This CPU was used in the Nuclear Power Plant Fukushima Daiichi?
ฝัง
- เผยแพร่เมื่อ 18 มิ.ย. 2024
- Find out More about Hetzner Managed Servers:
htznr.li/der8auer/managed-en
---------------------------------------------------------
*this is an affiliate Link!
Support me on Patreon:
/ der8auer
---------------------------------------------------------
Music / Credits:
Outro:
Dylan Sitts feat. HDBeenDope - For The Record (Dylan Sitts Remix)
---------------------------------------------------------
Paid content in this video:
- Hetzner Spot
Samples used in this video:
- /
Timestamps:
0:00 Intro
0:52 Hetzner ( Advertising)
1:34 Research on the CPU
4:15 Is the CPU real?
6:51 330.000€ per month?
8:55 The CPU in detail
11:02 Delidding the processor
12:10 The CPU cores
13:45 Summary/conclusion
16:32 Thanks for the support
16:58 Outro - วิทยาศาสตร์และเทคโนโลยี
I am a nuclear engineer with 30 years designing control systems for commercial nuclear plants. These BWR plants are quite old, and yes GE built the early ones. There was a technology transfer to the Japanese so the later units were progressively more domestic. I can assure you that the control were originally analog. It’s is possible that this computer was installed as and upgrade to the original equipment due to obsolescence. This computer probably didn’t operate the safety controls. These have been left analog where possible, but plant computers were installed for monitoring. Your dad was correct about changing out.
any chance they may have used a system like this for calculating optimal fuel loading / shifting bundles / re-fueling? Seems like this system might have come from a pre-cluster computing era, so maybe complex compute like fuel decay may have been something this was used for. I did maintenance work at a PWR years back, and it was still pretty analog for the controls as well so I'm not sure what they'd use a mainframe system for from an ops standpoint. Just spit-balling here, interested to hear an engineer's perspective.
can it run crysis?
@@stevemke at westinghouse most of our calculations are still deterministic. really does not require much compute to run programs like tracworks
@@Linuxfy enough with your can it run crysis non sense B$
Can it run Doom ?
I used to work at IBM on mainframes, and on an IBM mainframe CPUs come in pairs, the computation is duplicated and compared and if the results do not match that pair is turned off, a spare pair is turned on, and if the CPU book (what the CPU module is called) is low on spares a support technician is dispatched to replace it.
Assuming Hitachi are doing the same thing the 10 chips are probably compute, the 5 chips are probably L1 cache for each CPU pair and the other 7 are IO. That module would then be providing (at most) 4 usable CPUs of compute.
You can't find out the usable:spare ratio from the part number because that's configurable for the use-case. For this use-case I would expect it to have been deployed as 1-usable 4-spare because you would not want to get into multi-processing problems in something this critical, you don't need 2001-era multiprocessing for a 1967-era problem, and you would want maximum spares for that environment.
You are of course correct that millions a month would be a lot to spend on a normal CPU, but that's not what you are buying: you're buying near-absolute certainty that every calculation is done correctly and on-time.
Fun fact: there are microcontrollers with duplicate cores and ECC memory, for applications requiring high reliability (automotive) and high radiation environment (spacecraft).
Such microcontrollers aren't that expensive - TMS570 series starts from tens of dollars up
Wouldn't want a Single-event upset to screw up data.... like a boeing 737 max deciding to nose dive because the ex-McDD bean counters didn't think a second data check was worth the cost.
@@mieszkogulinski168 you can get even lower, there are c28x mcus that has ecc and lockstep. the tqfp48 variant of the f2800156 goes for around 2.8usd for qty under 100 if you buy direct from ti
@@mieszkogulinski168 Wow, I didn’t know about these. But I have wondered, how probable are these errors in something the scale of a tiny embedded microcontroller operating in normal conditions? Do cars actually need ecc and redundant core chips?
