IBM Power2 MCM from RS6000 Computer, Desolder, Teardown and how to Tutorial

[@adriansdigitalbasement]

Beautiful CPU and great work getting it out so cleanly!!

[@CPUGalaxy]

Thanks Adrian. Yeah, it was thrilling me doing that and making the video byside. I had only that one chance 😅. Most decapsulated Power2 MCM I saw so far in collections were damaged. And I am very much into keeping the things unbroken. Yeah, seems I found the right process. 🙃

[@hypergl6974]

@@CPUGalaxy now i get why is your channel called CPU Galaxy 😉 what a nice collection of chips you have!

[@CPUGalaxy]

thank you ☺️

[@pcclassic]

Мужик, у меня есть 3 штуки RS/6000 и все они прекрасны! И как раз в одном из них, model 530H, такие процессоры, я в восторге! Спасибо!

[@pyrotas]

This channel is kind of the lockpicking lawyer for CPU nerds

[@alensabanovic747]

It's a beauty :D That white ceramic really makes it glow.

[@JosiahGould]

That's a gorgeous chip. I remember seeing a shelf full of RS/6000s back in the day, running a fairly hefty website. They were quite loud if I remember correctly. Now you can give Wikipedia a good picture of the Power2 chip!

[@douro20]

Not ball-grid array, but a related (and much older) technology known as column-grid array, or CGA. It was developed in the late 1960s at IBM as a high reliability interface for integrated circuits and hybrid modules. It uses columns made of a high-lead solder (90% Pb/10% Sn) to interface a ceramic land-grid array package to a PCB. The columns are usually formed directly on a PCB by pouring the molten material into a steel mask and ground flat before bonding the package, but this process requires special machinery and is only practical in high-volume production. Newer techniques which don't use masks have expanded use of the technology in fields where high-volume production isn't practical such as aerospace and defense.

[@max1024]

White ceramic + IBM = fantastic look =)

[@CPUGalaxy]

definitely 👍🏻

[@cringemaki]

Man, back in the day when processors were made with ceramic. This is such a gorgeous piece.

[@RetroArcadeGuy]

Back when processors were made to last forever and not made on a wooden base that would bend and kill it. Or like certain overrated AMD CPUs getting ripped off when you try to take the heatsink off...

[@CPUGalaxy]

exactly 👍🏻

[@xeveniahdarkwind178]

Ceramic cpus so much easier to work with vs fiberglass composite substrate and you didn't have to worry about the core or chip cracking during heating or thermal solution installation... I miss those day the age of K6 and Pentium mmxsuper socket 7...do remember that The horrible silicon goo aka thermal interface material was soo hard to apply properly kinda like artic silver 3 and that stuff worked but was conductive under pressure granted the composite substrate helped protect the die from overheating and cracking Mid to late 90's early 2000's.. And if I ever run across Artic Silver 3 it's affordable I buy it or try to obtain it.. nanotherm silver xtc or tuniq tx2 and I keep it for legacy system work or repair.. My original November 2001 Xbox 1.0 system has all thermal goo changed out...

[@xeveniahdarkwind178]

@@CPUGalaxy ditto...

[@ЖИЗНЬВЛЮКСЕ]

amd epyc now made chiplets like this on 7 nm

[@derekkonigsberg2047]

I actually have one of these, sitting on a system planar for an IBM RS/6000 POWERserver 590 that I dismantled years ago. Probably the most impressive system for its clock speed that I've ever used. IBM really went superscalar-to-the-max when most others were only starting to explore the idea.

[@jaimeduncan6167]

Beautiful. The power1 architecture was very influential in the story of the superscalar computers, that today are the norm.

[@fsfs555]

IBM's chips are often very pretty: their high-end chips (including the PowerPC 970 series) are done on white ceramic as seen here, but others (the PPC 601, 604, and some 603s, all in QFP form) have previously been done using epoxy encapsulation on a ceramic substrate using yellow, blue, or blue-green epoxy. These were used in Power Macs and some low-end RS6000s along with a few odd network appliances.

[@arniceousmaximus2183]

Lach vip spot recognized for that beautiful chip who worked hard and deserves a purveyor such as yourself. Naja a pleasure watching!

