EEVblog 1380 - BBQ IBM Chips
HTML-код
- Опубликовано: 11 сен 2024
- It's an aussie barbie with fried IBM chips.
Part 1: • IBM BGA Thermal Futili...
Part 2: • Abandon Lab! - IBM TCM...
Ken Shirriff's blog: www.righto.com/
Forum: www.eevblog.co...
Subscribe on Odysee: odysee.com/@ee...
EEVblog Web Site: www.eevblog.com
The 2nd EEVblog Channel: / eevblog2
EEVdiscover: / eevdiscover
Support the EEVblog through Patreon! / eevblog
AliExpress Affiliate: s.click.aliexpr...
Buy anything through that link and Dave gets a commission at no cost to you.
Donate With Bitcoin & Other Crypto Currencies!
www.eevblog.co...
T-Shirts: teespring.com/s...
#BGA #IBM #Microscope
LOL, You did not only fry them but these TCM are also ESD sensitive.
I had to replace a whole bunch of these TCM's (working perfectly!) in the late 80's for an Engineering Change.
According to the documentation the price for each of those TCM's at the time was over US $300.000 !
At the time you could buy a house with each TCM !
A "TCM" is just a type of module.
You had many different types of TCM's ... CPU's, Storage, Storage controllers, Channel controllers and what not.
They were fitted on large multi layer (64 layers if I remember correctly) boards with large ZIF connectors. (hundreds of pins in a square matrix)
Those old 3080/3090 computers were very expensive toys..... but you got a lot of hardware for your money. these things were huge!
Kinda makes you wonder what the point of it all is right? :) All the effort and expense, few years later it's rubbish (cool rubbish nonetheless though).
@@VincentGroenewold Just like the latest and greatest phone... thousand or two bucks, and in 5 months it's obsolete junk!
@@VincentGroenewold They can't get the technology of later years if they don't do the technology for the current year. It is a step that has to be taken in order to move forward.
@Marcin Berman There hasn't been any innovation in phones for the last 10 years. They are just thinner with bigger screens now. A 10 year old phone can still run all the modern apps unless the manufacturer purposely slowed it down with an update.
Thanks for the insight, Spikejwh1.
But forget your concerns about ESD Damage. These chips are ground or etched in order to examine them under a microscope. A fault in the semiconductor junctions does not interfere. The purpose is explicitly not the use as a spare part.
To Phones(@Canadian Rocketboy, @rasungod0, etc): You should decide for yourself that something is obsolete and not the marketing department of any company, whose product you have paid for and are therefore the owner. The fact that millions like drunken minions are thirsting for the very latest hardware, at least pleases the shareholders .... ;)
Edit, I forgot something-> Back to our Australian: What causes problems, however, is if you don't remove the damn oil. And then also bring it to a boil/burn, LOL wonderful! This leaves residues that may make the subsequent examination impossible. It's so easy. Simply rinse and clean with commercially available solvents.
"If you want to see some balls, I'll show you in a minute." Umm umm no thanks Dave I'm good. 😍👌 Just kidding, good work,
Whip ‘em out Dave, they might need a Pinch&Roll scratching.
The fact I’m watching this on a 1/4” sheet of aluminium lithium glass is astonishing.
This is what I came to the comment section to see.
Billion dollar massively multicore cpu on the barbie. No whakas.
Isn't it wuckers, as in no wucking forries?
Bird poop is the least thing I would worry about in Australia. So many deadly platypuses about and poisonous wombats.
I look forward to Ken Shirriff's analysis. His posts are always excellent!
Putting it on the barbie was a great idea... But you forgot the garlic flux!
It never ceases to amaze me the complexity that is inside any tiny little piece of silicone chip that we use every day. but remember this is very old tech I have no idea what the nano meter which is on this but I’m sure it’s nowhere near 14. And stupid gamers have the balls to complain about ICs not being 5 nm. What an absolute art of engineering. Hats off to the engineers at IBM and to anyone who knows how to create these little wafers of wonder.
Screw that, find a big metal spatula and just flip the whole bloody thing over and see how many chips fall off! Oy, mate! 😎
I was thinking blow torch, but yes a chisel probably works better.. no smoke, but then, they are very brittle so... hmm
13:29 and this is where the movie Tron was shot.
