh, I'll just leave this gunked up and I'll just get some paper towels and wipe off all this because it'll be easier to see under the uh microscope and the macro lens.
Um, for the chip. so I'll get back to you anyway. that is the module in glorious 4k without the with still with the mineral oil on it and you'll notice that they're not all identical. Why aren't they all identical? Because these modules are supposed to contain uh, sram chips as well 128 kbit sram chips. and I'm guessing that these are down here. Why are they kind of like oddball? Um, sort of. not really. You know.
symmetrically placed. I don't know. Um, but anyway. like I assume that they're different types, but even they're a different color.
slightly different color to those ones. and I look that could just be light refraction of the dyes because, well, that's what dyes do. Pesky little things to look at. and uh, well.
quite beautiful. Actually, get them under the right, uh, light. and they're They're quite spectacular. but uh, yeah.
Anyway, this is not just one module with like 121 different processors on it. It's got tons of different logic elements, including 128 kbit S rams and that was. that. was pretty huge for 91 and their 10 nanosecond access time srams too.
Actually, it seems to be near impossible to just wipe this oil off yet. Yeah, there's another one that, yeah, I can see the pads on the bottom of that chip so they're coming off so these are not soldered down. I reckon these are, just. well.
are they just press fit? Wow, that'd be interesting. I don't even know. Soaking the whole thing in isopropyl do the job. Um, because yeah, that oil is going to be hard to get rid of.
Well, that's how you drop the frame out. Um, I just put it up here. so to raise it up just so that I can take some macro photo shots over always. Uh, my tear downs always.
Usually always have a high res tear down photos over on my Flickr account. so check that out. Um, and this just and the whole thing just fell off. So we're left with ceramic substrate which is pretty groovy though.
So anyway, I'm really having some fun. Uh, taking photos of all this at like different angles and stuff. It really is just quite something. And I can of course adjust the iris F 3.4 it's the lowest my camera will go at the zoom if I increase our aperture, everything becomes in focus or mostly.
And let's check this out under the tagano. Now this is uh, manufacturing material science at its absolute finest. It's like absolutely phenomenal. I'll link in the paper, uh, down below, you can read it for yourself talking about all of the uh construction technology that goes into this.
But we're talking about a Uh 63 layer ceramic substrate here. As I said, 121 devices on here, 144 caps all in here. We'll take a look at those. Each chip has 648 pads on it.
We'll check that out closer up. A mix of Cmos and Bipolar technology devices by the way. And there's 78 500 vias on this thing. I mean, it's just.
it's just absolutely incredible. And each chip there has 648 pads and uh, it's This is Not a Pcb. Okay, this is a ceramic substrate with a mixture of uh, thin film and thick film hybrid layer technology. So I believe the top layer is uh, thin film printed and the inner layers are thick film printed. Now those 63 layers? 0.2 millimeters. That's 7.8 foul thickness each. Uh, we're talking about 12 micron conductors. that's like half a thou half a thou conductors on this thing.
So yeah, it's just it's just absolutely nuts. And the top ceramic surface layer has a surface flatness of five microns. so all the mechanical engineers out there probably getting moist over that? I don't know. Let me know.
Is that good or not? Five micron flatness over this entire module? So the ceramic substrate here is made with a mixture of alumina, powder, glass, powder, organic binder, and plasticizers and all sorts of stuff. So you know, really, just incredible, uh, material science involved in this just. you know, the manufacture of this board. Anyway, you can see the capacitors down here and we let's find one that's ripped off because I did rip off a few.
There we go. There's the little pad for the Uh capacitor down there and yep, I can feel that oh look at that. look at that. The the solder.
I'm going to use that terming quote marks the solder. Look at how I can just like this has not been heated up at all. so they're obviously using some sort of I, I, you know, weird-ass um, metallurgy here. For the solder.
On these things, I'm going to assume that it's the same for the chips as well. But there for the Uh capacitor, there's the capacitors. Okay, I'm going to try and get this off. Okay, I'm going to put a little bit of force on that tongue at the right angle.