[@barryg41]

Never thought it would drop that easy. Brilliant work my friend. 🙌 Beautiful results!

[@dav1dbone]

Genius idea to use a regular heat gun to remove the cpu - probably could not have done that as easy with a hot air station - brilliant!!!

[@JW-uC]

I love how we're going back to MCM's with AMD and now even Intel is looking into it. I think AMD is also looking into chiplet'ing graphic cards. Especially interesting as I believe IBM did it for two reasons. 1) making physically large die chips was hard (and high fail rate, IBM even sold "cut and shunt" processors made from two half working muti-core processors wired together). 2) different substrates (chemical dies) had different electrical properties so it made sense to use a higher power dissipating, but slower, chips for "off processor" connections (that required more power to run over the motherboard traces) and lower power (but faster) chips for the "L1" cache and for the processor itself. IBM also used high speed caches of the memory (over and above the normal "on chip" style caches) directly on memory cards (you plugged the memory dimms into a card that then plugged into the main board)... which allowed both massive amounts of memory (for its day) but also massive data throughput (again for its day). It was quite amazing how much of their Mainframe tech was transplanted in to their Midrange (server) and Workstation computers.

[@gordonfreeman320]

What a beautiful piece!

[@PedroSantos-os4iq]

This CPU is amazing. Thank You for this nice tutorial.

[@Wacypro]

Oh lord I died a little inside when I heard that crack! So glad everything was alright

[@dieSpinnt]

Wonderful! I use this method, too. When harvesting components to play with:) Just one tip: I know you were excited, but first switch the heater off and let the fan blow a minute before switching off your Steinel (I mean the heat gun) It thanks you with a longer lifespan;)

[@arniceousmaximus2183]

1 month ago I try to catch up a little the side view was fantastic as the chip was removed in a very nice methode inmost say bravo you took extra time and methodical thinking paid off. Spitze!

[@Carstuff111]

Back in the days when technology was beautiful and unique just to look at normally, unlike the mostly cookie cutter tech we have these days that is only beautiful under a microscope. That is pretty awesome :)

[@pc-sound-legacy]

Awesome piece of compute history!

[@goldengerd6040]

Wahnsinniges Teil :o Wie gern würde ich mal deine Sammlung persöhnlich sehen, deine videos sind für mich wirklich eine gute Grundlage für meine Projekte die ich am laufen habe, Ich danke dir für deine Arbeit und vor allem dafür das du dein Wissen und deine begeisterung für diese Dinge! Alles gute für dich!

[@bjoernphotography]

Sehr Sehr geil - CPUs sind einfach immer wieder faszinierend...

[@taggxoc]

Neat, i have one of those on my desk, didn't really know what it was just thought it looked cool.

[@Killerspieler0815]

This deidding worked nice , unlike modern desktop CPUs ...

[@strife711]

This is like a low-key asmr channel and I am , 💯 interested in everything he does

[@osgrov]

What a beauty! Good job on getting it off the board too, that was rather scary-looking. :) You made me remember a bad day in history, when I failed to purchase a decommisioned RS/6000 workstation off a local business here. It was complete with tons of documentation, peripherals and all sorts of goodies, and fully working too.. Some dude had showed up, paid more than I did and walked off happy.. So much for having a deal eh. Oh well! Keep the videos coming, they always put a smile on my face. :)

[@spitefulwar]

What a behemoth! Sidenote: you are referring to "Inbus" screws. For our american friends you may refer to "Allen screws". ;)

[@diegodonofrio]

Awesome 👌 I have no words. Thanks for this channel, I am loving it ❤❤❤

[@B1G_Dave]

Now it needs a CPU Galaxy rating!

[@HuntersMoon78]

Your MCM CPU collection is gorgeous

[@CPUGalaxy]

Thank you

[@jj74qformerlyjailbreak3]

I 🤤 When I see that Collection. I thought the Pentium Pro Would Be The Centerpiece to my Collection. Now I Want This ☝️

[@chrisbenedictum1]

Hi, love your videos... thank you !!! Humbly speaking (I am old electronics tech (disabled now). I was thinking for the desoldering. Leave the heat sink on and unscrew from the back, with attached... greater thermal dissapation for dssoldering. Maybe avoid the heat sink till the most sensitive part is done... da solder hehee. Love science and tech 👍👍👍 Have a great one !!!

