Despite the never ending stream of bad decisions by UA-cam, it's nice to see at least one algorithm output is an improvement. Great channel, and a real joy to watch.
@@simon199418 Damn, not exactly, but if you want to have deep well meaned and thoughtfull conversations you might wanna hang up with smokers, it`s 1000% easier to focus deep on subjects.
You can`t find the intelligent people because they`re all smoking weed out of the class, obviously there`s some dumb people but even dumb people smoking weed are smarter than the average joe.
FINALLY. A professor that's figured out how to use a slideshow/presentation without fumbling around for the first 5 minutes trying to figure out why it's not working
It used to be that the glass pattern mask was the same size as the silicon wafer but as time moved on the computer chips got smaller and the wafers got bigger. Now the pattern masks do only one chip and the pattern is stepped across the wafer until all the chips on the wafer are exposed. The precision needed to step across all the chips in a wafer and then repeat it for the 40 or so different patterns masks while keeping all the lines aligned is extraordinarily complex.
A big thank you from a retired chip designer! In a remarkable short time, one becomes far behind the current technology. Thanks for the update. I am still in awe of the results achieved by the process development people...............
@paul snor "On September 10, 2019, Apple announced their A13 Bionic chip used in iPhone 11 built using TSMC's 2nd gen 7nm process." Try to keep up. 5 nm coming in 2020.
@paul snor www.eetimes.com/document.asp?doc_id=1333881# I don't know enough about advanced lithography to argue this any further. Press releases indicate that 7 nm currently in production uses 13.5nm EUV.
@paul snor You can, if you get clever with the masks and leverage diffraction effects. The light sources and optics get really expensive down in the EUV range.
I've just found this channel an hour ago. These lectures are so good and ther are so many of them I literally want to cry!!! oh my goodness, I sure hope this is not just a wet dream!
Not to be pedantic with a pedant, but 10^9 is a billion, not a billionth. Prefix giga. And neither dimensionless number is a unit of length... Sorry. Did I say "not to be pedantic?" I meant to say "to be totally pedantic."
I would love to see a similar lecture on the evolution (and perceived limitations) of network switches, from the single wire of the telegram to current optic fiber technologies. In all cases, thank you for an enjoyable and informative lecture
Working in dry etch was my favorite job, we made the optane memory chips for Intel and crucial, with some of our tools hitting 12nm and the aspect ratio of our tools and process was etching at 40:1, another interesting aspect is that the wavelength of light including uv used to expose the photo lithography mask is to large to exposes the design and uses water to refract the light to expose the mask clearly.
I work in the semi conductor industry myself and I find this video to be the best possible explanation of where the production comes from and where it is going to IMO the key phrase in the entire video is a bit tucked away: "at the same price".
Love how you flip education on it's head. FIRST you explain a complex system in simple language so a non-technical person gets excited (and not intimidated) by the subject......then they have the motivations to put in the effort to learn (and really understand) the math, physics, chemistry, material science, etc. needed to create and improve the system & subject that excites them. Current education system requires someone to master math, physics, chemistry, material science, etc. BEFORE, they really get a chance to play the various system and then hope they find an area that really excites them. As a semi-retired engineer w an MBA, it's still the "show me" (don't just tell me) professors & TAs I remember most! I also love how you tie everything back to the economics of the marketplace for the device or system. It takes money to improve technology, and investors to provide the money......but they need to be paid well for taking the risk. The economic circle of innovation.
Totally. It's really hard to keep the big picture simple and in context before and during a technical lesson like this. I've seen a few of EnergyProf's lectures and he seems remarkably good at it.
I have spent a lot of time researching how chips are made, I find it fascinating how small everything is. I have taken the covers off chips and put them under a microscope to look at the features. This was a great video, this technology has advanced more than I was aware of. New subscriber.
The graph at 06:17 has a glaring error. That spiked axis was accidentally labeled zero! Of course, you cannot pass through zero while asymptotically approaching zero. It is indeed 1 (one), or ten to the zero power. It must be fixed. It triggers my OCD.