16:45 - “If you want to see some balls I can show you that in a minute”
NSFW
Congrats getting the chips off, Dave. Good to see the magic smoke production machine outside!
The shot of the bottom sides of the chips seems to confirm what I initially thought, that the chips might have gotten stuck on charred oil. Giving the entire bugger a good rinse with petrol and letting it vent for a few days might have been a good idea to get rid of the gunk before toasting it.
Ken Shirriff does awesome work!
the ZL and RX numbers are the mask numbers used for that layer (Z usually indicated a metallic layer, I forgot wat R was)
RX is the mask level to define the diffusion layer to be doped to either P or N type.
I was the lead characterization engineer on the Liberty SRAM program. It was designed and fabricated at the IBM Essex Junction site in the early 90s. It os a CMOS "high speed" (something like 10ns access time) static memory. I honestly forget the density as I worked on several designs. Pretty cool to see after so many years. I'm still in the industry working for Marvell Semiconductor now.
When you first mentioned doing this and the soft solder I was thinking about pulling a single strand of steel wire under the chip. Wanna try it Dave?:D They do it to separate digitizer glass from LCD screens on phones with cracked glass.
Molybdenum wire is used for LCD separation. Out of curiosity I once bought some on *bay - 0.03mm can really cut your hands..
Chip-scale BGA - this was way ahead of it’s time!
I though the idea was to use the hot air gun as well? Similar to preheating for soldering/desoldering just a little hotter
and cleaning it a lot, and adding flux. This is kind of butchery there as it will take so much more efforts to clean the burnt oil residues.
Maybe the chips had underfill and that's the reason they started to desolder at 280c+ ?
No they were only reflowed according to the docs I found. Those thermal modules just have a lot of thermal mass while the dies do not, so the heat gets sucked right out.
Came for the chips, stayed for Daves balls.
Drop it in a deep fryer. All the crumbs wind up in the bottom as I've found out. :)
I've actually done this to recover some large mass multi pin parts at home. It worked great!
@@Audio_Simon :)
Its mind blowing that this is obsolete technology.
My dies usually flowed at 415°C within 30 seconds
Low temp ones will flow at 300°C
Those are big die chips.
"That's very cool" he says, looking at a smouldering heap of electronics, burnt oil, shimmering heat haze...
While it cooling you can smack a steak on it, and get a IBM branded steak. :-)
Winner winner chip dinner!!
The microscope stuff was cool! I may be insane, but on the last chip (the small one) are those transistors under the power lines? Cuz they sure look exactly like a drawing of a transistor! Amazing.
Cool video Dave. After the nightmare of frying the poor TCM I got the satisfaction with the nice microscope images. 👍🏻. Cheers from Austria 🇦🇹
It's always amazing seeing integrated circuits zoomed it, it's like a work of art.
it always amazes me how stuff like this is super microscopic it blows my mind how things that small can be made!
That memory is on the order of 288Kbits. Today the same size ASIC would have 1billion gates
Yea cooking with Dave ... sizzling on the pre heater, need some tommy k on those chips yummy. Winner winner chicken and chips dinner haha.
Nothing like throwing on some silicon waffles on the George Forman grill!
Send a few to zeptobars. Also noopy on the forum has some nice chip teardowns.
@16:43 "If you wanna see some balls I can show you that in a minute"
No thanks Dave, I'm good. :-P
Those large rectangular symmetric grid areas are probably rams. The void areas with lots of metal in them are probably capacitance fill which is used to add additional bulk capacitance, and to somewhat level out the surface to improve manufacturing.
15:39 those look like large buffers (basically an inverter), probably for sending the signal off chip with a lot of drive strength.
Absolutely amazing tech for the early 90s...
This is encouraging me to send my grad school die to Dave so he can take a look and roust my layout lol Unfortunately you can only see the top metal layers
Two scammers in the comments talking about crypto. Please report.
Ken won't have to use any dangerous chemicals on those because they're already bare dies.
He'll only need to clean up the oil residue and maybe wick off some balls.
Watch your language
He may dissolve off the metal layers with acid. He tends to do that in a lot of his teardowns. Particularly if the metallization is dense, like on the pink chips.