[@Wheatley_Darwin]

I really wish I could find a board like that, really been needing one like you got there

[@aurelioemilianomaltesmunoz9136]

This video was amazing, the CPU is beautiful thanks for sharing

[@Witq87]

Nicely done!

[@izools]

So freshing to see such care taken over this 🙂👍

[@CPUGalaxy]

thank you. 😃

[@PCsrot]

Really nice chip

[@stephenbeets00001]

That is one nice-looking processor. Admittedly, I don't usually like de-lidding processors, though I totally get why some collectors do that. Personally, I'm content to simply leave the heat spreaders on mine. If I was to run into one of these processor cards (not very likely any time soon, if ever), chances are I'd just leave all the chips intact as they are on the board and not desolder or de-lid anything. To me, THE WHOLE CARD as a unit is valuable and with so many ceramic packages that would require special techniques to remove, I wouldn't feel comfortable attempting such a feat. Maybe the tall, sharp-pointed/edged heatsink would come off, but that would be it. I did manage to buy a couple POWER processors of my own recently, none of which exceeded $40 each. I got one each of POWER9 Monza, POWER8 dual-chip module and POWER7 single-chip module, though in a 2,296 pin OLGA and not the bigger TCM-like ceramic one (maybe someday I'll get that one too, if I can find one on eBay for a price less than $50). I have your videos and website to thank for inspiring me to get what I got thus far. Have a great week, CPU Galaxy. :-)

[@ThiagoPerroni]

Better, better than actual BGA soldering

[@mce_AU]

Nice work.

[@KJohansson]

Proper chip "scavenging"! Break cleaning fluid is excellent to remove just about anything fat or greasy.

[@xeveniahdarkwind178]

Gotta love ether aka brake cleaner for those jobs

[@cedric0850]

Awesome! More, more, more, please!

[@PearComputingDevices]

That die looks as if someone took a PowerPC g3 and multiplied the cores.. I had a RS 6000 server once. It had a 604e cpu in it. Only a single core sadly enough but it was neat. IBM was at the top of it's technical advancements in this era. The 604 and G3 cpu was quite impressive as was the slightly less powerful 603 variants. Later with the 970 (g5) they got even more sophisticated. Sure the g5 cpu ran hot, but it was an engineering marvel.

[@krystian7246]

Świetny film, biorę się za następny :-)

[@MegaChrischek]

Das knacken fährt einem durch Mark und Bein ^^

[@CPUGalaxy]

lol. Ja, ich war live dabei 😂.

[@x4700]

This is totally awesome! Thank you so much for doing what you do.

[@CPUGalaxy]

thank you!

[@SkyOctopus1]

Hooray, only eight dies! I was a bit concerned when I saw nine areas.

[@brokeandtired]

Yeah one is just a blank spacer.

[@Privatier-ce6oy]

Thank you - your videos are so cool 😎

[@haltandcatchfire3626]

Great video, thank you for sharing!

[@jeremiefaucher-goulet3365]

I was surprised by the break cleaner. Isn't that a bit harsh/overkill to clean thermal paste? Curious about the rationale here, as I would have expected simple isopropyl alcohol to be sufficient for that step.

[@fox0ps22]

Despite how popular isopropyl alcohol is for removing thermal paste and general PCB cleaning, people don't seem to be aware of the fact that it does react with metals - ranging from the super common (aluminum) to the really exotic... and the results vary from metal to metal (ranging from no discernable effect to so strong that it is part of a metal treating process). I don't know exactly what kinds of alloys IBM used on this chip - but I know they are using at least one very uncommon low temp solder alloy, which is why those caps are notorious for ripping off at room temperature. I don't know which formula the brake cleaner used, or the purity (not likely a concern for brakes - so probably in the "meh /shrug" range), but it does a better job of leaving no trace than isopropyl alcohol (which contains water). It also does a better job on the glue (I'd have skipped the razor and just left the chip face down in a sealed container with a shallow pool).

[@jeremiefaucher-goulet3365]

@@fox0ps22 Thanks! After seeing how corrosive break cleaner is on plastics, I was a bit worried. Granted this isn't plastic.