@@abraarsameer9521 It's important to see these things on different perspectives. A normal CPU would be perfectly fine for your car, however if you sell million cars then suddenly a one in a million error event would start happening on a daily basis, which then would be a heck of a problem for any car manufacturer: A system error in a car wounds a person every day. The same logic applies for nuclear power plants (and many other things as well), but otherway around: a one in a million error event would affect million humans, meaning the risk is too high for the error to happen even once. Aviation has a lot of great examples and practices about this too.
$6mi/year for a critical device isn’t too unusual in certain businesses. I work on the medical industry and one of our machines used in the research labs and hospitals worldwide is leased, not sold, also by 7-digit figures per year. This gives the researcher an opportunity to refresh it when the lease expires, and is way cheaper than buying the instrument and later figure out what to do with an obsolete piece of equipment.
Also, a lot of supercomputers do run in nuclear reactor areas, for the power consumption and cooling requirements - guess where you find plenty of both?
yep, worked for philips as a imageonology technician, new wireless xray detector cost like 60 Millions CLP (minimal wage here is 450-500k CLP), a aprox 62k USD detector that i they dont last 1-2 years bc ppl are dumbass and doesnt understand that medical or any super precise instrument or parts, COST A TON bc theyre built in a super high precise manufacturer bc they have to works TONS of times with minimal errors
Modern NPPs built since 2000 actually use computers for what operators does BUT most of them still using analog IO and digital computer processing is only for quick response of the sensors and monitoring. What i've found was, the modern NPP using up to 5 different redundant IO and cable management, one line comming from the one PC , second from another and so on... , so for example you dont have one computer but 4 of them and the fifth is obviously Analog and much more reliable. But nowadays every NPP uses computers to monitor reactors's behaviour fast enough, not like a Chernobyl's SCAM where you must to wait good 30 seconds for the specific responded request and that also caused disaster. Now you have responses instant fast and precise. But also every analog output must to go through the computer to compare wheter is good or bad for the reactor so for example you can not pull all controll rods from the reacotor core, even when you shut down the main safety system there is another one to prevent such a catastrophe what had happened in Chernobyl or TMI
And also when you still gonna do such a thing and actually u did it, it will send all the input output data to the storage center far far away from reactor and also to the main CEO of that NPP and nuclear energy goverement so everybody will be informed just in case what could happen so they can be prepared for the worst case scenario.
@@camialehviva chile
It's never cheaper to lease! more convinient if the equipment regularily needs to be upgraded sure, but from an economical stand point it makes no sense to lease. The entire value of the product is paid down over the leasing period so it's exactly the same as buying new equipment after every leasing period + you have to pay a significant amount of interest. The only benefit is that you spread the cost over multiple years. Realisticly equipment and technology also lasts much longer than that of a leasing period before being obsolete. We want you to lease products because it allows us to regularily repeat business
It would be interesting to x-ray this cpu. You might get insights about the function of each tile.
The old process should make details easier to see.
Absolutely not, X-Ray has nowhere the resolution to tell you that.
@@FrozenHaxor sure, but specialized x-rays should be able to reveal blocks of contrasting densities which might clue us on their function. Also the substrate wiring should surely be visible since the traces there are fairly thick.
For the X-ray the silicon is completely transparent.. even for the super high resolution X-ray machine's..
You can use a good resolution X-ray to see on the ceramic layers and wires but nothing else..
@@ricardolmendes No. You can put a piece of silicon in front of highest resolution detector and it will act as a piece of paper. You need to section it, then metalize and look using a scanning electron microscope to get anything useful. There simply isn't enough difference in microstructure density to give you enough information through X-Ray analysis. All you would see at most would be bottom metal interconnects, which again, won't tell you much as it's mostly power rails.
Week out, week in I'm baffled "this one German dude" that runs a company and has tons of other things to do makes the most interesting videos on the side. German efficiency at it's best.
Respekt. Genießen Sie die Feiertage. Grüße aus Kroatien 🖖
Thanks :) you too
His videos have German build quality. I'd buy anything he makes tbh.
@@FirmB1ade true that.
Roman can spin off an antique pc show-and-tell channel.
Measuring radiation? that would probably be a Geiger-Müller counter. Hans Geiger was German, but device is still called the same in English
He's done a video before where he didn't know how to say dosimeter.
probably meant Zählrohr
Not only - there are also ion chambers (very simple, but not that sensitive) and proportional counters. Also, regular photodiodes can detect gamma rays
So is Alzheimer's.
