*Addendum* - The 486 came out in 1989 (which is what the copyright on the chip said!), while the 386 came out in 1985. Not sure how I goofed that up. Cheers to the viewers who pointed out the error :) - Consensus seems to be a diffusion barrier for the titanium, either TiW or TiN. I did see some tungsten in the EDS scans, but wasn't reliable enough to feel comfortable sharing. Might have been tungsten vias, or part of the TiW. Unclear! But we can probably say the titanium was for diffusion prevention. Thanks folks!
You did say "heavy metals" in the intro, but the metals you mentioned, Al and Ti, are not heavy. Perhaps you had meant to discuss W originally? When I heard that, I wondered what heavy metals are involved in these processes.
@@JohnDlugosz Ah, you caught me :) To be honest it was mostly just a nice aesthetic phrase :) But I _did_ have a section talking about tungsten that I found during EDS analysis, but ultimately cut it since I wasn't 100% confident on the result (although based on conversation here in the comments, seems likely it was TiW acting as a diffusion barrier. Or potentially just plain tungsten, which is used as via contacts to the transistor layer since it can withstand high heat). So we can pretend that's what I meant hehe :)
Titanium 7:45 - sounds like a primer. Titanium dioxide, white paint, primer, popular on the macro world - just enjoying the symmetry - Thanks for the clip
@@jacknolan6170 every comment here except for yours is utter garbage, thus can be used as an argument to say that no, we are not all as smart as the people that developed these....
@@FarmingwithZana101 So do it... No you can't because they don't do it magically, they DID it with a LOT of work, THEORETICAL physics, and dense mathematical tool... And OBVIOUSLY, not "everybody" is good in physics and in maths... Obviously, you're part of those idiots, because the first clue that someone is an idiot is "everybody can do it" "I can do it" "it's easy"... So we spotted you.
Around the time the 486 was being manufactured I worked at Intel in the memory components division in the yield analysis group. I worked in the area where we analyzed the electrical signatures of the failures. But I worked closely with the technicians that did the physical analysis. They had to be able to peel back the layers to see underneath to find and identify what the yield limiters were. They had a recipe book full of chemical methods to strip off the multiple layers that the chips were built upon. One recipe to strip off the passivation, another to strip off aluminum. Etc. It was pretty common back then to have a dip in yields that needed to be fixed quickly. Samples of the memory chips would have their failures cataloged. There were many cases where the particles that were contaminating the chip were analyzed in detail. Once a particle was found in the Scanning Electron Microscope, it would be bombarded with the electron beam. The SEM had a mass spectrometer attached that picked up the elements of the contaminate that was knocked loose by the electron beam. They would look at the mass signature of the particles and try to match it to a catalog of various materials in the factory. Sometimes they could tell right away what it was. Lots of carbon, lots of organics: Scotchbrite. Someone improperly cleaned a piece of equipment with a Scotchbrite pad. Other times it would be some new alloy they had not seen before. They would trace it to a hinge of a door in a wafer handler that was emitting particles every time it moved. The particles would land on the wafer and render the chip inoperable. It was an endless detective game.
Back in college one of my instructors told us a story of a yield drop that they traced back to a contaminant that came from a tech who had started wearing a new perfume. I would hate being one of the people doing failure analysis. I've had a couple opportunities to see some of the work they do and it's fascinating and I'm glad there are people with an aptitude for it, but I'm much happier designing the circuits.
Chip designer here, I am amazed by the amount of hand layout done in the late 80s to construct a processor. Today most of the functionality is automatically placed and routed with a ratsnest of wires going all over the place. This design is beautifully engineered and many thousand of man hours went into such a intricate layout.
@@EricBarthDev I don’t care what you think, but I can tell you that you are not contributing anything which improves the discussion. Hint: this is a comment section on a TH-cam video rather than a scientific paper and the audience is different.
This just feels so unreal. I can't even begin to wrap my head around how people came up with, and managed to achieve this. The scale is unfathomably small, and this is 30 years old! Had I not been sitting at a computer typing this, I wouldn't believe that this was even real. This is mind boggling to say the least
i look at these similar how minecraft redstones work, you start knowing the rules of redstones of what it can do, to certain materials and block heights etc. then you expand it it feels fimiliar like that except in a smaller scale
This is because we imagine people using tiny tools to construct each tiny wire and stuff but in reality this is more like stamping, or masking. They use UV light and certain chemicals to burn/eat away parts of a silicon plane (wafer) using cleaver masks that only let certain UV light pass or certain chemical reactions to happen, in certain parts of the wafer, is like taking a picture but at very very tiny size. This to me may remove some of the magic in making of this things but imo is not less impressive, the fact humans were able to come up with this stuff in less than 100 years, is mind blowing.
My first computer was a 486, purchased almost new for $1000 in 1996. I loved that thing. I bought memory for it at a computer show, opened it up, and installed it inside. I miss that machine.
I too bought my first 486DX for about $1000, had MSDOS 6.1 (faulty double space version) and spent another $500 on a VGA monitor (15 inch) and a 9 pin color dot matrix printer. I was king of my block back then. RAM was pricey back then because of a fire in Taiwan burnt down a RAM manufacturer. Started with 1 MB RAM (remember it was DOS). I eventually upgraded to the motherboard max of 4 x 1 MB sticks at $75 each. Then cam DOS 4GW and I was off to playing DOOM. Back then you could get free games in the box just for buying a CDROM drive. I chose the one where the Panasonic sound card also had the IDE driver built-in for the CDROM. I think I still have the ISA board somewhere.
Ditto, my first PC was a 486 DX-2 66Mhz, VESA 16 bit graphics card, only 4 MEG ram, though i did buy another 4 meg ram for £126 ( a bargain as the going price was around £150)....great for Quake
Tip: LCD screen. especially the colour ones. It's easy to magnify each pixel of one. But how do the layers really look like? the sections of the liquid crystal and the colour filters.
@@kkkloginPositively and negatively charged types of some semiconductor material are fused together, creating a PN junction. So a diode only really has 3 layers, the PN junction itself and the case. Some smart guys figured out that some semiconducting materials can glow. So those give us red, green and blue LEDs (and their derivatives). The blue LED is particularly important because it is the foundation for the white LED that is so ubiquitous nowadays. And that's about it, I find optoelectronics a bit boring.
What's more mind-blowing is the disposability of such effort and sophistication. No thought is given to the utter wonder, power and complexity of any modern device.
This is the kind of thing that runs through my head when I see the incredible complexity and design of technology even made 30+ years ago. It is so insanely easy to take everything we have for granted. Processors and devices simply become obsolete when the next shiny thing comes out, and people toss them in landfills or e-cycle grinders without so much as a second thought. It feels like a waste of human accomplishment. Despite their obsolescence, I still find it hard not to appreciate the unthinkable amount of effort put into developing these kinds of things.
I think gamers would do well to remember this rather than just blindly consume new things all the time. Being able to play a simple game like pong is a technological marvel, most of us will never understand how advanced it is. No need to buy the new call of duty/FIFA every year, the one you have currently is already a very advanced piece of software. Same goes with PC hardware, you dont need to upgrade to the newest Nvidia card every generation.
I am a layout engineer and I work on 3nm technology, pretty advance from this but It was so cool to watch those metal lines and vias placed, especially those titanium and aluminum contacts for source and drain!
I was a night manager in Intel. They had the masks for these chips displayed on the walls. I used to sit and stare and try to follow the Lions. It was cool.
Nice video. I am an IC Reverse Engineer and you actually almost succeeded with viewing the transistors. In fact, at 8:04 in your video, you can see the polysilicon gates (the visible tracks on the video) of the transistors inside some digital logic bloc. You can see around those tracks dots that are kind of outlining where the transistors' active areas are. Those dots are called contacts and they are tungsten vias linking source, drain and gate of the transistors to the upper interconnect layer. I can see some inverters and buffers there :-) Great job!!! I read my own comment and it sounds quite technical, sorry about that..
Oh, that's awesome! Thanks for letting me know! I saw some Tungsten on the EDS but it was a very small amount and not very consistent so I didn't want to mention in the video. But I was _fairly_ certain it was contact plugs down to the transistor layer. Super cool! Out of curiosity, do you know if there is a way to chemically etch such that doped regions become more apparent on SEM or optical? I've been looking around but haven't found anything definitive. Some references to HNA etching _faster_ in doped regions, but not as a way to add contrast to an image.
@@BreakingTaps You'll have to etch dopants to see them clearly. That's a technique used to read some ROMs for example. You can check dash etch recipe as a starting point. The chemicals you'll need are pretty nasty, be super careful ;-)
@@reivilo535 thanks! Will look into it...and see if I'm comfortable with the chemicals or not 🙂 good chance I'll pass of they involve HF, have had enough of that in my life I think 😄
@@Imkrhn Except they aren't 140 times smaller. The 7nm process node does not have 7nm features - they're larger. The process node name (7nm) is now just a marketing term and hasn't represented an actual measurement since 0.35 micron in 1995. This slide into marketing gibberish has gotten so bad that TSMC and Intel (and possibly others) have decided to stop describing their chips this way.