[@CPUGalaxy]

I agree fully on what you say 👍🏻

[@glennblank1364]

@@fox0ps22 Isopropyl does nothing to the silicon glue that IBM used to attach lids and heat spreaders to their modules. You must use razor blades and then ceramic polish to remove all traces of the silicon glue.

[@fox0ps22]

@@glennblank1364 huh, maybe that is why a wrote a paragraph explaining the logic of using brake cleaner instead of iso...

[@gamingthunder6305]

Its a really nice chip and good work getting it out. but i would prefer the entire machine restored.

[@OzzFan1000]

That's a fascinating CPU! What kind of edge connector is that?

[@adamsfusion]

I think it's a custom socket. These cards were used in the 7015-R20 and 7015-R30 in case that's a not enough info, but I've been having all sorts of trouble digging up any info outside of some really basic IBM documents. This definitely wasn't the first or last foray into IBMs obsession with processor backplanes.

[@InconsistentManner]

I was not expecting that bonk...

[@ryanmalin]

That's one way to de-bolt it!

[@eurocrusader1724]

Your videos are truly masterpieces and are very relaxing.. Tech ASMR! Edit. What will you be doing with the rest of the hardware if I may ask?

[@CPUGalaxy]

Thank you!

[@VincentFischer]

I'm really surprised that the heatsink is mounted without any thermally conductive stuff underneath. Is lots of pressure enough for this purpose or would this be designed in another way if they would made it today? Also what's this red thing in the middle of the side with pins? Some kind of spacer? Beautiful piece of tech!

[@fox0ps22]

They still do it that way - depending on the wattage. Typing this on a power9 that is around 130 watts, cooled by a high pressure mount heatsink. IBM starts recommending a thermal interface (that isn't paste) for anything running hotter than that - indium pads being the go-to solution for a long time.

[@VincentFischer]

@@fox0ps22 Thank you for the interesting insights. Any idea why not paste? they used it very generously on the die afterall :)

[@fox0ps22]

@@VincentFischer I can only guess, because IBM's papers are pretty well paywalled by the journals now: too much of a potentially destructive mess for too poor a result that lasts too short a time. You really don't want to be reenacting this video while reapplying fresh paste after the last application dries out on these machines. Did you notice how wet that paste still looked? Again - just guessing, but that stuff is very old now - and it definitely isn't made of whatever it is that passes for thermal paste these days (where you're lucky if it doesn't dry out after a year of gentle temperature cycling).

I love those old MCM Processors. Remember when I was a kid and the Cray SV2 came out with evaporation water cooling! I have non in my colection, are there ANY cheap entry level MCM cpu's that one could start collecting on a small budget? I use Electronic Cleaner from CRC, or really I use Biltemas Electronic Cleaner, works as well, more for less. It dissolves Paste just like Break cleaner did for you, I think its very similar all of them, petroleum based I would think like refined gasoline almost, (my guess from reading on the cans). I use that on my PPC G4 because the past gets under the plastic spacer so its easier to wash it out this way. Much easier then alcohol. I would love to have a MCM based processor in a working system to run some *Nix system on instead of my peasant Alpha/PPC run of the mill Workstations but I can probably keep dreaming XD

[@wishusknight3009]

I would love to find a complete machine intact to play around with. They had some interesting software options available to them. A stab in the dark for the chips on the PCB itself around the simm sockets. I would guess the larger one by the ceramic package is a bus interface, while the 3 medium sized ones are memory controllers. If memory serves they used a 128 bit memory interface, and the 3 chips would function in a similar way to the first pentium chipset, where 2 are identical and each take half the bus and act as half of the ram channels, and the other does the other half, with the 3rd chip being the arbitration and control logic. The smallest chip on the very far end could also serve a similar purpose and could be bus arbitration or clock splitting with its proximity to the oscillators. . (interrupt/dma controller?) This is just me completely guessing, I have no clue otherwise lol. But this is what makes sense to me based how other systems during that time were architected.