Yeah, all right, she's budging. There it is. there. it is.
Have you ever taken off a chip with that sort of ease? Ah, this is just. this is magical. This is A and floating around in oil. Oh, this is my favorite thing ever.
This is just incredible. A floating oil. A capacitor floating with oil. Oh my goodness.
Which you can just imagine being able to take off chips with that sort of ease. Ah, Lewis Rossman, Eat your heart out. Okay, people are going to be horrified, but I'm going to try and do this with one of the chips. Oh, people are going to be more.
I mean, we've got like hundreds more pads on here. Hundreds and hundreds of more pads. But let's see. Oh no.
I'm putting a lot of force on that. Oh, I'm putting all my hands slipping. Don't you hate it when your hand slips due to the oil? Jeez, Yeah, I can't I can't make that budge. But certainly the capacitors piece of cake.
They just go off like that. Oh, this is so much fun. Oh this is great. Let's flip the there's our cap.
There you go. We flipped our cap over and there it is. 16 pad capacitor. Which looks like it contains because we saw the top of it over here. Does it contain four individual caps like that? Perhaps. I don't know. Anyway, you can see that they're just like there's no no traces coming off there. They're just buggering off down in the internal layers.
and you can't see through these layers because they're like ceramic. They're a ceramic slurry, uh, substrate. And as I said, like individual vias. Um, down in there, we're talking hundred micron holes.
Um, and these are all you know? Like I don't know. Test pads. Not sure what the deal is. I assume you know some sort of test pads.
Something like that. I'm never going to get this oil off. I'm not even going to try. Ugh.
I'm already starting to get oil over everything here. So yeah. But anyway, this is absolutely remarkable. We've got 648 pin, so I guess you could call that Bga.
Uh, although you know they wouldn't have used that term back in the day. I don't think. and the inner layers apparently are made of a, uh, moddy molly denim. if I'm pronouncing that correctly.
molly Denim? um. powder. So yeah, this is not. These are not like copper, exposed etched Pcbs.
This is not what's happening here. This is an entirely different technology to what you're used to with your fiberglass. uh, circuit boards. Um, it's just yeah.
it's not the same thing. All right, I'm going to get Medieval on a task. I'm going to get in there with a big pair of pliers. Ah, sorry, this has got a flame comment down below.
Go ahead. I don't care, I'm I'm gonna kind of give that a little. Uh, you can get that a little twisty. Life's pretty straight without twisties.
Oh yeah. yeah, yeah. no. Oh, I know, I've come a gutter look at that.
but at least we can see under it. We can see under it, so that's useful. There you go. Wow.
Okay, yeah, you can see a similar similar thing to the caps in there. You can see the individual solder pads. Once again, I look, I'm going to clear this away. Yeah, yeah.
look. look, you can see the solder. Yep, it's the same thing. It just spreads like that.
So this is not your regular solder that you are used to. Because this is room temperature here. this has not been heated up. Wow.
So yeah, we just completely shattered that dye. This absolutely butchered it. Sorry about that, but that's fascinating to see under that and confirm that it's basically the same interconnection uh, technology as what the was what was under the capacitor there. Which makes sense of course.
So is this some Doug Henning metallurgical magic? Well, I don't think so. I had a look at the document and it actually specifies. Yes, it's solder and it specifies it as 97.3 solder. So another 60 40 Rubbish.
But yeah, 97.3 It's not some magic room temperature solder or something like that. so I've got to assume that's what's here. So why does it seem to wipe off like that? Well, I can. Actually, this is not feeler vision, but I can actually feel the bumps in there. What's actually happening here, and why I was able to magically use the Jones method for desoldering this chip. Just push it right off. Um, is because of the tiny pitch and the tiny amount of solder that we've got here. Uh, we're talking 16 pads here.
But these pad dimensions I checked are only 180 microns wide. So that's like seven thou wide pads on here, so there's practically no solder on there. These aren't balls on the bottom of the chip, right? These are just like they've just applied the solder paste. However, they apply it and they've refloated.