I work in CD-Sem for Intel currently. This is very interesting to see. It's pretty crazy to see what used to be state of the art vs what we see nowadays.
This is from a time where the majority of cars on the roads were carburetted, with some still being produced that way: Fuel injection in most cars was a single injector atop a, well, carburettor.
While in the US Navy, being a Sonar Tech, we had to repair our own equipment. Was a security clearance issue…ET’s didn’t have that high of one. In that capacity, I was sent to 2M school. Mini-micro mini repair. I was soldering integrated circuit chips, circuit boards, etc., with the aid of an intense microscope. The solder tip looked like a baseball bat and one little shake was enormous. Great video! 🫡
The last figure I saw, not too long ago, was we're up to 15 billion transistors on a single die. That's 15,000,000,000. A typical 8 bit microprocessor in the 1980s in our home computers contained around 9000 transistors
@@deang5622 Apple’s M1 Ultra has over 100 billion transistors, however that is not impressive - the chip is enormous. What is more profound is the 1nm transistor size made by some MIT researchers a few years ago. That’s 2 orders of magnitude smaller (roughly) than the current transistors used. That’s around 10 hydrogen atoms in width, which is an insane achievement.
@@binaryalgorithm total number is not the right metric to be impressed by, it is transistor density. Still extraordinarily impressive, but Cerebras uses a whole 12” wafer for that count, not just a single die (about a few cm^2). That said, today’s technologies are using transistors with several nanometer gate lengths (although that’s not quite half pitch, for a number of technical reasons). What is almost more impressive is that we can fabricate 100T transistors on a wafer and nearly all of them work. The yield is impressive and unimaginable (failures < 1e-8 or fewer).
I had so many 486 chips and motherboards as a kid, but I never knew how unbelievably complex they were at the microscopic level. It gives me a whole new respect for chipset architects and engineers, a group of people I didn't know I could respect anymore than I already had.
When people think of exploration, they probably think of going somewhere like outer space. But sometimes the most amazing places to explore are the tiny landscapes and structures that are so often a mere arms length away from us, which we never even notice are there.
I like how you explain all the features because it makes the video a lot more interesting when we are seeing the components up close. Makes it feel more magical
This is fun to see, as it brings back memories. On the day of my in-person interview at Intel, I was given a tour of what was then Fab 5, where a technician bored with the usual fab tours showed me the copyright etched into the silicon on an electron microscope, at the time the chip was manufactured new in 1997. I worked in Fab 15 and Fab 20, until I left in 2005. I was a technician in diffusion, which was depositing very thin layers and oxides, or as we liked to say, "making silicon rust".
Although my comment here appears to be a year late, I wonder whether anyone in the know could tell us whether miniturazion has merely provided greater speed, or actually enabled computations previously very difficult, or whether if it were possible to run a i486 100+ times faster if it could match the computational capacity if a more modern chip.
@@dannypal123 Modern microarchitectures absolutely do enable computations that were difficult on earlier CPUs, particularly through hardware acceleration. AVX, SSE/SIMD, AES, etc.
This is impressive. The fact this tiny structure was manufacturable 30 years ago means to me technology is way more advanced than I believed it was. Now, they're on a nanoscale level (7 to 14, and soon as low as 4), meaning everything you see there is actually between a hundredfold to a thousandfold smaller. So they can litterally put over 100 486 CPUs in the same space than a single 486 CPU. And that's not accounting for the 3rd dimension: add several 10s more layers, and you're in the 500-1000x density range. Awesome.
Nothing on a 7 nm chip is actually 7 nm. On a 0.350 micron process the gate pitch and the minimum metal pitch are both 0.350 micron. This has never been the case since; with RAM optimizing for minimum metal pitch and CMOS logic optimizing for gate pitch. Since FINFET the no dimension anywhere on a chip is even close to the node name and they're not even very comparable between manufacturers.
Maybe you just aren't old enough. Most people started getting computers in their homes in the early 80's (although the Altair 8800 came out in the mid 70's.) They were little more than toys even when compared to this 486. However, they were truly amazing when compared to how we had solved problems for thousands of years
@@matthewbanta3240 My age is fine thank you very much. This CPU is still younger than me. I just never had the opportunity to see how small those silicons already were at that time, and that's just putting everything technological into perspective to me. I used to be a machine designer, so I am well aware of precision manufacturing techniques. But this is still order of magnitudes smaller and more precise than what was achieveable in my domain 20 years ago. And that's the fun about learning new stuff: despite looking like an idiot to the people you're expressing your awe to, you end up growing as a person nonetheless.
I always find content like this fascinating. Many things have been reduced to a Nano scale both manmade and in nature. It would be great to see more content like this.
I'll never forget when I was working on bare ICs at Hughes Aircraft (we were making multichip modules), and I was bored and looking at an IC with a high power microscope. Suddenly I saw a "happy face", circle with eyes and mouth, upside down. I turned the whole thing around, and sure enough, there was a very distinct happy face on the die. Of course I told my coworkers about it, and they just smiled. IC makers put lots of easter eggs on their chips just for the fun of it.
This is fantastic content! The EDX overlays are really cool. I do have a few tips that I’ve picked during my time as a metrology engineer… Allied High Tech’s Epoxy Bond110 works really well for mounting protective glass slides to your cross-sections. (Their diamond polishing films also work really well). Proper polishing takes time you can’t rush it. Pt or Au sputter coating also goes long way. Fun note about AFMs… if the object you are trying to image/measure is smaller than the radius of your AFM tip, you are actually measuring the tip itself. We would use this phenomenon to determine when a tip’s useful life has ended.
Rad, thanks for the tips! Will look into that, I mistakenly assumed a random bottle of cheapo 5-minute epoxy sitting around would be sufficient... and then was horrified how many voids there were and how poorly it held up to polishing. :) Did you mount to the glass slide and then polish both the slide and the sample simultaneously (e.g. the slide was part of the cross-section)? Or was it just used as a mounting substrate but the cross-section was just sample + epoxy?
@@BreakingTaps I would polish simultaneously. I’ve found when using this epoxy, less is more. A small drop of epoxy then clamp the slide to the sample for curing.
Small correction: A finished chip is not "a wafer". It's a chip. A wafer is the term used for the circular silicon substrate (typically 300mm in diameter today) that chips are made from. They're "photographed" onto the wafer in repeating patters, as many as there is room for, one layer at the time. Once all the layers are complete the wafer is sent for testing, cutting and packaging.
I remember my 486sx from the 90s it was my first computer. I ran a B.B.S. called Gateway. It had only 4 telephone lines running at 14.4bps. This is totally nostalgic 😊
It's an unfair comparison. The fly's processor had 3 billion years to evolve to where it is right now. You need to compare it to a human made processor made 3 billion years from now.
@@BoyFromMa I don't think we are late to the game. We are working on it, but why? is this something we used to do? we might be lost trying to get back to the Science/Creation Barrier.
@@delphicdescant Oh but the Fly incorporates multiple processors, Bios and CPU with flash. Don't for get the Ultra low voltage rechargeable power supply. A damaged wing? it will rewrite data in the flight computer to compensate. On The Fly!
This is absolutely the most beautiful video i have watched in TH-cam, which up until now was Life beyond by melody sheep for me. This is also the first video of yours i have watched and it made me sub. Amazing work!
I started work in a Failure Analysis Lab 40 years ago and worked with microscopes and basic tools. A few years ago I returned to the work and now have access to a SEM. You're right about the polishing process. It's a skill and an art that requires a lot of experience to become good at it. I don't have that much either but there isn't much demand to see chip defects in the OEM electronic manufacturing arena these days. My predecessor did this type of work 20 years ago and used plasma etching equipment and chemical etches to remove the IC layers. Removing the layers requires knowledge of the device's construction as one of the other commenters has pointed out. Otherwise you're experimenting and hoping for hte best.
Fascinating! Thank you for showing this. I got my start on the Internet with a 486 based computer. Things learned from the Internet have been equally important as my formal education. This started with my first computer with its 486.
when looking only at the transistor count (i know, wiring and stuff couldnt be scaled down 1:1), you could actually put close to 20.000 486 dies in place of one AMD zen 3 die. its 20 billion vs one million transistors. or the other way around: you would not be able to see the 486 die with your naked eyes when done in modern lithography, because it would only be some µm² in size. even the optical microscope would have trouble resolving anything useful. and yet, even the original 486 feels like sci fi magic when thinking about it. its totaly astonishing.
Extremely cool! I work for the company that manufactures your SEM. Primarily I work with the image data after they have been acquired with an SEM, FIB-SEM, TEM, etc. Nowadays a FIB or Broad Ion Beam (BIB) are often used for milling and polishing the surface before imaging. You can even repair a bad circuit with what we call a Circuit Edit system, which is basically a special purpose FIB-SEM. What you are doing on your channel is one of my dreams and inspired me to purchase an old Hitachi SEM at auction. It will take some effort to get it working again, but your hobbyist results keep me interested in the process. I wish I could own one of our own microscopes, but it is not a charity, lol. They are pretty expensive. Good luck with your new channel - subscribed!
