Physics of Computer Chips - Computerphile
ฝัง
- เผยแพร่เมื่อ 24 ส.ค. 2024
- You can't beat physics. Why the chip manufacturers are heading for a wall. We asked nano-scientist Phil Moriarty Professor of Physics in the School of Physics and Astronomy, University of Nottingham
EXTRA BITS: • EXTRA BITS: More Silic...
Teaching Physics in Ethiopia: • Teaching Physics in Et...
Is it the end for Moore's Law: • Is it the End for Moor...
Password Cracking: • Password Cracking - Co...
Gamer's Paradise: • Gamer's Paradise - Com...
/ computerphile
/ computer_phile
This video was filmed and edited by Sean Riley.
Computer Science at the University of Nottingham: bit.ly/nottscom...
Computerphile is a sister project to Brady Haran's Numberphile. More at www.bradyharan.com
"You don't understand it, you get used to it" sooo relatable
Yep
It's not sad, it's just the way it is. No one can fully understand the implications of quantum mechanics (namely entanglement), which is why there are so many interpretations of the same observation. Look up the double-slit experiment and tell us that doesn't blow your mind.
females xD
Isn't it like that for everything we "know"? Classical physics is just as strange, we are just used to it.
Not really, quantum physics is very un-intuitive. The fact that an electron doesn't exist in any particular place or time (and is just a probability wave of being somewhere) until it is observed is something spooky that no one really understands. There's nothing like that in classical physics. Or the fact that two entangled protons can be a universe (or any length) apart and still affect each-other (instantaneously, breaking the laws of relativity).
Phil, aren't you too large to call yourself a nanoscientist?
Dad, I told you not to make these jokes in TH-cam comments!
its okay, he identifies as a nanoscientist!
cissized pig
He was referring to little phil... downstairs.
If he's a nanoscientist, I hope I never meet even a microscientist, let alone a full scientist!
"You don't understand it, you just get used to it"
Probably one of the best physics quotes ever
Yeah that is actually very profound , it keeps you going.
It’s sounds like the food that’s served in US prisons and jails!
John von Neumann was the first one to say it. Give him the credits
How do the individual atoms feel about being manipulated?
#METoo
electronized
Uncertain. Of course.
The electrons are TRIGGERED
Try to learn about transistor...
My favorite connection from physics to electronics, is the fact that quantum tunnelling effects are at the heart of how flash memory and EPROMS work.
I don't always watch computerphile (over my head) but you could put Phil Moriarty in a video about paint drying and I would watch it........oh wait.
It seemed like the video ended while he was still explaining something.
Prof. Moriarty actually never stops talking. The best you can do is turn off the camera just as he changes topics.
He was about to unify classical and quantum physics
He was about to show his pee pee
I want more of this guy. He could start his own channel and just talk and I think he'd have thousands of subscribers;)
He has his own channel. Called Phillip Moriarty.
He does have his own channel! Moriarty2112, or you could follow Sixty Symbols where he has many videos about physics!
You should watch Sixty Symbols, he features in literally several dozen videos there.
+Phi6er Aww really? I liked this guy. :/
seconded
Being a chemist, just having listened to a physicist, talking about mechanics, for the purpose of computing, I just realized that the electron couldn't care less about how it's manipulated and by whom.
Why don't they just download more RAM into the electron beam to make it go faster?
I'll create a GUI interface using Visual Basic to see if I can track down an IP address for the download.
CSI?
Usual Hollywood hacker nonsense, that particular excerpt is from CSI:NY.
Justin Bell I thought so
"we need to hack faster !!!"... 3 people typing at once... on the same keyboard
This was a really informative video. Some illustrations/animations to visualize what he said would have made the video even better, although I understand that they take quite some time to make.
Check out en.wikipedia.org/wiki/Photolithography
i agree. His experience as a professor really shines through.
I'm not sure if there's a definite correlation, but all the physics professors I had in college were the best teachers I ever had. Along with their ability to explain things they were passionate/excited about their field
More of this guy please.