[@wishusknight3009]

Ok according to wikipedia the memory interface for the POWER2+ is 128 bits in 4 buses.. And it would at first glance with it having L2 cache slots that you have a Power2+ there with 128kb Dcache. But it also identifies this as being a 6 chip SBC package and not ceramic. Where as you have an 8 chip Ceramic package and L2 cache slots. So Wikipedia is not entirely accurate. Or perhaps the external cache slots were there but never used on the original power2. I would suspect yours is certainly a version with 256K DCache. The first POWER2 is listed as having a 256 bit memory interface in 4 busses, with the package containing 4 chips for the Dcache, 1 for a 32kb Icache, 1 for storage controller (probably just a fancy name for cache coherency or bus interface chip) , 1 for 2x ALU's or Fixed point maths as its described, and 1 for 2x FPU's. And die size is listed as 1215 mm2 on a .72 micrometer process with a .45 micrometer feature length. So I would revise my guess to the small bga packages on the PCB as the one closest to the CPU being an MMU of some sort in addition to I/O interface chip. The rest I think I will keep as I guessed them before.. back to readinglol...

[@BGTech1]

The closest chip to the cpu is probably a northbridge chip. And the other chips are memory controllers.

[@jankomuzykant1844]

Those are also nice looking chips :o)

[@BGTech1]

@@javiermesa-martinez8731 thanks for the clarification

[@tommyovesen]

Great video. The break cleaning fluid was new for me, what is the active substance in that ?

[@sherannaidoo2712]

Usually carbon tetrachloride. Some formulations use diethyl ether in countries where ccl4 is banned.

[@virtualtools_3021]

cocainum

[@danthompsett2894]

Yep thing of beauty, i geuss since its never gonner be used again it doesnt really matter what cleaning product to use as long as it doesnt damage the look of the cpu dye's, only ive always been advised to use TIM cleaner or 99% isopropanol alcohol not brake cleaner.

[@alessandrosamperi8087]

Bellissimo!

[@EdwinSteiner]

Nice CPU! In the end it looked as if there was still some residue on the dies. Is there something or is it only a trick of the light?

[@necro_ware]

Nice collection! Is your museum open for visitors? (of course after the covid hell)

[@CPUGalaxy]

Thank you. My small museum is entirely private. But you are invited to come anytime. 😉. I would anyhow appreciate a coproduction of our channels. 😅

[@necro_ware]

@@CPUGalaxy Thank you for the invitation! :) And I remember your offer and will definitely contact you as soon as I have an idea, what we could do together. I always watch all of your videos and find them great, but we were acting in different areas up to now. Despite, that we both are into retro computing, I'm currently out of ideas for cooperation. However, I keep it in my head all the time :)

[@fox0ps22]

@@CPUGalaxy Dunno what the tax situation is for you, but in the US you can run a non-profit museum like that, while collecting tax advantaged contributions and writing off expenditures - so long as you post public visiting hours every so often in the local paper and fill out the taxing authority's forms. You are providing a public service, you should take advantage of anything that would lessen the personal financial burden and incentivizes the continued production of this sort of material. I'm not an attorney or a CPA, but you can totally take that as my liability inducing legal and financial advice :)

[@gnif]

Beautiful chip, a shame you used brake clean to remove the thermal paste though as it's very likely what has caused the markings on the dies. IPA is commonly used to clean thermal paste of all kinds while being extremely gentle on the parts (they are washed at the factory with IPA), and is extremely cheap and safe to handle.

[@glennblank1364]

No, the brake cleaner doesn't ever hurt the coatings that IBM used on their dies. The markings are due to thermal effects from the chips on the coatings over the years of use.

[@charonunderground8596]

Amazing !

[@BiggRanger]

What temperature is your heat gun running at? I have an Alpha AXP processor that is soldered to a PCB, and want to remove it. I've tried some other parts, but end up cooking the board, so I'm running my heat gun too hot.

[@Damien.D]

Wonder what the blue gunk is?

[@onometre]

that cracking noise got a wince out of me lol

[@CPUGalaxy]

lol. me as well. Its hard to make a video and being excited disassembling sthg like that at the same time 😄

[@rsuryase]

how do you dispose of the liquid waste?

[@MJ-uk6lu]

So pretty!