When something is name based, it's not translated for obvious reasons.
@@incumbentvinyl9291 but that's probably not what he meant, Geiger-Müller measures something specific, a Zählrohr can measure more relevant things.
I worked on mainframes starting in the late 80's. We used ICL Fujitsu stuff. They never disclosed the pricing as everything was leased. We went to client server distributed computing in the early 90's using Unix servers. When we had the mainframe, the ICL engineers had their own office on our site. After we had dumped the mainframe and were clearing out the engineers office we found a confidential price list. The nodes, which is what they called the processors, the top of the line nodes we had were £10M each, and we had 2. These things were massive thou, about 8 foot tall, and about 5 foot square.
I love mainframes. Big iron is still so important. While working for the Canadian government I learned a bunch about it - so awesome and surprisingly cost effective.
This is a close relative of the Fujitsu - they kinda co-stole the tech from the System 390. Mum used to work with one in the Australian Bureau of Stats
Yeah they get BIG. Imagine a set of 20 or so of these nodes in a tower, with all the storage space to feed data into the CPU cluster at a rate fast enough to get your money's worth out of paying for the whole thing in the first place. Plus all the networking gear to get that data on-site in the first place. That's why the things are cabinet sized by the time they are actually working in the real world.
The node thing is because in design for a long time these things were basically discrete cpu's that just happened to share a mounting surface and cooling solution. There's a bit more too it than that considering they are running different instruction sets and doing redundancy checks and I'm sure I/O routing is a nightmare on these things, but essentially, its somewhere around 4-16 different full cores and the supporting infrastructure for them. The nomenclature has stuck around even in the age of multi-core computing because they still have multiple of what we would normally consider a full CPU on the die.
Back when I was getting some training on mainframes for IBM, I think I remember them calling their "hot swappable compute package" "pucks", they were just a node like the one in the video, though with some design considerations for the hot-swappable part. Also yeah let that sink in, hot-swappable CPU clusters.
@@zippydaspinhead I'm happy just with hot swappable RAM, hot swappale CPUs are insane, but an entire cluster hot-swappable, amazing
This was super interesting and I had no idea such large CPUs exist. Thanks for the video.
Check out Tesla Dojo cpu. one of the newer 'generations' from elon
Always remember what you have at home as CPU - GPU - RAM etc.
In the computer industry the single component is 100 to 1000 times larger.
For example search for "Cerebras WSE-2" or "IBM 305 RAMAC 1956 5MB".
As you see IBM is always at the largest scale with all things related to computer technology, you would almost think they are not from earth.
Another thing to think about is, how is TH-cam storing all the 10* Peta Bytes of data weekly that humanity is making on their platform?
Do you think they are using just regular enterprise HHD/SSD for that or use a super secret storage box that is big as a car?
there is an soc called the Cerebras wafer scale engine, and as the name implies, it is a wafer scale chip, meaning it uses a whole entire wafers worth of silicon for its die. I think its like 20,000mm2. The second gen chip uses TSMC 7nm hpc and packs over 2 trillion transistors.
@@crowntotheundergroudThat's awesome, thanks for sharing. Hilariously, they have a table on the product page comparing their wafer chip to an Nvidia A100, and they cite their 56X larger chip size as an "advantage." 😄
"The most obvious thing to start with is Googling anything that's written on the CPU" That's what I had to do when I was doing electronics recycling. Annoyingly it rarely helped for anything actually interesting we found. That label on it made me instantly know it was going to be the best thing ever, or leave me questioning why it's even there in the first place. It's even more fun when they don't even have the manufacturers name on it. Good times...
an easier way to find out is to get a Geiger counter
@@MikeZak101 If it was inside a server case then it'd be unlikely to register other than normal
@@MikeZak101 huh? Not sure why you're mentioning that since it has nothing to do with finding out what you actually have...
my bad, didn't realise server cases were radiation proof@@WyndStryke
Its an IBM system/370/390 clone, whole industrial espionage thing. If you start on the wiki for IBM MVS you will get some clues.