You know, in a not too dissimilar manner to what you're used to. But there's just so such a tiny little piece ant amount of solder on there that the sheer force of that just my force was enough to just break all of those pads at once. It's only 16. I guess the mechanical engineers out there please.
You can, you know if you can do a like, a some sort of sheer analysis or something I'm probably not using the right terminology. but anyway, with 180 micron wide, 97.3 solder, you'll probably find that. Yeah, you can just push these chips off with a bit of force. Now, why don't the pads rip off? Because this was what would happen if you tried this at home.
Don't try it at home kitties. Uh, you'll The pads will just rip right off a regular fiberglass board, even though like the real, high quality high temperature ones. Most likely because a tiny little seven thou pad in their 180 micron pad size. Uh, there's going to be virtually no adhesive under there.
although it might not come off because you've got the Vr in the middle, but I don't Anyway, if you relied on a tiny little pad like that, there'd be virtually no like an adhesive on there to hold those pads in place and you just shear them all off. Um, you'd completely come a gutser. And yeah, Do Not try the Jones method for, uh, desoldering your Bga parts because it's not gonna work. And look over here.
We haven't ripped off a single pad. Why is that is? because Well, this is like, you know this is what you get when you pay. You know, like a hundred thousand dollars. Um, this is like, you know the best that the Ibm Research scientists can come up with.
This is, you know, not just regular fiberglass. This is like, you know, some ceramic woo woo mixture of stuff and it's all just embedded in there. And these pads are. They're probably never coming off.
I put a lot of force on that anyway. That'd be fantastic if you could repair chips like that. But unfortunately, yeah, don't try it on anything but Ibm Magic. Woowoo! So I mentioned this document and I won't go through it I'll just link it in down below, but it's uh, done by J.u Nicobaka? Winning! Um, and friends of course.
Um, and it's basically them. They're a bunch of Ibm research scientists, is basically them boasting about all this marvelous technology that they've got in these, uh, you know, ceramic, uh, modules in the construction of them and everything. And it's highly recommended. Worth a read. Absolutely fantastic. So just think of the people that went into actually making this thing, um, and the technologies involved in manufacturing this, and they're thanking these various divisions within Ibm for doing it. anyway. Ju Nicobacca apparently had a distinguished career at Ibm.
I don't know where he's now. I published like 90 papers or something. I had to look impressive. So there you go.
I hope you enjoyed that as much as I did. and I've got oil all over my fingers. And thank you very much Jim for sending this in. This was absolutely pornographic technology.
Absolutely fantastic. I'd love to like, maybe you know, like x-ray this or something like that. I don't know what we see, you just see 63 layers of interconnections and stuff like that under here. So it's got 400 meters of wiring inside this, by the way.
Um, for those wondering, uh, which doesn't sound like a lot, but I guess it is when you stretch it all out anyway. Um, yes, 63 layers of state of the art ceramic Pcb manufacturing technology from 1991. Absolutely incredible. And this had like half the number of Mips as an 80 486 at the time.
But as I said, you know it's you're not comparing apples to apples there. So wow. Um, that's absolutely incredible. I hope you enjoyed that as much as I did.
If you did, please give it a big a thumbs up. And as always, you can discuss down below or over on the Eev blog forum or over on any all of the alternative uh platforms that I'm on. And I'm also on the Twitters, I'm on the Instagrams, I'm on the flickrs. I'm everywhere.
Catch you next time you.
😱😱😂😂😂🥰🥰🥰
Life's pretty straight without crispies! — sounds gay mate
For the solder on the inset article it refers to 97 Pb(lead)/3 Sn(tin) but in the caption overlay you put 97 tin /3 copper was this changed at some time in production or something?
processor module looks like nvidia's recent grace hopper superchip
Sadly while the hardware was really slick, it was a last gasp of dinosaur technology. Also they often ran dinosaur OS and dinosaur software so the actual performance was often unimpressive. At one University one of these monsters was replaced by one rack of blade processors at less than 1/20th the cost and higher performance.
The original chiplet CPU …..