[@WooShell]

I'm surprised the brake cleaner didn't etch away the blue lacquer. IMHO that's far too corrosive liquid to use on such delicate parts.. I'd have tried it with isopropyl alcohol first. Still, absolutely amazing result. I've got several Power4 MCMs (the smaller dual-core only, not the chunky palm-sized ones) that I need to clean up, but it looks like IBM sintered some sort of 5mm thick graphite heat conductor plate onto it. If anyone has an idea how to remove those, I'm glad for every hint.

[@glennblank1364]

The brake cleaner doesn't hurt anything on these IBM ceramic MCM's. It's no where near corrosive enough to damage ceramic and metal or the coatings that IBM used on their dies. The ceramic covers on Power 4 MCM dies can be easily removed with a heat gun set on high for a few minutes. After the cover comes off you'll need to scrape the cement off of the dies with a sharp razor blade while the cement is still hot.

[@exaltedb]

Would the heat gun method work for removing a PowerPC G4 cpu from its daughtercard (for the sake of collection)?

[@CPUGalaxy]

yes of course.

[@exaltedb]

@@CPUGalaxy thank you for letting me know. It’s a dual-processor card and I think both of them would look cool in my collection

[@Иван-ъ2ы4б]

Красивый, но система охлаждения очень замороченная можно было cpu поставить на примую на радиатор например медный никилерованный а на кристалы жидкий метал или туже терморасту с добовлением металла я думаю это было в разы эффиктивние

[@ErraticPT]

Nice chip. Its just a pity these aren't practical to run as part of a system in the average home environment.

[@CPUGalaxy]

yeah, thats indeed a pity...

[@jevgenijobzigailov2376]

Today's AMD Threadripper would look similar with heatsink removed, and even with 3000$ price per CPU it's a fraction of IBM beast price, no matter adjusted for inflation or not. And still not practical for average home user.

[@warrax111]

Hello, can be am5x86-p75-x5-133adw downclocked to 33x2 Mhz? Somehow? Motherboard is socket 3 , Asus PVI-486sp3. Or there is no way, and 25x3 = 75 Mhz, is lowest ?

[@izzieb]

I've always been curious about IBMs server POWER processors, but I have absolutely no reason to buy anything with them as I feel it'd be a waste - better bought by someone like yourself who can appreciate them more.

[@stephenbeets00001]

I found a POWER9 Monza on eBay for a mere $39. No guilt at all. Never too late to get into them should you re-consider.

[@GeomancerHT]

Isn't pointing components on a motherboard with a graphite pencil extremely dangerous?

[@marcinpl3157]

Only if the board is powered.

[@virtualtools_3021]

if it is made of carbon-14 maybe

[@phillycheesetake]

How funny that the world is moving back to MCM's.

[@charonunderground8596]

Maybe its time to Collectible Chip Rating for Pentium Overdrive 83Mhz ?

[@radarmusen]

I happen still I have to reboot one of them, funny to see apple and ibm copyright on the same time.

[@jordydrop]

quite the specimen

[@Clancydaenlightened]

14:00 note if you can use some liquid flux😉, would've been a bit quicker amzn.to/2ZyvzpW Or amzn.to/3bCa519 If you're lazy 🙃

[@loqAtMefi]

Future computer historians 100 years into the future: "Well, here we have another bare early multi-chip module from IBM but we're not sure which system board it belongs to since every example we can find has been unmounted by a chip collector and all the documentation was lost in the global computer virus pandemic. Anyway, look at how big those gates are. You can very clearly see them without a quantum field effect microscope and little more than a magnifying glass. As you know your personal embedded computation module has more than a billion times more transistors."

[@ruthlessadmin]

Leave it to IBM to make an expensive CPU that can't be replaced independently of an even more expensive board.

[@yeslatino1668]

8:19, there is other way to do this .... :D

[@e8root]

reminds me of Pentium D 9xx, just 4x times better :)

[@CPUGalaxy]

😄

[@mireknovacek9680]

Why we dont have such cpus in todays computers?

[@KonuralpBalcik]

Rare

[@heylel1841]

Thank God it didn't into gold scrapes hands

[@CPUGalaxy]

yeah. thanks god

[@uhohwhy]

big black clock

[@phrtza]

Čistič brzd :)

[@jensdroessler3575]

I like Italian paste better!

[@vladimiro8171]

Sacrilege

[@rameshleo21]

Worst CPU ibm, sorry sir that's Truth