How on earth have I never found this channel? Amazing writing, pleasing voice, *brimming* with information, which is eloquently and intelligently conveyed, and the video length is *just* right to give information and kindle a yearning for information in me, which made me research this further. Keep up the good work. Thank you.
This is so incredible. I’m amazed by how small and complex it is and how many transistors it has. And now I think my mind would be blown away if I see the new ones as even each year so many improvements are being done and now the chips have reached 3nm process. There are really genius people out there and I still can’t believe how they can manufacture these without an error so we around millions of all these individual transistors. Amazing is not enough as a word.
Nope every wafer not always perfect. Sometimes some core cannot function and the cut by laser. And manufactur selling less core chip. Or sometimes even the chip function properly they consume more or less power and they called sillicon loterry
Thank you so very much for allowing to see inside one of the best micro-processor chips ever made. I'm an electronics major and really appreciate the splendid beauty of these wonderful devices. Yeah, it's now old school stuff but it's still a work of art. Like the Mona Lisa !!! 💎💎💎
In a past life, I was a tech which my job duties were to maintain 10 beautiful Zeiss microscopes plus a small desktop SEM within a geology research facility. Although I never polished a chip down, ( I could have as we had a Thin Section Lab also ) I did view a photo sensor array, at the limits of visible light which required immersion oil to do so. It was quite the eye opener! Just the pixels of an array needed 1000x magnification to view them. Incredible! Thanks for the journey.
Wow. I remember building and programming 8086 computers back in the day...16k of ram and 5.25" floppy disk drives. Amazing how technology has jumped over the years. My best project was a 486DX40 with 512kram and an 80meg disk drive, a 5.25" and 3.5" floppy drive. Ah, the good old days. Thanks for this incredible video.
@Jason Fitch 16k was not a lot on an 8086. The 8086 could address 1MB of RAM. 16k was where the first PC started at (though it used an 8088 - the 8 bit I/O version of the otherwise identical 8086).
512 K of RAM on a 486? Man, even my shit-ass Deskpro 386 SX had 1024k. On a DX-40, I'd assume at LEAST 4 Mb RAM. Maybe you meant 5120k RAM? I've seen (even recently) weird setups like that - I have a 286 in my basement with something strange like 4768k of RAM.
@@the_kombinator If you re-read my post, I said my best project, not what the system was capable of. You really do not need to be so critical, I did not have unlimited financial resources to build a maxxed out system.
Beautiful shots... chips from close r truly both amazing technological achievement as well as artistic, thx for this vid! Jewelry made from chips look also incredible😉
As a hobby photographer and someone who loves tech (especially pcs) this is crazy, I have never seen stuff like this so close up, and every render from 3:18 just looks like a perfect image. It's chaotic. but also somehow organic, it is really fascinating to me. Thx for the experience
Science is a fractal beauty of infinite complexity. I wish I had been exposed to this when I was in grade-school... 31yo and three years into my engineering degree, I guarantee I will never look back. Only forward to a passionate life of learning, with dreams ever dominated by fantasies of earning that elusive "PhD" next to my name. The world is a fascinating place.
Seeing the elemental colormap of the cross-section makes me realise how well structured and carefully designed the whole process is. Seeing silicon and oxygen was intriguing (but not that surprising), but seeing the aluminium microwires having a coating of titanium underneath was eye-opening. And seeing that the Electron Microscope is able to discern between atomic elements is something that blew my mind! 🤯 The amount of planning and work involved made me really appreciate the sheer complexity of microchips!
oh yes. If there was a lofi or ambient/ space music soundtrack for footage like this, that would be some awesome background study ambience, and some badass yoga background ambience =]
AMD's Zen cores are very interesting because instead of straight or right angles, the chip features blob-like structures as a product of AMD's High Density Library. I hope you guys can take a close-up look of Zen soon.
Credit where due, I stole this idea from Wendover Productions :) They recently launched a second channel (Extremeties, I think it's called?) and did a similar setup: had a few videos in the new channels backlog, and aired an episode on the "main" channel to help advertise it. It seemed like a nice way to launch a channel so I poached the idea :) Thanks for watching!
Minute Correction: The 80486 (DX) was introduced in 1989 not 1985 (1985 saw the advent of the 80386). The 80486 (SX) chip in the video actually was never available until 1991 when Intel offered a cost reduced version with a defective, disabled or (later) omitted FPU. The copyright is still that of the original 80486 (DX) from two years prior and the inscriptions carried over.
Very interesting stuff. The technical details of these things has always been shrouded in mystery and obtuse academia - sparse charts and illustrations, but not actual examples. This is so close to being an actual breakdown of the process - I only wish to know more about how these things were actually manufactured. It's mind blowing to think that vias could be done on a silicon wafer. How much is even silicon anymore with all those other metals added? How even were they added, aligned, printed? Incredible stuff, and a great video.
3D? Blender? I guess it's CAD based and designed, even if it's 3D. Blender or anything like that would likely crash instantly trying to show this amount of elements, vertices, polygons, shading etc. Not talking render here.
Great video. I would like to suggest that it might be easier to understand what you're referring to if you have a pointer or in some way highlight the feature you're referring to. Often there are many different features visible in an image and it takes me a long time to determine which one is correct. In grad school I examined thousands of SAR images (synthetic aperture radar) and could instantly recognize features that took people who weren't immersed in satellite remote sensing a bit longer to understand.
That's a great point, thanks! Will try to incorporate more pointers/arrows/indicators in future videos. Appreciate the feedback, trying a new format is are always a little scary! ❤
That's an understatement! This full size desktop processor has 3.3 million transistors and the newest iPhone has almost 16 billion and is smaller than a pinky nail, while using vastly less power. A top end cell phone literally has more power than the fastest supercomputers of just a couple decades ago. The fastest supercomputer in the world in 1999 had about 15 percent more processing power than an iPhone 14. Insanity.
This is mind boggling. I have seen similar microscopic views before, but this brings the circuitry into much better detail. I just can’t imagine how these lines and connections are laid down (or etched, I guess). In the future, I hope you’ll do a video on this process.
This video is too fascinating, it's like when I was analyzing a ram memory module and in the program I could separate the layers and see how everything was interconnected through the tracks, it's like a small futuristic world of the kind we see in the movies, really fascinating, Greetings from my beloved Venezuela.
There were some people that reverse engineered the 6502 and z80 with just some microscopes. these are MUCH MUCH simpler 8 bit single layer CPUs however. still impressive. one of the projects was called "visual 6502" recommend checking it out some time
Yep, what Piipolinoo said. EDX/EDS. The electron beam, particularly when cranked up to max intensity and accelerating voltage, will kick out characteristic x-rays from the sample. And the energy of the xrays emitted is specific to each atomic element. So by blasting the sample with the beam, you can collect xrays from each location and categorize semi-quantitatively what elements are present. There's a lot of nuance, a lot which I don't even know yet, which makes analysis tricky. Detector location, how flat the sample is, how deep the beam is penetrating, overlapping spectra, etc etc. But for some things like the aluminum interconnects it's easy to determine what's going on. Didn't want to talk about it too much in the episode, trying to keep that channel more "mainstream"/documentary-like. I'll probably talk more about the technical aspects of the SEM itself on this channel in the future though!
@@BreakingTaps but I thought you were doing these with your scanning probe microscope that pokes things with a small needle, I didnt realize it was a SEM. You have also shown the wiggling needle so I was pretty confident you somehow poked their chemical composition with a needle lol.
@@the_kombinator - beyond that, is repeatability and testing... which involve mind boggling integers at this scale.... both in terms of high binary operation count, and the miniscule physical size.
@@driverjamescopeland That and also, if there's a mistake in a subsequent pill, you lose 4 cents. If your wafer isn't clean or there's a manufacturing defect on a die, well, you just threw out a lot more than 4 cents. Having said that, I'd love to get a failed 3 or 486 substrate.
@@the_kombinator - I didn't know until recently, most of those failed substrates can't be recycled into new modern chips. They can be used for larger resolution transistors, but not microprocs. It's simply not economically feasible to use anything but virgin silicon for new microprocessors.
Very cool and zen look at the microworld. A note on the afm scans, it seems your tip was likely not tracking the traces as it fell off the tops of the metal edges. You can often spot this when the trace and retrace scans are not very similar especially when coming off sharp edges. This can be improved by either slowing the scan speed, decreasing the tapping set-point voltage, or increasing the gains on the PID loop.
Yeah, I noticed that as well. In this case, I think the sample itself was slightly tilted with respect to the tip (and/or the cantilever's arc). The forward and reverse scans actually look pretty similar and not much on the error map. So my theory was that the sample was tilted slightly, allowing the tip to access one side of the vertical wall but not the other wall (since it would be "shadowed" by the overhang). But I could be wrong! Still learning to spot artifacts and their causes :) Cheers for the advice! Will keep that in mind for future scans.