Asianometry definitely would be able to find out more details on this CPU!
The engineers that created those things were superstars of their time.
"Superstar" implies they got some serious recognition from their work but in reality no-one knows their names or who they were. Pioneers? Maybe, but definitely not superstars.
@@anteshell such is computing
@@anteshell How about slave labor. No credit for work done.
@@user-xn4cy8js7c Slave labour means getting completely inadequate financial compensation for the work or not at all. "No credit" is not synonymous to "no compensation", which means it does not imply slave labour.
In fact, I'd rather take monetary pay over my name in credits any day of the week, because my name on a page 152 on a manual in small print doesn't pay for bills.
The common name in English for a measuring device for radiation is a Geiger Counter (or Geiger-Müller counter more correctly). Another name thats more agnostic of the measuring method is a Scintillation counter.
He was probably confused because it's Geigerzähler in German, but as both Geiger and Müller are German names, they rarely translate this directly into English.
Also, if this thing was still so radioactive, there would be no way you would get it shipped.
@@p_mouse8676 I mean does anyone at the post office truly check that?
Absolutely they DONT
@@cromefire_ in modern sorting centers it just goes automatically through a machine. Especially when it has to go into an airplane.
It's entirely possible they had a CPU that was made after the plant was designed, as mainframe-type systems were designed with insane backwards-compatibility. They could have upgraded the hardware to something more readily available without having to modify the software running on it or any of the secondary systems or peripherals. In some cases this can even be done one CPU at a time in a larger system, while it's running.
It's probably a clone of an IBM mainframe build on licence
If it's a late upgrade (well yeah, it's going to be considering the building era cpu's were slot type :P) some of those cores are probably for emulation / translation processes to make it compatible with the other old hardware / software stacks, right?
@@libaf5471 I think it's very unlikely it is using emulation. That would create a massive software dependant single point of failure and completely unneeded layer of complexity on top of any possible hardware defects. Not a good design choice for something as critical as nuclear powerplant.
less of a license, more of an IP theft sting and lawsuit, but yeah, @@ciano5475
Theres a few details on the IBM MVS wiki page. Hitachi and Fujitsu ended up making systems compatible with old school IBM mainframes even after IBM changed to z/arch
@@libaf5471 Slot was a solution to solve problems toward mainstream market. This module was for a very spécific market where price wasnt a problem. You can find pretty old hardware still running in specific cases. Because they do peferctly the job + a new hardware would require dev, time, money, tests, and more dev, time, ... So it's easyer to keep spare parts.
Control system engineer here (DEH and DCS especially for nuclear power plant). they need to upgrade their control system due to lack spare parts. of course there are other reasons, like better protection, better control outcome. this cycle typicaly is 20-25y like you said, but NPP user does not like cutting edge tech, they prefer product has been on the market and been tested for quite some time. and before the retrofit actually implemented, at least 3 years designing time needed. that Cpu or technology could be already 5-10 years old when it finaly been used or implemented in the NPP. so i think the last retrofit (which use that big mainframe cpu) was carried out at about 2000. evidence? PGA on the CPU maybe? i'm not saying that cpu is 100% from fukushima, just saying that make sense.
This CPU is such a cool piece of tech!
I would love to have it and put it on the shelf...
As always: Very nice video. Thank you for the insight into some technical history of these kinds of cpus.
Thank you for all the great videos this year and happy holidays to you and all your family. All best for 2024 also btw.
Thanks again for all the great content! Looking forward to more in the new year.
Fascinating stuff as usual! Always glad for a shout out to the POWER family. Not a lot of those were anywhere close to $6M, but some of the biggest were a lot of fun to work on.
Definitely needs a follow up to scan the dies on this thing and fix the pins and create an acrylic plate to go around them, it might not be usable again but can be preserved and documented! someone might pay you even more to put it in a museum!
Thanks for your videos, attention to detail, and down to earth approach to the items and topics you cover. Appreciated from Sacramento/LA - California
This is a great insight as to what was occuring with CPU design and pricing in that era, thanks for the indepth information and research
This video is the perfect Christmas present!! Thanks a lot, Roman.