@@PeregrineBF Ah, OK
Oh! Dude - just put it on the heater upside down. When they drop off, you are done.
Hi Dave, maybe you would have had better results if you placed a thermal blanket over the PCB.
The amount of heat energy being lost to the atmosphere must have been very high.
Amazing! Thankyou for sacrificing that big cpu.
Now we have to wonder how many different die are on that substrate, but it would take to long to take them all of and compare them
I don't think Dave would ruin that rare piece of art. But if someone knows the layout and can point out which dies are different, he would probably happy to desolder one sample from each type (unless most of them are different).
I have no clue what Im looking at when you zoom in, but its definitely cool!
I worked 35+ years at the IBM foundry where those chips were manufactured. (Wafer-test, burn-in, module-test.) We used VACUUM PENCIL to handle the wafers and chips. (not tweezers) I spent my share of time looking thru microscope at wafers to align the probes to touch those solder-balls so the chips could be tested. (we used joystick to move the wafer under the microscope... the wafer was on a vacuum-chuck)
Fishkill or BTV? I thought CMOS5 was BTV.
@@kk1l - BTV (Burlington Vermont) was the foundry for wafers/dicing/chips. EFK (East Fishkill NY) did the packaging of chips onto TCMs.
@bruce peebles BTV is still active as Globalfoundries FAB9. It is 10min from my home and where my office was located for 35 years. The Fishkill fab is GF FAB10, soon to be owned by ON Semi. Surprised we don't know each other IRL.
Of course.i You try not to break it, and they just fall off willy-nilly, but when you WANT to remove them, they refuse to budge!
Are you sure it is oil? It looks like protection resist (usually photoresist) that is applied before cutting. You would strip resist with acetone and IPA in an ultrasonic bath, if it is not haveily crosslinked.
The big pads on the die are for electrical testing. The cross in the corner is an alignment mark for the stepper. The numbers are the mask desginations. Purple is oxidized silicon.
I thought you were barbeque-ing with the heat from IBM chips. I immediately though of BUNT chips (Burlington Universal New Transceivers) used in the big IBM disk systems. Plenty enough heat from therm to slow cook a brisket.
get some flax on it.😂
Look forward to seeing what Ken does with them...good work sir !...cheers.
What's Ken's channel?
@@mikeissweet I found him on " AVR Freaks" he did a piece on un capping a 7805 regulator..it was great, try there !
@@andymouse do you have the link?
@@domtom128 just google " avr freaks " you will see it
@@andymouse I know the forum but I can't find Ken's post
Are those going to Curious Marc's friend Master Ken?
Yes
As someone who had the misfortune of using that same Puhui preheater, it's not a great preheater. It heats really unevenly, the temperature measurement is pretty off, and the "clamp" rails only attach on one side, and have enough play as to make them worthless. But hey, sometimes brute force works. :D
Yeah, it's not great, but I got it for 10 bucks on ebay, no other bidders :-D
@@EEVblog It got the job done at least :)
Persistence! Good work Dave. You’re an inspiration to all.
That is one ballsy little chip!
chips on a barbie, wait, that's not how it goes
16:48 "if you wanna see some balls I can show you that in a minute"
Gorgeous pictures
Thanks
Ahhhhhhh a thing of beauty. Joy forever ;)
Thing of beauty, joy forever!
16:47 I thought this was a family channel :P
Can anyone post Ken's channel or site? Dave mentions Ken, but no link to him ...
Link added
@@EEVblog Thanks
I did like this video and I did give a thumbs up. I also, as always, enjoyed your joy.
Yeah! You get these bad boys of.. 👍 Great job, Dave!
Guessing this is 95/5 (reflows around 230C) if your IR is even close (doubt it is 63/37 because that's ~183C). Too old to be SAC-305.
a layer of aluminium foil over the whole barbecue could have sped up the process by keeping the heat from escaping off the top.
Fascinating work with the microscopes in real-time. What a difference good lenses & lighting make!
"How to get the balls off" ... I suppose it will be encapsulated in epoxy and then ground down with a very fine emery cloth or something similar, at least the guy from the YT channel "electron update" does it that way. Interesting video Dave, thanks ! Take care !