@@BreakingTaps That's true, it could be due to the angle of the cantilever or the tip geometry. Many tips are also asymmetric along the cantilever axis allowing the tracking of steeper walls on one side.
Lovely - a minor correction, the 486 came out in 1989 (as shown on the copyright on your images!). The 386 came out in 1985 but would have been a bigger processor manufacturing node I believe.
Oof! How did I miss that?! Thanks, will start up an addendum for corrections. I assumed the copyright was about the "SX" version, not sure why I assumed it was wrong hah
@@BreakingTaps Lovely video, thanks! Actually, the 486 SX was introduced in 1991, as a modified version of the 486 DX with the FPU disabled. The 1989 date seen on the wafer is probably the manufacturing date.
@@ofloveandliquor Introduced in 1991 but manufactured in 1989? ;-) Nope, Intel doesn't sit years on their wafers. It's simply the copyright date for their 80486(DX).
Yep I remember looking at the back pages of Computer Shopper magazine (almost as thick as a phone book!) back in 1989 looking for the best deal on a 486 motherboard/CPU combo - the prices were such that today (even adjusted for inflation) you could buy a nice complete gaming rig for the price of a 486 motherboard+CPU back then! I also remember buying a 387 FPU coprocessor for my Zeos 386 around the same time! The 486 was the first Intel to integrate the two on chip with over a million transistors!
Thanks for showing this! I have always admired the i486. :) It blows my mind how advanced microprocessors had already become at the end of the 1980s! The i486 DX4/100MHz is an absolute BEAST of a CPU under Windows 95 (with 32MB 60ns RAM and 256k 15ns cache). Not even joking. The i486 DX4/100 is so powerful that it will do *true* polygonal 3D graphics (it'll play Descent and Descent 2) all in software at 35 fps _without_ hardware acceleration with only a 512kb 16-bit ISA VGA video card under MS-DOS. 8 slot (1x8 bit ISA, 4x16-bit ISA, 3x 32-bit VESA) 25/33/40/50MHz (adjustable by jumper) Socket 3 motherboards are the best motherboards ever made! As for your video narration, it would be much more listenable if you moved further away from your microphone and spoke louder. Narration is an art form I've studied for decades, and the best style that suites a general audience is a smooth, yet vocal and gritty, authoritative tone (Mythbusters narrator is an absolute prime example). Speaking softly and calmly close to a microphone and letting your voice granulate isn't how professional narration is done. Try taking my advice because I love your videos. :)
younger people today have no idea how much work a 486 is capable of. I can't think of anything we do at our office, other than the demands of today's internet, that couldn't be accomplished with a 486 just about as well.
This is so goddamn impressive. The machinery and skill that is needed to create such detailed and tiny things Is far beyond everything I could imagine!
Amazing to see this up close. It still boggles the mind to know that this was thought up by a human. I had a 8088, 286,386 (with math co ) and skipped the 486 and went straight into the first Pentium ........Cool times. Editing the Autoexe bat file, config sys and making sure my 640 KB or RAM on my 386 (w/math co) spread out correctly. I bumped it up to 720 KB after installing 19 chips on a ram board.
I bought an expansion board for my Tandy1000 (8088) that gave me 640 KB. That was so much RAM that I used some as a RAM disk that I copied my compiler onto so I could edit and compile programs without switching floppies.
Does the 486sx not have easter eggs? I've taken micrographs of a i486DX and it has a block full of initials, other initials hidden in the traces, and the big 486 etched in one area. That 486 had the most graffiti on it, the 386 and pentium 90 I imaged didn't have any. Another cool chip to image is the Weitek 3172A, that had entire names etched on empty spaces. Last chips to have removable covers were from the late 90s, after that they started making them solid for better heat transfer.
Oh, it did! I found the big block of initials ("JL") and the 486 logo. Didn't see the hidden initials, but will go peruse my stitched scan to see if I can find it. I actually captured some _really nice_ images of the initial block... but lost the data :( The next day I was imaging some other parts of the chip and the software was happily reusing names and overwriting my prior images, so I lost a bunch of stuff. Unfortunately only realized after I had embedded in epoxy and started grinding away at the chip.
@@BreakingTaps If the 496 logo is at the top and readable, the larger somewhat hidden set of 3 initials are in the large trace to the right and a little down from the block of many initials. There's also something written in the lower right corner.
*Addendum*
- The 486 came out in 1989 (which is what the copyright on the chip said!), while the 386 came out in 1985. Not sure how I goofed that up. Cheers to the viewers who pointed out the error :)
- Consensus seems to be a diffusion barrier for the titanium, either TiW or TiN. I did see some tungsten in the EDS scans, but wasn't reliable enough to feel comfortable sharing. Might have been tungsten vias, or part of the TiW. Unclear! But we can probably say the titanium was for diffusion prevention. Thanks folks!
I think TH-cam unpins comments when they are edited. Just in case this was intended to be a pinned comment...
@@max_kl TIL! Thanks, will start double-checking after editing
You did say "heavy metals" in the intro, but the metals you mentioned, Al and Ti, are not heavy. Perhaps you had meant to discuss W originally? When I heard that, I wondered what heavy metals are involved in these processes.
@@JohnDlugosz Ah, you caught me :) To be honest it was mostly just a nice aesthetic phrase :) But I _did_ have a section talking about tungsten that I found during EDS analysis, but ultimately cut it since I wasn't 100% confident on the result (although based on conversation here in the comments, seems likely it was TiW acting as a diffusion barrier. Or potentially just plain tungsten, which is used as via contacts to the transistor layer since it can withstand high heat). So we can pretend that's what I meant hehe :)
Titanium 7:45 - sounds like a primer. Titanium dioxide, white paint, primer, popular on the macro world - just enjoying the symmetry - Thanks for the clip
Every time I see how detailed and complicated chips are I am reminded how incredibly smart the people who figured out how to manufacture them are.
When you fall into imagination nothing is difficult they are humans like us, this means if they can do it we can do it aswell
@Joe Duke man how do people manage to bring politics into every. single. thing.
@@jacknolan6170 every comment here except for yours is utter garbage, thus can be used as an argument to say that no, we are not all as smart as the people that developed these....
@@FarmingwithZana101 So do it... No you can't because they don't do it magically, they DID it with a LOT of work, THEORETICAL physics, and dense mathematical tool... And OBVIOUSLY, not "everybody" is good in physics and in maths... Obviously, you're part of those idiots, because the first clue that someone is an idiot is "everybody can do it" "I can do it" "it's easy"... So we spotted you.
@@garryiglesias4074 Indeed. Most, but not all, of those who post comments on TH-cam would be hard pressed to make a sandwich.
Around the time the 486 was being manufactured I worked at Intel in the memory components division in the yield analysis group. I worked in the area where we analyzed the electrical signatures of the failures.
But I worked closely with the technicians that did the physical analysis.
They had to be able to peel back the layers to see underneath to find and identify what the yield limiters were.
They had a recipe book full of chemical methods to strip off the multiple layers that the chips were built upon.
One recipe to strip off the passivation, another to strip off aluminum. Etc.
It was pretty common back then to have a dip in yields that needed to be fixed quickly.
Samples of the memory chips would have their failures cataloged.
There were many cases where the particles that were contaminating the chip were analyzed in detail.
Once a particle was found in the Scanning Electron Microscope, it would be bombarded with the electron beam.
The SEM had a mass spectrometer attached that picked up the elements of the contaminate that was knocked loose by the electron beam.
They would look at the mass signature of the particles and try to match it to a catalog of various materials in the factory.
Sometimes they could tell right away what it was. Lots of carbon, lots of organics: Scotchbrite. Someone improperly cleaned a piece of equipment with a Scotchbrite pad.
Other times it would be some new alloy they had not seen before. They would trace it to a hinge of a door in a wafer handler that was emitting particles every time it moved.
The particles would land on the wafer and render the chip inoperable.
It was an endless detective game.
Fascinating details, thank you!
I've definitely wondered about how contamination vectors were detected in a busy clean room.
id love to see an in depth video about all the different practices and solutions for contamination control, if anybody's familiar with one
sounds like a cool job
@@barfoom Air filtration. Protective suits and hoods and overshoes. Air showers on entry to the clean room.
Back in college one of my instructors told us a story of a yield drop that they traced back to a contaminant that came from a tech who had started wearing a new perfume. I would hate being one of the people doing failure analysis. I've had a couple opportunities to see some of the work they do and it's fascinating and I'm glad there are people with an aptitude for it, but I'm much happier designing the circuits.
Chip designer here, I am amazed by the amount of hand layout done in the late 80s to construct a processor. Today most of the functionality is automatically placed and routed with a ratsnest of wires going all over the place. This design is beautifully engineered and many thousand of man hours went into such a intricate layout.