Looking forward to the Cinebench results😉
Super cool find, thank you. Merry Christmas.
REally a nice one! Thanks and wish you a great year ahead, too!
What an awesome piece of history! Love the video, nice work.
Best tech channel alongside with BZ's. Always liking your videos first then proceed to watch them :D
The units I work on used PDP-11s for the control computers built in the 80’s. They were then upgraded to use emulation in the early 2000’s, so could have been used as a midlife upgrade.
Were they Kent P4000 running the K90S operating system.
@@macman6421 No. forget the name of the original computers, but they run a custom OS built just for the plant I work at. I knew all the details at one point, but those Cheerios fell off the stick long ago
@@MartycraneI was thinking of the system they had at BNFL
@@macman6421 I’m on CANDU’s. the computers got upgraded as I was starting so I very briefly saw the old stuff. Still used reel to reel for some programs we had to load in infrequently. Same with paper logging. That’s all gone now though, it’s all emulation
Very cool item, nice work getting it opened and cleaned up.. I wish you a Happy New Year also Roman.
Awesome find and thanks for sharing. I love obscure stuff like this.
Excellent Video! Always appreciate your insights.
I have studied electrical engineering with specialization in powerplants 2004-2009. The stuff we learned at technical university was how to steer the PP using the manual switches and dials. Basically the green cabinets that you show @3:48. In 2008 with about 20 of my colleagues we were for a month allowed as interns into the biggest conventional power plant in central Europe that was build in the 80's. We probably went through all the departments there, but I do not recall seeing such mainframe computers. We did spend few hours in the steering room, and they used computers for steering the systems (green cabinets for manual operation were still there, as well as fish tank - which is mandatory thing at every steering room). But no pictures were allowed and I don't recall anything except modern at that time LCD monitors, mouse and keyboards. The software they were running looked modern for sure (fully colorful graphs, diagrams representing technological flow with steam parameters, gauges). So I'm certain they did retrofit from fully manual steering into computerized steering and they could make additional upgrades / retrofits every several years. They have scheduled maintenance periods, so if power plant has let's say 10 blocks than one or two may be in maintenance (they may replace turbines, or generator with newer one, cutout entire boiler and put in new one and at the same time change the steering system), when this block goes back to operation they will do the next one and so on. The change for entire power plant may take like 5-10 years to complete. I don't know much about nuclear power plants, and computer steering was never mentioned at the university in my day, but at the same time I don't feel there would be a need to have a massive mainframe system to operate it. I mean you have hundreds of sensors to read information from, but how many refreshes per second you actually need? When in manual operation maybe someone went to do the readouts once per hour and it was enough.
A fellow electrical engineer here, but with a focus on micro electronics: Such a CPU isn´t used for its compute power, it´s purely redundancy. I suspect that the ten large chips are all individual cores, functioning in pairs with the same L1 cache chip. Each calculation is made in both cores of a pair and the output is compared. If the result from both cores in a pair is the same, all is good. If it´s not, that pair is cut from power and the calculation is repeated in the next pair. When the second to last pair is reached, the CPU will be replaced asap to keep the redundancy in place. That way the complete CPU would be able to spit out errorless results for tens of years. In this case I wouldn't be surprised if some of the cores in this CPU never saw any real use besides a periodic errorcheck, as the whole design of the cores themselves would've already been focussed on integrity rather than speed.
For nuclear at this scale you absolutely need mainframe class computers to monitor and operate the core. Nuclear is much more complex to operate than any other type of power plant.