This is one of my all time favorite videos on TH-cam! I have watched this video about 7 times now and I just absolutely love how well Mr. Moriarty explains the semi-transistor manufacturing.
hi there camera man!
Hello!
Was the camera man sitting on a basketball?
not enough giggles for that to be the case.
And this why movies tend not to use real mirrors!
movies use real mirrors, they just don't have the camera face on with the mirror
Prof Moriarty is one of my favorite guests on any of the “phile” videos. Awesome guy and very good at breaking things down to a level I can understand.
2019 update: 5nm in the works, 7nm in production (AMD Ryzen 3000 series, e.g.)
So if 14nm = 50 atoms, 7nm = 25 atoms, 5nm = 17 atoms. Getting there
3nm in lab
yet they are too inneficient
Thats not the case there. 5nm isn’t exactly 5nm. It does not represent a geometrical shape on transistor. It’s just the technology’s name. You can call it “marketing”. The real limitation for geometrical shape of transistor is 7nm. Nothing more than that.
For silicon ^^
Watching this in 2019, they are now manufacturing 7nm microprocessors, how things move on.
intel 14nm have 8nm wide fins in a finfet transistor, I think they are making features bit smaller than 7nm in the absolute sense.
And on what he call Extreme UV (EUV)
@SuperTanner But how can they go smaller than atoms
And tomorrow morning a 5nm machine’s getting delivered to my home.
Let’s come back in a year, see where we’re at
we are in 5nm stage now
I love how excited he gets to answer each question and you can tell it’s genuine too
"If you find that confusing.. Good" - soo funny and soo true
That was a brilliantly clear and energetic overview of modern chipmaking. Professor Moriarty explained how a Silicon transistor works, but didn't label it as such. The Silicon substrate is formed into the transistors by adding impurities (doping).
Three years ago they were talking about 14 nm, today we're talking about 5 nm; so this video becomes history in less time than it would take to study Electronics.
Unfortunately the 5nm and 7nm are really brand names than actual sizes of features - they are usually much more refined 10nm process which are capable of higher densities by eliminating issues with previous 10nm and 14nm processes.
Essentially we are right up at the limits of manipulatable sizes when it comes to computing - most every improvement now comes from extremely complex and well designed architecture - cache for instance which is what amd rides on for their superior processing of late with 5000 and coming 6000 series chips
@@aravindpallippara1577 You need to search before replying, and address what was written.
Plants are being constructed in Arizona, Tainan, etc. with 2 nm coming in 2025 - call it creative naming or fudging on the numbers - each new plant builds a smaller process.
People don't invest and they don't spend over a hundred billion to convince you, you are not convinced.
Instead the money is spent to place billions of transistors in the space previously occupied by one transistor, decades ago.
They really are getting smaller.
Phil Moriarty is badass
You said the wavelength is a limit but the nobel prize last year was for the invention of microscopes which overcome this wavelength barrier. I think they used the light emitted by proteins and blocked the light emitted by neighbouring proteins so that the resolution was down to one protein. There might be a way to use that for making smaller chips.
Moriarty is one of my favorite science communicators
"Quantum Mindset" #bandname
+MaxPower ^ rofl
But damn if you try to find the location and time of any given concert in particular.
+scabbynacker "Are you thinking with quanta yet?" #tagline
I don't know why but he reminds me of Roy from The IT Crowd
Just of course more intelligent than, "Hello IT, have you tried turning it off and on again?"
Similar accent
He's also got the Roy-esque quality of talking about intricate stuff in a non-jargon way
Was about to comment this, brilliant.
...and gestures.
I was wondering who he reminded me of!
Videos like this immunize me from the Kruger-Dunning Effect.
same here my friend. i hqd to look up the effect so im even more ignorant ;-)
putting you in your place is a simpler way of saying it
If you think it did that, you should be worried.
sugarfrosted Okay... it was just a booster.