@Anno Kitsune you're a bit late lol
Why do I not believe you ? oh, it's because there are actual terms for "rats nest"
@@EricBarthDev I don’t care what you think, but I can tell you that you are not contributing anything which improves the discussion. Hint: this is a comment section on a TH-cam video rather than a scientific paper and the audience is different.
@@EricBarthDev what's a rats nest
@@EricBarthDev bro thinks "rats nest" means an actual rats nest
This just feels so unreal. I can't even begin to wrap my head around how people came up with, and managed to achieve this. The scale is unfathomably small, and this is 30 years old! Had I not been sitting at a computer typing this, I wouldn't believe that this was even real. This is mind boggling to say the least
i look at these similar how minecraft redstones work, you start knowing the rules of redstones of what it can do, to certain materials and block heights etc.
then you expand it
it feels fimiliar like that except in a smaller scale
@@AAGWGEGEGEFEG 😆 stupidest thing ive read on youtube for awhile
This is because we imagine people using tiny tools to construct each tiny wire and stuff but in reality this is more like stamping, or masking.
They use UV light and certain chemicals to burn/eat away parts of a silicon plane (wafer) using cleaver masks that only let certain UV light pass or certain chemical reactions to happen, in certain parts of the wafer, is like taking a picture but at very very tiny size.
This to me may remove some of the magic in making of this things but imo is not less impressive, the fact humans were able to come up with this stuff in less than 100 years, is mind blowing.
Yes but the first prototype design has to be drawn by hand. @@Argoon1981
Definitely crazy, and now your phone has a hundred times more power.
My first computer was a 486, purchased almost new for $1000 in 1996. I loved that thing. I bought memory for it at a computer show, opened it up, and installed it inside. I miss that machine.
I think we had one? Zenith Data Systems. Windows 3.1? 1,000 was a lot back then I'm sure.
@@ayliniemi Yes! Zenith. And After Dark screen saver.
I too bought my first 486DX for about $1000, had MSDOS 6.1 (faulty double space version) and spent another $500 on a VGA monitor (15 inch) and a 9 pin color dot matrix printer. I was king of my block back then. RAM was pricey back then because of a fire in Taiwan burnt down a RAM manufacturer. Started with 1 MB RAM (remember it was DOS). I eventually upgraded to the motherboard max of 4 x 1 MB sticks at $75 each. Then cam DOS 4GW and I was off to playing DOOM. Back then you could get free games in the box just for buying a CDROM drive. I chose the one where the Panasonic sound card also had the IDE driver built-in for the CDROM. I think I still have the ISA board somewhere.
it was mind blowingly better than the 386
Ditto, my first PC was a 486 DX-2 66Mhz, VESA 16 bit graphics card, only 4 MEG ram, though i did buy another 4 meg ram for £126 ( a bargain as the going price was around £150)....great for Quake
Tip: LCD screen. especially the colour ones. It's easy to magnify each pixel of one. But how do the layers really look like? the sections of the liquid crystal and the colour filters.
please include the LED and the different layers in it.
@@kkklogin The LED is a diode. So you need to understand how pn junctions are constructed.
EINK also interesting. And MEMS ICs. They looks great.
I'm very impressed with the fact that you can get a panel which can display billions of subpixels at once each individually for little money.
@@kkkloginPositively and negatively charged types of some semiconductor material are fused together, creating a PN junction. So a diode only really has 3 layers, the PN junction itself and the case.
Some smart guys figured out that some semiconducting materials can glow. So those give us red, green and blue LEDs (and their derivatives). The blue LED is particularly important because it is the foundation for the white LED that is so ubiquitous nowadays. And that's about it, I find optoelectronics a bit boring.
What's more mind-blowing is the disposability of such effort and sophistication. No thought is given to the utter wonder, power and complexity of any modern device.
This is the kind of thing that runs through my head when I see the incredible complexity and design of technology even made 30+ years ago. It is so insanely easy to take everything we have for granted. Processors and devices simply become obsolete when the next shiny thing comes out, and people toss them in landfills or e-cycle grinders without so much as a second thought. It feels like a waste of human accomplishment. Despite their obsolescence, I still find it hard not to appreciate the unthinkable amount of effort put into developing these kinds of things.
I think gamers would do well to remember this rather than just blindly consume new things all the time. Being able to play a simple game like pong is a technological marvel, most of us will never understand how advanced it is. No need to buy the new call of duty/FIFA every year, the one you have currently is already a very advanced piece of software.
Same goes with PC hardware, you dont need to upgrade to the newest Nvidia card every generation.
Now that is something I say all the time
I am a layout engineer and I work on 3nm technology, pretty advance from this but It was so cool to watch those metal lines and vias placed, especially those titanium and aluminum contacts for source and drain!
is there anything here that you specifically see as being inefficient or "bad" relative to what you see today?
It still amazes me how far we’ve come and how the tools have progressed. It’s still a cool job to have!! Floor planning is no joke !!
I was a night manager in Intel. They had the masks for these chips displayed on the walls. I used to sit and stare and try to follow the Lions. It was cool.
I would do the same 😆
Nice video. I am an IC Reverse Engineer and you actually almost succeeded with viewing the transistors. In fact, at 8:04 in your video, you can see the polysilicon gates (the visible tracks on the video) of the transistors inside some digital logic bloc. You can see around those tracks dots that are kind of outlining where the transistors' active areas are. Those dots are called contacts and they are tungsten vias linking source, drain and gate of the transistors to the upper interconnect layer. I can see some inverters and buffers there :-) Great job!!! I read my own comment and it sounds quite technical, sorry about that..
Oh, that's awesome! Thanks for letting me know! I saw some Tungsten on the EDS but it was a very small amount and not very consistent so I didn't want to mention in the video. But I was _fairly_ certain it was contact plugs down to the transistor layer. Super cool!
Out of curiosity, do you know if there is a way to chemically etch such that doped regions become more apparent on SEM or optical? I've been looking around but haven't found anything definitive. Some references to HNA etching _faster_ in doped regions, but not as a way to add contrast to an image.
@@BreakingTaps You'll have to etch dopants to see them clearly. That's a technique used to read some ROMs for example. You can check dash etch recipe as a starting point. The chemicals you'll need are pretty nasty, be super careful ;-)
@@reivilo535 thanks! Will look into it...and see if I'm comfortable with the chemicals or not 🙂 good chance I'll pass of they involve HF, have had enough of that in my life I think 😄
This i486 was built to the 1000nm process node, today's chips are laid down using a 7nm process, 140 times smaller
@@Imkrhn Except they aren't 140 times smaller. The 7nm process node does not have 7nm features - they're larger. The process node name (7nm) is now just a marketing term and hasn't represented an actual measurement since 0.35 micron in 1995. This slide into marketing gibberish has gotten so bad that TSMC and Intel (and possibly others) have decided to stop describing their chips this way.
Wow! Please, please do a series on the transistors over the years!
No I already know what transistor is it like on or off 1 or 0 very ez 👍
@@AboveEmAllProduction everyone simply knows what you said ez, we just need a microscopic view of transistors over time from 130nm to 5nm these days
I work in CD-Sem for Intel currently. This is very interesting to see. It's pretty crazy to see what used to be state of the art vs what we see nowadays.
This is from a time where the majority of cars on the roads were carburetted, with some still being produced that way: Fuel injection in most cars was a single injector atop a, well, carburettor.
what I find really interesting is how much smaller they've gotten everything to be over the years
This was so long ago, a heat sink was _optional._
@@nickwallette6201 On a DX, sure. On a DX2, you'd better have that AND some active cooling.
@@the_kombinator Pssst... this is an SX. :-)
While in the US Navy, being a Sonar Tech, we had to repair our own equipment. Was a security clearance issue…ET’s didn’t have that high of one. In that capacity, I was sent to 2M school. Mini-micro mini repair. I was soldering integrated circuit chips, circuit boards, etc., with the aid of an intense microscope. The solder tip looked like a baseball bat and one little shake was enormous. Great video! 🫡
This has to be the most exquisite introduction to an i846 I've ever heard xD
I find it mind-boggling that electronic circuits can be made on this scale.
In 1989. Todays processors contain many, many more components at a much smaller scale.
The last figure I saw, not too long ago, was we're up to 15 billion transistors on a single die.
That's 15,000,000,000.
A typical 8 bit microprocessor in the 1980s in our home computers contained around 9000 transistors
@@deang5622 Apple’s M1 Ultra has over 100 billion transistors, however that is not impressive - the chip is enormous. What is more profound is the 1nm transistor size made by some MIT researchers a few years ago. That’s 2 orders of magnitude smaller (roughly) than the current transistors used. That’s around 10 hydrogen atoms in width, which is an insane achievement.
@@deang5622 Cerebras launched a new AI supercomputing processor with 2.6 trillion transistors, but that's a different sort of processor.