@@binaryguru Yes, a nuclear power plant is complex relative to other power plants, but don´t forget this is a ´94 era CPU in a ´70 era power plant. This CPU was most likely hundreds or thousands of times more powerful than whatever it replaced. I wouldn´t even be surprised if the original plant design didn´t even include a digital processing unit and was purely analog. Development of these nuclear power plants doesn´t take several years as derbauer suggests in this video, but rather takes tens of years. Nuclear reactors being comissioned today run on handware from early 2010 or even the zeros. This means that the original system in the fukushima power plant is from the 50s or 60s, well before the first real CPUs were launced (Intel 4004 in 1971) and bizarrely simple compared to what we are able to do now (or in 1990). To add to that, while a nuclear plant is much more complex than a coal plant for example, a coal plant is utterly simple and even the processing required for a nuclear power plant isn't very complex either by modern standards. Furthermore, these systems are kept as simple as possible in an effort to reduce the risk of errors. The risk of errors grows exponentially with complexity, so every added parameter or processing step has disastrous effects on reliability.
Hehe. Steer the pp.
That is crazy and thanks for the video and entire year of videos. Frohe Weihnachten
HDM is probably "high density module". I found a very similar item on an auction site. Compare the serial number sticker with yours. Look for "Hitachi F7/MP5600 High-Density Module 21A XSL01-D DH-PF75-A1M1 Rarity Rare TOP"
I find learning about parts like this very interesting.
Being a machinist who makes mechanical shaft seals for nuclear power plants and actually seeing the computer parts that run the equipment is awesome!
We love you R! This is the kind of content i love to see the most, thank you
Frohe Weihnachten! This is very informational.
This was fascinating!! Thank you 👍
Frohe Weihnachten und ein glückliches neues Jahr, Roman, auch an die Zuschauer und Familien.
Asianometry video vibes. Was captivating from beginning to end. Thank you and Merry Christmas 🎄
so cool man thank you for your awesome videos and doing them in english!
@der8auer-en The pad pattern of the missing chiplets in your Hitachi CPU are identical to the pad patterns on the CRAY X1 Multi Chip Module shown on the CPU Galaxy channel. It seems very unlikely they would be identical unless the same IC was intended to be placed there. See frame at 4:31 in the Cray video and compare with 12:47 in your video. I grabbed the frames and compared them. The pattern of missing pads around the center of the pad pattern is a 100% match. There must surely be a story here, maybe even one never exposed before.
Amazing content, congrats!
Very interesting video! Merry Christmas from windy old England!
I love the engineering part of things so love what this channel has to offer 👍
Super Cool . . . . Merry Christmas, Roman.
I am so jealous when folks like yourself end up with these cool old server/mainframe/supercomputer CPUs. They are both engineering and objective works of art. Even if that is not from Fukushima, it is still a wonderful piece of history to have, and it is a shame more information is not freely available. However, given the longevity of these systems and the places they are often used, it makes sense why we in the public still know so little. Thank you for sharing some history and hopefully more can be learned about this beast as time goes by!
"Now let me just de-lid this---" *BOOOOOMM*
Got Demon Core vibes with the screwdriver and everything...
Maybe steaming ham with it is a better option?
Happy holidays Der8 and DerCatz
Lovely Christmas present
12:30 Looks like its symmetrical, theres 4 sets of processor modules on each side, pairs of IO in the middle so they can be shared and memory controllers up either side on the edges
So cool, happy holidays Roman!
Thank you too Roman, for all the great videos 😉
Super awesome video!
Super interesting video mate.
Absolutely Fascinating Stuff!
Merry Christmas Roman to you and you partner/family!!
From the mid aughts to just before the pandemic, my employer was a contractor working at Oak Ridge National Laboratory in the US. One of the teams we worked with at ORNL was tasked by the US Department of Energy to help with the Fukushima disaster. Generally, they were helping analyze the data coming in from Japan and making suggestions for containment and such. That's about the extent of what I was allowed to know -- as you mentioned, the nuclear types are touchy about their secrets. As a contractor, I was tasked with maintaining some of the hardware being used to review and discuss the simulations they were running. It was kinda cool to learn from the end user of my (very small, indirect) role in figuring the disaster out. And, as a personal point of pride, that equipment I maintained ran trouble-free for the duration of their high-profile project. 😊
I don't know these items, but I want to have one because of you xD
Thanks for the video and Happy Christmas.
grate videos all year round.