When talking about the 97/3 he has words saying "97% Tin / 3% Copper" Pb is the symbol for lead, not Tin. The 3% Sn would be Tin, right? Where did the copper come from? Thanks for the teardown!
Thats a 4,500 sq ft house !
Fun fact: Every smartphone today has more computing power.
Why do you talk as if you are Murray Walker commentating an exciting formula 1 race? Highly annoying.
The computers were designed to run for a decade with 0 downtime even if you put a bullet in the computer.
When making the ceramic substrates I described what we did as making metallic spider webs inside of rubies
Get your paws off those pins! ES 9000's had 4-8 Hp electric motors to pump coolant (2 for the supply 2 for the system) through the systems. The TCM s had to be recycled to recover the gold they contained because we had government contracts.. Prior to Y2k we recovered millions of dollars just for the scrap recovered. Hitachi and Crays were equally over engineered
It is crazy how much effort is put into building something that will be scrapped within a couple decades.
This the mother of the future epyc chiplet cpus with 10K cores
i guess they got their money's worth out of it, probably many times over in different companies before it got torn down, but it's still amazing that this piece of equipment is now worthless and you can take it apart!
The empty spots are circuit fails. The units with the most failed spots become the lowest end units, the ones with the least become the highest. They still do it that way today with the failed chips on the same die. It simply becomes a lower core product.
Those copper slugs are worth about $250,000 at today's scrap prices
Lol, I remember we had an Engineering Change (hardware update) where we had to replace dozens of them !
The old ones were "Scrap Locally" …. meaning throw them in the the dustbin….
Realize the old ones came out of a fully functioning 3080 and were working perfectly….
encore une pièce historique qui part en fumée😓😤
I worked for a company in the 80's that made these aluminum cooling hat bodies by the 100's weekly. Along with many of the testing and inspection equipment used in Fishkill IBM.
ES/9000 was air cooled. Two previous generations, 3080, and 3090, were water cooled.
The solder connection between the chip and the ceramic substrate was called a C4 – Controlled Collapse Chip Connection.
I am a simple man. I see torx. I smile.
you can see hotspots on pistons / chipside .. awesome !!!!!
Wow! This brings back memories. I worked at the IBM East Fishkill plant back in the 1980's. I worked in the manufacturing technology areas where the ceramic modules were created. At its peak, the plant employed about 15,000 people, and the weekly production could be driven away in the back of two station wagons. An unbelievable amount of engineering went into the design of these modules and the manufacturing processes to make them. The ceramic green sheet material that formed each layer of the module was created by a long sheet roll process on a machine that was approximately 63 feet long. The roll was then taken to a die machine that precisely cut the dimensions of each sheet. The sheets were precisely punched for layer-to-layer vias, then fed into screening machines that spread molybdenum paste through masks to create the pattern on each layer. The next step was a stacker that aligned and stacked each of the 63 layers so that vias from layer-to-layer would align. The stack was then pressed and readied to go through the sintering process. The sintering or firing, of the ceramic substrates was done in huge sintering furnaces that had a hydrogen forming gas. Each furnace was loaded by robotic arms that moved assemblies that weighed hundreds of pounds. When the process was running, the excess hydrogen gas was burned off in a stack. It looked like something from the Wizard of Oz. Occasionally, there would be an air leak into one of these furnaces where the oxygen/hydrogen ratio would be just right to cause an explosion of the entire furnace, shutting it down for an internal rebuild. The ceramic module would then go to the thin films processing section, where it was CMP polished for flatness and the multiple thin film layers were deposited using technology adapted from wafer manufacturing. In the newest generation of TCM, there were six layers of thin film redistribution wiring on top of the ceramic, before the chips were attached. This process had to provide for corrections in lower layer yield defects, and had to achieve 100% yield over the entire substrate. There were multiple customization laser repair processes that added/subtracted single trace defects between the chip sites. Wow! I could go on, but I'm running out of steam… maybe someone else can pick up and describe the additional complexity in the manufacturing processes of this unique assembly.