Or even the Dunning-Kruger Effect.
I'm just glad there's someone that can see that physics chemistry and computer science are integral to each other than the normal 'brinksmanship' that you see in these fields.
Hearing about 14 nanometers being the smallest you can go, while my cpu is made with 7 nanometers manufacturing just shows how fast technology moves on. This is 4 to 5 years ago, and at that time, their ultimate goal was 13.5 or something like that! Science moves fast!
Professor Moriarty.
Who's first thought isn't Sherlock Holmes.. :)
Sorry just had to mention it.
I can never get enough of Professor Moriarty. Such a fantastic and interesting person!
You should have seen him in Sherlock Holmes
Very good speaker, very well explained, and engaging. This really helped me understand exactly what the subject was about. I'd love to have this guy as a teacher.
I brought up the fact that Chemistry and Physics and ultimately everything that follows is a seamless whole just divided into digestible parts to a chemistry instructor once, and he almost flipped his lid. It was almost the same reaction from the physics department, yet they worked together constantly on things, though the chemists tend to be more reserved and the physicists tended to let their reservations go a bit, especially on things that went "boom". Nice to hear the Professor say nearly the same thing, about the relationship.
God, I love how professor Moriarty explains stuff !
this was incredible to watch. the passion and conviction he showed was amazing
5:40 "as a physicist, it's not that you understand it. you just get used to it" wise words. I tried understanding the wave nature of electrons, lost half of my hairs just to get my head around that thing and I am not even a physicist. Physics can be addictive. also, it can be intuitive and unforgivingly confusing at the same time. I know a person who'd agree to the last statement. That is, Mr. Erwin Schrodinger.
To elaborate on the "Layers" question, yes, it's very much done in layers. In fact, even the first layer wasn't fully described here. After the exposed (or unexposed, depending on the process) areas of the polymer are washed away, another layer of some material is applied. Then the places where the polymer remained are washed away, leaving the new material only in the gaps. The material can be dopants for the underlying silicon, metal layers to connect components, insulating layers to separate things, etc. It can even be exposed to etchants, rather than a new material, to remove whatever layer is showing.
I really admire Phil Moriarty's ability to talk around naming concepts like quantum tunneling, and keep his talk on track even with tangential questions.
I usually just understand 25% of these talks but just love this channel and will keep coiming back to it again and again. Thank you for this!
I see Phil Moriarty. I watch. I upvote
indeed,his videos are better , because he actually is more specific than others...
face01face f
And that's how you identify a redditor.
I see Phil Moriarty. I upvote. I watch
This guy won't shut up. I love it!!!!
The energy of his explanations.... HE should be duplicated in the kindergarten, school, high school... NOBODY cannot be energized and curious the way he is talk and explain things! His energy in language would make me a damn poet!!
This is a great effing interview.. awesome enthusiasm and passion.
I thought that the wave interference of the photons (as in, interference you see with the double split experiment) would be come a problem. But instead you could create a mask that actually utilizes this phenomenon to create interference patterns that match the target pattern on your silicon sheet.
The physics and technology in making the 13.5nm light that is talked about here is some of the most advanced applied physics in industial RnD going on im the world today. Primarily this work is done close by to you, in The Netherlands. I would love to see you or Brady make a video on it. great job on videos. From another Irish physicist.
clever semiconductor industry: how about we name nanometers stop meaning nanometers?
people: so, like lies?
clever semiconductor industry: no, no, we'll just pretend that it's just a name and call it an advertising term
"I don’t think a regular person appreciates how insane it is that computers work. I propose we stare at each other mind-blown for about 1 hour/day, in small groups in circles around a chip on a pedestal, appreciating that we can coerce physics to process information like that.""
One of my favorite Computerphile videos. Great explanation of concepts I've always wanted to understand. Thank you!
4:56 You are doing it wrong !
You are supposed to say "You can't brake the laws of physics !".
XD
I'd love to hear Prof Moriarty talk about spintronics and photonics if you have him on again.