*Professor* "Your CPU is hitting the level of mouse brain for computational power...isn't that impressive" *People* "Yes it is...Now lets watch TikTok on that machine!"
The nm scale is not an industry standard. In general the Xnm scale refers to the smallest unit of measurement that can be addressed not the size of a transistor. Marketing has required chip makers to lower their nm quotes to denote a new node. A 14 to 7 nm shift doesn't always work out to double the capacity. AMD is using TSMC's. 7nm node for the CCX but are running all of the I/O on Global Foundries' 12 and 14 nm node. A Frankenstein monster that is a very effective means of producing high core count microprocessor.
What Engineering Mathemagical devices really are, in applied techniques. When human DNA was finally deciphered, the real discovery was how much further understanding there was to go, so I'm not worried about robot take over, more like the work required to usefully apply the simpler devices this far. Must be another word another order of intensity beyond awe, for each advancement for this stuff? Awe to the nth +1...
In the seventys I wish I had fuel injection instead of carburetors thinking fuel milage would be better I had a 350 chevy and could tune it to get 20 miles per gallon, now I have a 6 ltr and fuel injection the best I can do is 14.4 mpg. Very Disappointed
Even the air change constantly. Deepholes must drilled into ground inorder to prevent vibrations from earth. Nearly all chips are designed in India. They dont however make any chip except for military purposes. Same method used for anti radar coated air craft cockpits i.e. sputtering.
we should apart from improving the computations per unit time per unit volume of circuit, try to decrease the ultimate need for large quantities of simple arithmetic computations per unit time, or lots of arithmetics, per user or per mankind. better mathematics and physics should in theory reduce the dependence on lots of arithmetics. better formulae should reduce the need for large quantities of simple arithmetic operations. there is after all a finite limit to miniaturisation of physical devices. human progress is not likely to have any limits. therefore the ultimate future of human progress lies not in lots of arithmetics but in freedom from lots of arithmetics.
Kinda ironic, but the more complex something is, the better it gets with time. But it also makes sense, since more complex things have way more room for improvement.
Absolutely fascinating! The sophistication level of that collection of technological innovations is a mega achievement. When you think about what most electronic circuits involve ie: resistors, capacitors, transistors, inductors, and you think back to just a few decades ago where all of these components were placed onto a circuit board that had visible metal traces connecting them all with holes drilled through contact points for soldering on the components. Now it's at the atomic level using photon energy to install the circuits at microscopic scale.
Thanks lot Sir, the almighty creator have given you a best brain and you also do better for us. May the almighty give u best ability and oportunity to do better something for the world. May the creator keep feet u every time.
ive been using the same cell for last 4 years and same with PCs & laptops. android 9 runs terrible because it is a full blown dragnet surveillance OS. same for windows 10.
EUV - he is actually describing what intel is having trouble with the last 5 years plus ... and moore's law is dead. At 3nm or less it will be very very hard to manufacture (down to single atoms)... every gen 18 months or so, new CPUs are only faster by 10-20% and the improvements are getting slower and slower and taking longer and longer. Without a huge breakthrough by 2030 computers will last 10 years plus.
That's what the extra cores are for. Going faster (smaller) is becoming harder, instead thread parallelism shares out the workload. Core to core todays CPUs aren't much faster than 5 years ago but when running software that scales over multiple cores they can be several times faster.
Amazing the drive that capitalism produces. If only there were a way for each individual person to have to buy their own personal energy source for everything they do. Just imagine the results!
Right. Some people would buy from wind and solar, paying fifty cents a kilowatt hour. Some people would buy from nuclear plants, and pay eight cents a kilowatt hour.
At 1:05, the inventor was Jack Kilby (not Kirby), who subsequently won the Nobel Prize in Physics for the discovery. IIRC, he held a Master's Degree, not a PhD.
This was filmed at a somewhat perfect time because Moore's law pretty much petered out around 2015 2016 it's now 2021 and Intel's 10nm is only now a viable node at scale. TSMC is arguably continuing Moore's law as they're now upto 5nm which is utilised by Apple but the costs of wafers on new nodes is growing massively. I think 5nm at TSMC is now above 10K Us a wafer whereas 16nm was less than half of that I think when new.