Good video, fun to see other than Intel and AMD cpu. Wish you happy holidays and new year. 👍🏼
The most popular secure way is actually to run 3 of them. I have no knowledge of Nuclear Plan setups, but this is the most common setup in general.. could also be, that other reactor's Mainframes worked as one of the backups. Having this setup its actually very easy to hotswap Mainframe HW, so its not as big deal as you say. Usually in Mainframe everything is hot swappable and anything works with back compatibility since 50's.
Well, it’s cool that you have this because like I feel like this would be something you would see from only like really really big tech youtubers
Thanks for your sharing
Very cool stuff , love the huge cpu sizes. Same with modern ampere it’s big and awesome
Was hoping in some benchmarks lol, but very cool mate ty
That chip is beautiful, thanks for sharing.
So I wouldn't be over concerned with the cpu date and the commission dates. Something this big was probably on the design phase customer list for the Hitachi guys. They may have even provided input to the design.
DerBauer playing Homer in the nuclear plant control room lol
Less funny is that Germany shut down all their nuclear power plants after Fukushima due to Green outrage, and is now in permanent outage and has the most expensive and polluting electricity in the world...
But hey, Happy Xmas !
Fascinating video :)
Merry Christmas and a Happy and Healthy New Year.
its been a while since i watched a video of yours and man the beard looks good on you
You're awesome my German brother... Have a marry Christmas
wow! what an incredible cpu. a true marvel of engineering. your very lucky to have gotten this chip
Thank you for the good content.
Cool. Thanks for sharing.
Merry Christmas and a happy new year Roman
It's cool thanks for buying stuff like this for us
Amazing subject that was so interesting, thx you
Enjoyed the history lesson, thanks.
Fascinating video.
A very interesting and thought provoking end of year video.
An eye opener to custom silicon and a world away from what we are used to.
There's a CPU that's made today that goes all the way to the end of the wafer. Cerebras Wafer Scale.
Be sure to let us know how it does on LN2
Cool topic, I'd be interested in seeing more of the non-retail hardware that's out there. Cheers and Merry Christmas Roman and team!
See you soon Roman :) Amazing videos. Thanks for all
This was fun, thanks.
@der8auer : Roman, the 36 (6x6 grid) tiles (small, medium, big cores + other stuff) in the CPU block, could they have been installed on a per-customer basis ? (so that in theory Hitachi could have many different versions depending on the needs of the customer, i.e. more or less big cores, more memory, ...)
Awesome and super interesting video. Wonder if there was a lease price discount over the long period of time this CPU was operational.
Damn! 10K for a CPU!? I wonder how the seller came up on that price. Hopefully you can verify the authenticity of it being used and is from the Fukushima Nuclear Power Plant. Keep us updated Roman.
well considering they had to rent it for 200K per month when it was new i guess it a steal.
That's true. 👍@@hallgrimurkhallgrimsson1200
Considering its a rented CPU that was supposed to be returned to the manufacturer, it's age and that it cost millions to run this might very well the only example of this CPU in circulation in the whole world. Worth the price I would say.
@@NoiseStaticBlur wouldn't that mean that there is no demand for theese things, therefore completely worthless? i would pay like max 100$ for this. but in reality a video is enough.
wow nice. Thank you. I think this is the first time I've seen such a CPU
Very interesting Video D8R.
Ah yes, SuperH. The architecture of both mainframes and calculators.
I love the difference between the German and the English thumbnails , the dollar must have dropped in value some 😂
wdym, they same?
In the German thumbnail he seems to reference the exact monthly cost, where in the English version he seems to just reference that it's in the millions quickly.
Wow what a cool relic and just so precision waterblock wow
I worked in communications in the military and we regularly had to inspect and certify nuclear shielding for our ground based electronics to make sure they worked in the case of a nuclear blast. I also happened to be stationed in Japan during the 2011 disaster and saw a lot of footage of the inside of that reactor room. They kept their electronics in similarly shielded HEMP hardened enclosures so I doubt any EM or ionizing radiation got in or out of the computer systems, even right next to the reactor.
Happy holidays Roman🎉
This is really cool!!
This would legitimately be such a cool museum piece.
Ok now I want to hear more about your dad sounds like a very interesting man
happy holidays