Just a silly question: doesn't diffraction screw up with the lithographic process considering that light has to go through such tiny apertures?
Yes, which is why you use light of very small wavelengths.
The wavelengths they're using in mass production today are not that small. I believe 193nm is still standard, even though they're making features as little as ~14nm in size (actually many parts of a "14nm" process are not 14nm, but it's all pretty much to scale compared to older process nodes like 22nm).
The next step is (supposed to be) EUV, where they do drop to very short wavelengths and high energies, as discussed in the video. They are having a lot of issues getting that to work for mass production though.
The diffraction of light really is a limitinig factor when reducing the size of features. One of the ways semiconductor manufacturers get around this is phase shift masking, which Prof. Moriarty explained as two masks just slightly offset from each other.
i didnt get how shifting the two templates helps
+sewer renegade There are actually several ways of masking phase shift masks. I think Professor Moriarty is portraying the general concept of setting up having the edges of the photomask phase shift the light passing by so that when the light reaches the photoresist on the wafer, the edges of the patterns are being enhanced by constructive and destructive interference of the light waves, thus making features smaller than the wavelength of light possible.
Physics AND computer science, all with Professor Moriarty. Great combo, I really enjoyed this episode. Thanks guys!
Wow, that was was barely within what I could follow all the way through. The few times I started to get lost he stopped and explained it a bit more. Very well done, and I even picked up a few new things.
silicon is reflective? MIND BLOWN.
This guy is magic, and clearly loves his stuff. Excellent.
Even though it's microscopic, the offset mask technique made my OCD flame up. Which direction do you offset? Who decides that? What if the mask goes out of bounds? My heart asks for eternity
Big man working on small stuff, respect
This interview elevated my understanding of how we're able to manipulate atoms. Thank you.
Great video; thanks Computerphile and Dr. Moriarty :)
The next stage is going to be optical computing. We will be operating in the THz range, because we will no longer have to wait around for transistor states to switch.
I swear every time a computerphile video ends, when I hear those beeps, I start singing "Askepios" by the Mars Volta.
I love it.
At the end of each video I literally start singing "I'll be there waiting..." and start asking myself "damn what song is that?"
My personal guess on what happens when the industry hits the wall, is that they figure out ways to build them larger.
They are already doing this with RAM and once they figure out how to build chips which emitt almost no heat they will do so with processors aswell.
I'd love to hear an update, now that Industry has processors at the 5 nm level.
it may be off topic:
ion computer
00 North up spin CW (volt amp)
01 North up spin CCW (volt -amp)
10 South up spin CW (-volt amp)
11 South up spin CCW (-volt - amp)
-------------------------------
photon computer-
00 photon (0,90) no mirror no lens
01 photon (0,270) no mirror lens
10 photon (180, 90) mirror no lens
11 photon (180,270) mirror lens
Nice to see you again Dr Phil!
My fiancé worked for Intel, he was sent around to various clean rooms and such to work on the computers that were running the electron scanning microscopes that they were using for debugging chips.
The whole process is pretty cool to me, as I just stopped learning the abstractions of the CPU at the logic gates, and VHDL design.
A curious wonder, what kind of feature size would a hobbyist be able to achieve? I mean, there's that guy who built a macro-computer by using full chips for his transistors, and I know most of us are better off using FPGAs anyways. But say someone wanted to get into etching their own silicon, what do you think would be the range of quality that they could get to?
My "like" happened at 7:05. The "quantum computing" question wasn't very well informed, though it may have highlighted a common misconception. The answer, spurred from the question, about how "classical" computing must necessarily exploit the quantum nature of matter if it intends on reaching ~1nm scale features is totally spot on. Looking forward to the next ~5min of video!
This is probably my favorite video on the channel.
I could listen to this guy for hours
these videos are like the best thing in my life sometimes. thanks for continuing to make them. :)
i don't know if i understood at least half of the video, but what i do know now is that things just do their thing in physics
please, more on the physics of computer hardware! there's been so many amazing inventions and discoveries through the years in the semiconductor industry so we can use computers as we know it...