@@binaryalgorithm total number is not the right metric to be impressed by, it is transistor density. Still extraordinarily impressive, but Cerebras uses a whole 12” wafer for that count, not just a single die (about a few cm^2). That said, today’s technologies are using transistors with several nanometer gate lengths (although that’s not quite half pitch, for a number of technical reasons). What is almost more impressive is that we can fabricate 100T transistors on a wafer and nearly all of them work. The yield is impressive and unimaginable (failures < 1e-8 or fewer).
I had so many 486 chips and motherboards as a kid, but I never knew how unbelievably complex they were at the microscopic level. It gives me a whole new respect for chipset architects and engineers, a group of people I didn't know I could respect anymore than I already had.
Thank you for your kind words.
Yes, true, that's just a fact, people now throwing some old intel chips and crushing them like junk but they don't know how dense and complex they are
The 486 is my favorite CPU. It brought so much to personal computing.
@@volvo09 Got me through college.
they were all hand made.
When people think of exploration, they probably think of going somewhere like outer space. But sometimes the most amazing places to explore are the tiny landscapes and structures that are so often a mere arms length away from us, which we never even notice are there.
I like how you explain all the features because it makes the video a lot more interesting when we are seeing the components up close. Makes it feel more magical
Thank you.
All info, no fluff, no music…perfect.
This is fun to see, as it brings back memories.
On the day of my in-person interview at Intel, I was given a tour of what was then Fab 5, where a technician bored with the usual fab tours showed me the copyright etched into the silicon on an electron microscope, at the time the chip was manufactured new in 1997. I worked in Fab 15 and Fab 20, until I left in 2005. I was a technician in diffusion, which was depositing very thin layers and oxides, or as we liked to say, "making silicon rust".
What did you think about the Pentium 4 processor?
@@kaan4864 I had one and for me it worked really well.
@@310McQueen interseting, i had one and it was very slow and overheated a lot.
Although my comment here appears to be a year late, I wonder whether anyone in the know could tell us whether miniturazion has merely provided greater speed, or actually enabled computations previously very difficult, or whether if it were possible to run a i486 100+ times faster if it could match the computational capacity if a more modern chip.
@@dannypal123 Modern microarchitectures absolutely do enable computations that were difficult on earlier CPUs, particularly through hardware acceleration. AVX, SSE/SIMD, AES, etc.
This is impressive. The fact this tiny structure was manufacturable 30 years ago means to me technology is way more advanced than I believed it was. Now, they're on a nanoscale level (7 to 14, and soon as low as 4), meaning everything you see there is actually between a hundredfold to a thousandfold smaller. So they can litterally put over 100 486 CPUs in the same space than a single 486 CPU. And that's not accounting for the 3rd dimension: add several 10s more layers, and you're in the 500-1000x density range.
Awesome.
Nothing on a 7 nm chip is actually 7 nm. On a 0.350 micron process the gate pitch and the minimum metal pitch are both 0.350 micron. This has never been the case since; with RAM optimizing for minimum metal pitch and CMOS logic optimizing for gate pitch. Since FINFET the no dimension anywhere on a chip is even close to the node name and they're not even very comparable between manufacturers.
@@soylentgreenb Ok. You kind of told me I was wrong, but without correcting the mistake. Would you have that information handy by any chance?
Maybe you just aren't old enough. Most people started getting computers in their homes in the early 80's (although the Altair 8800 came out in the mid 70's.) They were little more than toys even when compared to this 486. However, they were truly amazing when compared to how we had solved problems for thousands of years
@@matthewbanta3240 My age is fine thank you very much. This CPU is still younger than me. I just never had the opportunity to see how small those silicons already were at that time, and that's just putting everything technological into perspective to me.
I used to be a machine designer, so I am well aware of precision manufacturing techniques. But this is still order of magnitudes smaller and more precise than what was achieveable in my domain 20 years ago.
And that's the fun about learning new stuff: despite looking like an idiot to the people you're expressing your awe to, you end up growing as a person nonetheless.
@@matthewbanta3240 what an unbelievably useless comment. “Back in MY day we actually understood how amazing technology was”. GTFOH
I always find content like this fascinating. Many things have been reduced to a Nano scale both manmade and in nature. It would be great to see more content like this.
Yes, and imagine, that isn't even "nano" scale.
A trip down memory lane. My first PC was a 486SX. Many good memories of playing X-Wing Flight Simulator on that. Good Times!
How about "Myst"? By Cyan Studio
One of the best videos ever seen of the microscopic world, i like it, i Love it, keep up bro
I am beside myself with the quality of this! I am hopping up and down in my seat with excitement.
I'll never forget when I was working on bare ICs at Hughes Aircraft (we were making multichip modules), and I was bored and looking at an IC with a high power microscope. Suddenly I saw a "happy face", circle with eyes and mouth, upside down. I turned the whole thing around, and sure enough, there was a very distinct happy face on the die. Of course I told my coworkers about it, and they just smiled. IC makers put lots of easter eggs on their chips just for the fun of it.
This is fantastic content! The EDX overlays are really cool.
I do have a few tips that I’ve picked during my time as a metrology engineer… Allied High Tech’s Epoxy Bond110 works really well for mounting protective glass slides to your cross-sections. (Their diamond polishing films also work really well). Proper polishing takes time you can’t rush it. Pt or Au sputter coating also goes long way.
Fun note about AFMs… if the object you are trying to image/measure is smaller than the radius of your AFM tip, you are actually measuring the tip itself. We would use this phenomenon to determine when a tip’s useful life has ended.
Rad, thanks for the tips! Will look into that, I mistakenly assumed a random bottle of cheapo 5-minute epoxy sitting around would be sufficient... and then was horrified how many voids there were and how poorly it held up to polishing. :) Did you mount to the glass slide and then polish both the slide and the sample simultaneously (e.g. the slide was part of the cross-section)? Or was it just used as a mounting substrate but the cross-section was just sample + epoxy?
@@BreakingTaps I would polish simultaneously. I’ve found when using this epoxy, less is more. A small drop of epoxy then clamp the slide to the sample for curing.
@@BreakingTaps yeah you might void your warranty doing that 😂
As much as I think I know, seeing this makes me feel incredibly inadequate and humble. What an incredible video.
just love this invention, seeing semiconductors and transistors, is like watching a beautiful mountain or the sunset, its a masterpiece.
This is incredible! Very cool seeing the "3D" images of the 486. One of my favorite TH-cam videos in a while.
The most mind blowing part is how we develop the means to do it, the tools to do it, then to scan it then to see it then... that's really incredible.
Small correction: A finished chip is not "a wafer". It's a chip. A wafer is the term used for the circular silicon substrate (typically 300mm in diameter today) that chips are made from. They're "photographed" onto the wafer in repeating patters, as many as there is room for, one layer at the time. Once all the layers are complete the wafer is sent for testing, cutting and packaging.
Still gonna call it a Wafer cos it sounds cool and reminds me of food
I remember my 486sx from the 90s it was my first computer. I ran a B.B.S. called Gateway. It had only 4 telephone lines running at 14.4bps. This is totally nostalgic 😊
Gorgeous, I love some of the spots that look like a very large industry from a sky point of view.
This guy not only shoots the electron microscopic view but also uploads an edited video of it in 4K!
What a legend!
The Fly still has a better processor and comes with an OS. The Fly can duplicate itself. Great Video I subscribed!
It's an unfair comparison. The fly's processor had 3 billion years to evolve to where it is right now. You need to compare it to a human made processor made 3 billion years from now.
@@BoyFromMa I don't think we are late to the game. We are working on it, but why? is this something we used to do? we might be lost trying to get back to the Science/Creation Barrier.
The ability of the microprocessor to be OS-agnostic, able to load any arbitrary code, is actually a win compared to having it built-in imo.
@@delphicdescant Oh but the Fly incorporates multiple processors, Bios and CPU with flash. Don't for get the Ultra low voltage rechargeable power supply. A damaged wing? it will rewrite data in the flight computer to compensate. On The Fly!
The fly cannot perform data analysis from a wide variety of sources, much less play DVDs or stream video from the 'Net.
This is absolutely the most beautiful video i have watched in TH-cam, which up until now was Life beyond by melody sheep for me. This is also the first video of yours i have watched and it made me sub. Amazing work!
It's like viewing a Google maps with lots of large building structures, roads intersections and parking lots... Amazing...
I started work in a Failure Analysis Lab 40 years ago and worked with microscopes and basic tools. A few years ago I returned to the work and now have access to a SEM. You're right about the polishing process. It's a skill and an art that requires a lot of experience to become good at it. I don't have that much either but there isn't much demand to see chip defects in the OEM electronic manufacturing arena these days. My predecessor did this type of work 20 years ago and used plasma etching equipment and chemical etches to remove the IC layers. Removing the layers requires knowledge of the device's construction as one of the other commenters has pointed out. Otherwise you're experimenting and hoping for hte best.