So was were these things a WLCSP before WLCSPs became a thing, some 25 years later?
Looks like an Arial view of New York city streets with a nuclear bomb mushroom, "solder ball"
Acetone with an old toothbrush would more than likely have removed that burned oil...
Try some soap and water first...
I love Dave and EEVBlog...There I said it!
About 25 years earlier, the densest logic was in cordwood modules in the CDC 6600 like this one: www.computerhistory.org/revolution/supercomputers/10/33/60. The 6600 had 40,000 transistors. Wonder how many times more transistors are in just one of these barbecued chips.
Hope there's no PCBs in that oil....
I remember reading a story about gangs in Africa who would drain the PCB oil from electrical transformers and then sell it to street food vendors as cooking oil. Dave's got a setup here!
You can remove the oils with breaker cleaner.
How is it possible to see through the layers of silicon and also through the solder balls? Is silicon slightly transparent on that thickness?
And are the "see through" solder balls just because the optical properties of the microscope do that to the image? 🤯
why didn't you put a cake pan over it to keep the cool air off the top?
Why don't blast it with some brake clean to ged rid of the oil?
Remarkable and cutting edge for the time
"Not fish... Snake!"
got the same idea: Expected to read "tyrell corp"
Now we need an answer to more important question - can you cook good steak on IR preheateter
Please also send to the Zeptobars guy. He does ultra high resolution IC images.
I suspect that's mainly aluminum metallization we're seeing. Is it possible to count the metallization layers on the various chips?
From that era yes it it Al. Speaking for the Libery CMOS memory chip IBM was just transitioning from non-planarized metal to planarized. I think we were at 4 metal layers at the time.
Well worth the effort for those die shots. Glad you didn't give up.
What causes the optical distortion around the ball.
You really have to wonder how much work there was involved in stopping cross talk, ringing etc.
We had proprietary CAD software that did signal analysis (including crosstalk) as well as strict design groundrules on the length of nets, the number of loads per net, etc.
Coupling is a big deal especially in the sensing circuits where you are dealing with trying to read a 100mV differential off the bitlines.
Arnie in Commando I recognise, watched that a ton of times and maybe, I think Ghostbusters, I can't really remember I wasn't really a fan; though I remember it being pretty good. Cheers Dave 👍😅
Well that was a relief. For a while there I thought you might have to resort to drastic measures to really properly destroy this thing.
This is fucking awesome!
throw some PCBs on da barbie mate
That ball at 13:41 almost looks like a bad 3D rendering using a standard cloudy day skymap!
Why don't you even try to clean the oil in some ultrasonic cleaner bath before heating it up?
Struth Bruce throw another CPU on the Barbie mate..
Would you say them chips are well done or medium rare
really pretty dies... I am excited to see what Ken thinks of these ! He might be a huge nerd ! All Marcs buddies are some smart, smart people.
Welcome to NerdBBQ now with Kelvins.
15:47 i expect to read "tyrell corporation"
Weyland Corp
I was waiting for some flamethrower action.
joy for ever
anyone googled that PN?, there's absolutely NO results
The number you Googled will be a serial number, a production part-number or the Engineering Change (=version) number of the TCM and not a P/N you can order.
What Dave showed was not the complete TCM but only the substrate with the chips.
There should be big heat-sink with a plunger to each individual chip to conduct the heat away from that chip.
It is rather a complicated module, not only electronic-wise but also mechanical-wise.
Yep, zilch!
Those are the part numbers for the masks to print those layers (ZL and RX). You can find others arounf the chip perimeter typically in the corners. The chip is Liberty, a CMOS SRAM designed in Essex Junction, Vermont.
Toluene or benzine would be better for flushing the oil off the chips.
Well try dish soap to clean it. It's burned oil after all.
Soap is good for non-burned oil. Once it's carbonised even solvents are not very effective.
@@ferrumignis Dish soap is designed to solve crusted oil and fat.
@@kyoudaiken Dish soap is simply a detergent, it will emulsify fats and oils in their normal state, but when burnt into carbon, dish soap wont touch it. Even soaking in boiling water with dish soap has little effect on burnt fats/oils, it requires manual removal by scraping or abrading with a pan scrubber or caustic chemical remover.