Watched extra bits, still want more.
Double patterning is so arcane; especially when looking at super small designs like SRAMs
This is brilliant! I would be really glad if you made more videos on this topic
It's very interesting to dive into the physics and chemistry of electronic computing, it's not a subject I've explored much as a computer scientist.
Watching this on a phone with a snapdragon 845 based on a 10nm process. Interesting to hear him talk about pushing the limits to get to 13.5 and here we're are just a couple years later at 10nm
Speaking of the 'integrated whole'... Years ago I read Richard Feynman's 'Lectures on Computation' (certainly less popular than his lectures on physics, but Feynman did a bit of computer science as well) and he described a model of computation that theoretically required zero energy. I don't know enough quantum physics to understand the exact mechanism (it had to do with particles moving from an excited state to one of multiple rest states I believe) but I've never heard about it anywhere else. I would love to hear some academicians talk about this idea. Or about amorphous computing, although that is still a fairly specialist field... Oh, but memristors! I'm sure you could easily do a video about them! Please don't get distracted by the whole "neuromorphic processors" stuff, that's all anyone ever talks about. Talk instead about the impact they could have on computer architecture - Everything, CPU, registers, L1 cache, L2 cache, L3 cache, RAM, mass storage, all of it could be done with a single large array of memristors. And portions of the array could dynamically change from providing memory functionality to providing computational capacity faster than a RAM read. That could change everything!
I believe what you are essentialy talking about is quantum computing. It uses photon as information. By emmitting light on surfaces on the atomic scale you 'bounce' the photonic energy inside an atom causing the electrons that are 'orbitting' the nucleus to up energy levels. When atoms become more complicated these energy levels gain sublevels. The energy photons carry which equals the Planck constant * the frequency of the light is the lowest obtainable energy possible(zero energy). This energy can be transported just like electrons(i.e. information) and can be stored in the energy levels(i.e. memory). If you can read/write(manipulate) photons you can make a computer using the lowest resolution our universe has to offer.
this is great info, since I come from a physical science background this shows the application of what I learned, wish they had a class on the physics of computers
Really cool to see semiconductor fab processing explained here!
That polished silicon would make a super nice (and extremely expensive) mirror
This material is way over my head, but this video was fascinating. Thanks for putting it together.
AvE has something he'd like to say to the cameraman
YOU FAK
he says his camera is voice activated as a joke but it seriously is.
focus caNUCK
I love how excited he got for the silicon question.
i love this guy he explains so well.get him on the show more often plz
I love how you can see how passionate this man is about what he does.
Computerphile needs to bring us on a fieldtrip to a chipfab
Welcome back professor. I had wondered about your absence.
I really liked this sort of unplanned interview
Chemistry is a bottom up approach to the macro world, and physics is a top down approach to the micro world. Two different paradigms of the same thing: material science. They both inform the other.
It would be cool if Computerphile did an episode on 3D chips. With physics wall coming soon, the only place will be up.
There are metal layers on top of the transistors. E.g. Intel 45nm has 9 metal layers. You always needed room for these since invention for the microchip. So, chips were never flat.
Yeah, i know that. That's not what i meant. More along the line of what memory chips been going through lately, with 32 or 48 layer stacking, except for CPUs. TSV, etc.
The people that say we live in a boring period in history should watch this video
I'll follow you anywhere Phil
Now we have 7nm, quite astounding.
We missed you!
The industry will go to Silicon-Germanium, maybe graphene at some point, then photonics. If they can't then software engineers better start creating some crazy optimized code.
' its not that you understand quantum mechanics, you just get used to it '
wow
I learned SO much from this. Thank you.
I also love his passion. It excited me to learn this.
Please do an updated one of these about the current physics of the newest chips
6:20 The sound of going from one very difficult topic to an other very difficult topic fast.