Fascinating! Thank you for showing this. I got my start on the Internet with a 486 based computer. Things learned from the Internet have been equally important as my formal education. This started with my first computer with its 486.
when looking only at the transistor count (i know, wiring and stuff couldnt be scaled down 1:1), you could actually put close to 20.000 486 dies in place of one AMD zen 3 die. its 20 billion vs one million transistors. or the other way around: you would not be able to see the 486 die with your naked eyes when done in modern lithography, because it would only be some µm² in size. even the optical microscope would have trouble resolving anything useful.
and yet, even the original 486 feels like sci fi magic when thinking about it.
its totaly astonishing.
Extremely cool! I work for the company that manufactures your SEM. Primarily I work with the image data after they have been acquired with an SEM, FIB-SEM, TEM, etc. Nowadays a FIB or Broad Ion Beam (BIB) are often used for milling and polishing the surface before imaging. You can even repair a bad circuit with what we call a Circuit Edit system, which is basically a special purpose FIB-SEM.
What you are doing on your channel is one of my dreams and inspired me to purchase an old Hitachi SEM at auction. It will take some effort to get it working again, but your hobbyist results keep me interested in the process. I wish I could own one of our own microscopes, but it is not a charity, lol. They are pretty expensive.
Good luck with your new channel - subscribed!
This is and will be a permanent saved video for me, truly mind-blowing and informative. Thanks for creating this
How on earth have I never found this channel?
Amazing writing, pleasing voice, *brimming* with information, which is eloquently and intelligently conveyed, and the video length is *just* right to give information and kindle a yearning for information in me, which made me research this further.
Keep up the good work. Thank you.
The 486 chip looks like an aerial view of a cityscape under the scanning electron microscope.
This is so incredible. I’m amazed by how small and complex it is and how many transistors it has. And now I think my mind would be blown away if I see the new ones as even each year so many improvements are being done and now the chips have reached 3nm process. There are really genius people out there and I still can’t believe how they can manufacture these without an error so we around millions of all these individual transistors. Amazing is not enough as a word.
Nope every wafer not always perfect. Sometimes some core cannot function and the cut by laser. And manufactur selling less core chip. Or sometimes even the chip function properly they consume more or less power and they called sillicon loterry
Didn't even wait for the end of the video to subscribe. Card came up and I clicked it. Let the Micrographing consume me.
Thank you for doing this by the way. idk why other creators haven't done this yet, but I am sre glad you did!
Amazing! When I was a kid, everything was electro-mechanical and vacuum tubes. Pencils and AM radio.
Thank you so very much for allowing to see inside one of the best micro-processor chips ever made.
I'm an electronics major and really appreciate the splendid beauty of these wonderful devices.
Yeah, it's now old school stuff but it's still a work of art.
Like the Mona Lisa !!! 💎💎💎
The furnace in the background gives me hope!
In a past life, I was a tech which my job duties were to maintain 10 beautiful Zeiss microscopes plus a small desktop SEM within a geology research facility. Although I never polished a chip down, ( I could have as we had a Thin Section Lab also ) I did view a photo sensor array, at the limits of visible light which required immersion oil to do so. It was quite the eye opener! Just the pixels of an array needed 1000x magnification to view them. Incredible! Thanks for the journey.
I've been looking for a deep dive video like this for years, thanks for the insight!
iconic. I have one. its was the biggest thing in my life in the 90s when we got our first 486sx CPU!
Yes! The 486 was awesome, my favorite CPU. It brought the modern computer as we know it into existence. I will always have a few in my collection.
Wow. I remember building and programming 8086 computers back in the day...16k of ram and 5.25" floppy disk drives. Amazing how technology has jumped over the years. My best project was a 486DX40 with 512kram and an 80meg disk drive, a 5.25" and 3.5" floppy drive. Ah, the good old days. Thanks for this incredible video.
@Jason Fitch 16k was not a lot on an 8086. The 8086 could address 1MB of RAM. 16k was where the first PC started at (though it used an 8088 - the 8 bit I/O version of the otherwise identical 8086).
512 K of RAM on a 486? Man, even my shit-ass Deskpro 386 SX had 1024k. On a DX-40, I'd assume at LEAST 4 Mb RAM. Maybe you meant 5120k RAM? I've seen (even recently) weird setups like that - I have a 286 in my basement with something strange like 4768k of RAM.
@@the_kombinator If you re-read my post, I said my best project, not what the system was capable of. You really do not need to be so critical, I did not have unlimited financial resources to build a maxxed out system.
@@tonyv8925 No worries mate, I doubt I'll at any point in time be putting 4 GB of RAM into my 386 ;)
I dont know if you are a Factorio gamer but... you may be amazed by that Megabase 2:55 (really looks like that)
This is the best video on the topic I've ever seen, thank you so much
WOW how breathtakingly beautiful and complex these steam chugging archaic technological dinosaurs were!
So cool! I love how the traces at 6:48 look like an IR satellite image of an abandoned city.
I could watch this all day. Integrated circuits are beautiful
Wow what a fantastic video! Glad I discovered your channel.
Could you also do a memory scanning (DDR1/2/3/4 whatever you'd like).
Will add it to the todo list! Have some old sticks collecting dust in my drawer, shouldn't be a problem!
@@BreakingTaps Thanks! 😁
I owe these people a round of an applause for making the devices and transistors that we are using right now.
Beautiful shots... chips from close r truly both amazing technological achievement as well as artistic, thx for this vid! Jewelry made from chips look also incredible😉
As a hobby photographer and someone who loves tech (especially pcs) this is crazy, I have never seen stuff like this so close up, and every render from 3:18 just looks like a perfect image. It's chaotic. but also somehow organic, it is really fascinating to me. Thx for the experience
Science is a fractal beauty of infinite complexity. I wish I had been exposed to this when I was in grade-school... 31yo and three years into my engineering degree, I guarantee I will never look back. Only forward to a passionate life of learning, with dreams ever dominated by fantasies of earning that elusive "PhD" next to my name. The world is a fascinating place.
Seeing the elemental colormap of the cross-section makes me realise how well structured and carefully designed the whole process is.
Seeing silicon and oxygen was intriguing (but not that surprising), but seeing the aluminium microwires having a coating of titanium underneath was eye-opening. And seeing that the Electron Microscope is able to discern between atomic elements is something that blew my mind! 🤯
The amount of planning and work involved made me really appreciate the sheer complexity of microchips!
I suggest adding it to the featured channels list on this channel
Good idea! Just added it :)
Where is the AND gate in this cpu that I learnt at my university?
Thats an incredible footage! I have disassembled my old CPU as well and was really surprised to see that masterpiece! Its is beautiful!
Absolutely mind blowing. The video feels painfully short. I could easily spend hours looking at this! Kudos for the outstanding job!
oh yes. If there was a lofi or ambient/ space music soundtrack for footage like this, that would be some awesome background study ambience, and some badass yoga background ambience =]
AMD's Zen cores are very interesting because instead of straight or right angles, the chip features blob-like structures as a product of AMD's High Density Library. I hope you guys can take a close-up look of Zen soon.
Wow that’s a well planned way of introducing a new channel. Well done 👍🏻 I’ll be def a subscriber
Credit where due, I stole this idea from Wendover Productions :) They recently launched a second channel (Extremeties, I think it's called?) and did a similar setup: had a few videos in the new channels backlog, and aired an episode on the "main" channel to help advertise it. It seemed like a nice way to launch a channel so I poached the idea :) Thanks for watching!
Minute Correction: The 80486 (DX) was introduced in 1989 not 1985 (1985 saw the advent of the 80386). The 80486 (SX) chip in the video actually was never available until 1991 when Intel offered a cost reduced version with a defective, disabled or (later) omitted FPU. The copyright is still that of the original 80486 (DX) from two years prior and the inscriptions carried over.
Thank you! I've been hypnotized the technology of the Ic since I was thirteen...
I really enjoyed how you described everything here! It felt like a calming story haha. Thanks!
Very interesting stuff. The technical details of these things has always been shrouded in mystery and obtuse academia - sparse charts and illustrations, but not actual examples. This is so close to being an actual breakdown of the process - I only wish to know more about how these things were actually manufactured. It's mind blowing to think that vias could be done on a silicon wafer. How much is even silicon anymore with all those other metals added? How even were they added, aligned, printed? Incredible stuff, and a great video.
I'd love to get a super high resolution 3D model of the chip to drop into Blender and make some really cool renders
and post it and share with the world :shy:
3D? Blender? I guess it's CAD based and designed, even if it's 3D. Blender or anything like that would likely crash instantly trying to show this amount of elements, vertices, polygons, shading etc. Not talking render here.
That was only 1989, imagine today. Incredible!
Words and feelings can’t describe how I’m blown away by this. Wow… this is just.. WOW AND ITS 30 ish years old just.. I’m speechless
Great video. I would like to suggest that it might be easier to understand what you're referring to if you have a pointer or in some way highlight the feature you're referring to. Often there are many different features visible in an image and it takes me a long time to determine which one is correct.
In grad school I examined thousands of SAR images (synthetic aperture radar) and could instantly recognize features that took people who weren't immersed in satellite remote sensing a bit longer to understand.
That's a great point, thanks! Will try to incorporate more pointers/arrows/indicators in future videos. Appreciate the feedback, trying a new format is are always a little scary! ❤
This has natural flow without pointers.
I second this
This is honestly mind blowing. And the fact that we all have one of these but way more advanced in our pockets.
That's an understatement! This full size desktop processor has 3.3 million transistors and the newest iPhone has almost 16 billion and is smaller than a pinky nail, while using vastly less power.
A top end cell phone literally has more power than the fastest supercomputers of just a couple decades ago.
The fastest supercomputer in the world in 1999 had about 15 percent more processing power than an iPhone 14. Insanity.
This is mind boggling. I have seen similar microscopic views before, but this brings the circuitry into much better detail. I just can’t imagine how these lines and connections are laid down (or etched, I guess). In the future, I hope you’ll do a video on this process.
The complexity of manufacturing this must be mindblowing
This video is too fascinating, it's like when I was analyzing a ram memory module and in the program I could separate the layers and see how everything was interconnected through the tracks, it's like a small futuristic world of the kind we see in the movies, really fascinating, Greetings from my beloved Venezuela.
There were some people that reverse engineered the 6502 and z80 with just some microscopes. these are MUCH MUCH simpler 8 bit single layer CPUs however. still impressive. one of the projects was called "visual 6502" recommend checking it out some time
why didn't you explain how the chemical composition analysis thingie works? I have no idea, but it looks amazing
The techniquce is called Energy-dispersive X-ray Spectroscopy, short EDX. Almost all SEMs are equipped with it.
Yep, what Piipolinoo said. EDX/EDS. The electron beam, particularly when cranked up to max intensity and accelerating voltage, will kick out characteristic x-rays from the sample. And the energy of the xrays emitted is specific to each atomic element. So by blasting the sample with the beam, you can collect xrays from each location and categorize semi-quantitatively what elements are present.
There's a lot of nuance, a lot which I don't even know yet, which makes analysis tricky. Detector location, how flat the sample is, how deep the beam is penetrating, overlapping spectra, etc etc. But for some things like the aluminum interconnects it's easy to determine what's going on.
Didn't want to talk about it too much in the episode, trying to keep that channel more "mainstream"/documentary-like. I'll probably talk more about the technical aspects of the SEM itself on this channel in the future though!
@@BreakingTaps but I thought you were doing these with your scanning probe microscope that pokes things with a small needle, I didnt realize it was a SEM. You have also shown the wiggling needle so I was pretty confident you somehow poked their chemical composition with a needle lol.
Can we all just take a moment to appreciate Intel's rate of production at this scale?
The first pill costs two billion dollars, the following ones, 4 cents.
@@the_kombinator - beyond that, is repeatability and testing... which involve mind boggling integers at this scale.... both in terms of high binary operation count, and the miniscule physical size.
@@driverjamescopeland That and also, if there's a mistake in a subsequent pill, you lose 4 cents. If your wafer isn't clean or there's a manufacturing defect on a die, well, you just threw out a lot more than 4 cents. Having said that, I'd love to get a failed 3 or 486 substrate.
@@the_kombinator - I didn't know until recently, most of those failed substrates can't be recycled into new modern chips. They can be used for larger resolution transistors, but not microprocs. It's simply not economically feasible to use anything but virgin silicon for new microprocessors.
So once in a very blue moon TH-cam algo spits out something actually fascinating. Today was the day.
My terminology has upgraded from "thinking rocks" to "thinking rock lasagna"
Very cool and zen look at the microworld. A note on the afm scans, it seems your tip was likely not tracking the traces as it fell off the tops of the metal edges. You can often spot this when the trace and retrace scans are not very similar especially when coming off sharp edges. This can be improved by either slowing the scan speed, decreasing the tapping set-point voltage, or increasing the gains on the PID loop.
Yeah, I noticed that as well. In this case, I think the sample itself was slightly tilted with respect to the tip (and/or the cantilever's arc). The forward and reverse scans actually look pretty similar and not much on the error map. So my theory was that the sample was tilted slightly, allowing the tip to access one side of the vertical wall but not the other wall (since it would be "shadowed" by the overhang). But I could be wrong! Still learning to spot artifacts and their causes :) Cheers for the advice! Will keep that in mind for future scans.
@@BreakingTaps That's true, it could be due to the angle of the cantilever or the tip geometry. Many tips are also asymmetric along the cantilever axis allowing the tracking of steeper walls on one side.
Lovely - a minor correction, the 486 came out in 1989 (as shown on the copyright on your images!). The 386 came out in 1985 but would have been a bigger processor manufacturing node I believe.
Oof! How did I miss that?! Thanks, will start up an addendum for corrections. I assumed the copyright was about the "SX" version, not sure why I assumed it was wrong hah
@@BreakingTaps Lovely video, thanks! Actually, the 486 SX was introduced in 1991, as a modified version of the 486 DX with the FPU disabled. The 1989 date seen on the wafer is probably the manufacturing date.
@@ofloveandliquor Introduced in 1991 but manufactured in 1989? ;-)
Nope, Intel doesn't sit years on their wafers. It's simply the copyright date for their 80486(DX).
@@oldguy9051 Makes sense, thanks for the correction!
Yep I remember looking at the back pages of Computer Shopper magazine (almost as thick as a phone book!) back in 1989 looking for the best deal on a 486 motherboard/CPU combo - the prices were such that today (even adjusted for inflation) you could buy a nice complete gaming rig for the price of a 486 motherboard+CPU back then! I also remember buying a 387 FPU coprocessor for my Zeos 386 around the same time! The 486 was the first Intel to integrate the two on chip with over a million transistors!
Thanks for showing this! I have always admired the i486. :) It blows my mind how advanced microprocessors had already become at the end of the 1980s! The i486 DX4/100MHz is an absolute BEAST of a CPU under Windows 95 (with 32MB 60ns RAM and 256k 15ns cache). Not even joking. The i486 DX4/100 is so powerful that it will do *true* polygonal 3D graphics (it'll play Descent and Descent 2) all in software at 35 fps _without_ hardware acceleration with only a 512kb 16-bit ISA VGA video card under MS-DOS. 8 slot (1x8 bit ISA, 4x16-bit ISA, 3x 32-bit VESA) 25/33/40/50MHz (adjustable by jumper) Socket 3 motherboards are the best motherboards ever made!
As for your video narration, it would be much more listenable if you moved further away from your microphone and spoke louder. Narration is an art form I've studied for decades, and the best style that suites a general audience is a smooth, yet vocal and gritty, authoritative tone (Mythbusters narrator is an absolute prime example). Speaking softly and calmly close to a microphone and letting your voice granulate isn't how professional narration is done. Try taking my advice because I love your videos. :)
younger people today have no idea how much work a 486 is capable of. I can't think of anything we do at our office, other than the demands of today's internet, that couldn't be accomplished with a 486 just about as well.
I'm an assembly nerd and I enjoy so much this type of content !
This is so goddamn impressive. The machinery and skill that is needed to create such detailed and tiny things Is far beyond everything I could imagine!
Around a few 1000 times more transistors would fit in the same die size today :)
I doubt the average person could build this if you gave them 1000 years starting from scratch.
Amazing to see this up close. It still boggles the mind to know that this was thought up by a human. I had a 8088, 286,386 (with math co ) and skipped the 486 and went straight into the first Pentium ........Cool times. Editing the Autoexe bat file, config sys and making sure my 640 KB or RAM on my 386 (w/math co) spread out correctly. I bumped it up to 720 KB after installing 19 chips on a ram board.
I bought an expansion board for my Tandy1000 (8088) that gave me 640 KB. That was so much RAM that I used some as a RAM disk that I copied my compiler onto so I could edit and compile programs without switching floppies.
It almost looks like a tiny city in there!
It's hard to believe that this chip we are looking at was made 32 years ago!
Imagine what they look like now?
Those detailed scans truely facinating, wouldnt mind this as my wallpaper.
What a nice video. Thank you for creating and posting it.
Does the 486sx not have easter eggs? I've taken micrographs of a i486DX and it has a block full of initials, other initials hidden in the traces, and the big 486 etched in one area. That 486 had the most graffiti on it, the 386 and pentium 90 I imaged didn't have any. Another cool chip to image is the Weitek 3172A, that had entire names etched on empty spaces. Last chips to have removable covers were from the late 90s, after that they started making them solid for better heat transfer.
Oh, it did! I found the big block of initials ("JL") and the 486 logo. Didn't see the hidden initials, but will go peruse my stitched scan to see if I can find it. I actually captured some _really nice_ images of the initial block... but lost the data :( The next day I was imaging some other parts of the chip and the software was happily reusing names and overwriting my prior images, so I lost a bunch of stuff. Unfortunately only realized after I had embedded in epoxy and started grinding away at the chip.
@@BreakingTaps If the 496 logo is at the top and readable, the larger somewhat hidden set of 3 initials are in the large trace to the right and a little down from the block of many initials. There's also something written in the lower right corner.