I used to write control software for Plasma Deposition and Sputtering machines. It was great fun making all the devices work together to process wafers. Turbo vacuum pumps that ran at 33,000 RPM, Cryo Pumps that cooled down to a few Kelvin in order to scrape a few more gas molecules from the chambers. All the way to rotating assemblies weighing several tons spinning at 60 RPM with enough energy to throw the machine through the wall if they suddenly stopped. Oh, and we also had plenty of flashing LEDs ;)
I've always thought the biggest leap in computing would be going from binary computation to omninary (my crappy made up word). In other words moving away from on off bits to multi-value bits. Like a bit not just equaling on or off but say 256 levels of on or off. The entire software industry would have to be reworked of course. But you could move a variable of 256 (or whatever number) in the same time as on or off. It would be huge.
We can make chips with 10 nm features, but we still can't spell "transistors" reliably, and we still write with markers (although on a transparent surface, and "backward"-got to admit, that's pretty cool).
4': Good point: Moore's Law doesn't happen because of any hard law of physics. Physics simply allows it, but what really makes it happen is economics: Businesses and other organizations make plans to cause those increases to happen.
This is a great lecture. There is a minor error though: The exposure performed by scanner-steppers is performed one dye (circuit) at a time (=stepping). Each dye is exposed by scanning a slit shape image across the dye (=scanning). en.wikipedia.org/wiki/Stepper#Scanners. I cannot grasp how it is still possible to expose 100 wafers per hour with a precision in the nm range.
I believe the economics of computer chips is very simple, the smarter the chip, the smarter it _can_ make you and it's always economical to be smarter.
Chips are small, but what if we could make these same fast chips but in a larger NM. I could live with a physical size of s chip like a CPU if the chip is faster and cheaper. I’m a novice but does use of larger NM chip slow it down?
It comes down to economics. You can fit more smaller chips on a 300 mm wafer than larger chips. It would cost more money to scale up than to scale down.
It's about the speed of electricity. If the chip is larger, the electricity has to run more km's to do the same amount of work, which means it takes a longer time.
Intel actually reintroduced a 22nm node because their fabs have not been able to meet demand for 14 nm while they have been stuck trying to transfer to a 10 nm node. The main reason that smaller is better is because the biggest margins are from the data center where the cost of operation in electricity is as important as compute performance. Smaller nodes can traditionally run at higher efficiency.
@@puffthemagiclepton7534 is right, its economics a 300mm wafer where i used to work was about $20k from all the different processes, materials and energy. It would take over an hour and a half to run a single process in my area and could have over 800 processes to complete so realestate on the wafer is super important, you want to fit as many chips on it to get the biggest yield possible
Lol I recently built a new computer after having mine since the i7 920. And I was SHOCKED to see how little wattage PSUs were nowadays. My ATI 5870 + 920 required a 1500 watt PSU. Now I have 1000x the processing power on 650 watts.
19:49 I am doing a FAB build now 2019. This is a clean room with people. This area is considered to dirty for the chip and will never see this room. Instead the chips live in a box that can be even more cleaner. The box itself is only open at inside the tool.
Until AMHS takes your FOUP across the fab with the door still stuck to the ATM lol. We had N2EFEM ATMs that would purge the FOUP and transfer module with nitrogen to protect coatings from oxidation as well.
I’m pretty sure that Gordon Moore said that the price per transistor would halve every 18 months, later revised to 2 years. Nothing about speed. Moore’s Law is often misrepresented or misquoted. The end result has been similar. Processor clock speeds have stuttered and the increases have gotten smaller, getting 5% to 10% faster each generation rather than twice as fast, all whilst chips are produced with multiple cores on a single die as transistor density have increased, delivering cheaper transistors, just as Moore predicted. Apart from this little thing I found this tremendously informative. Subscribed.
Thank you sir, I wish my educators were as clear and entertaining as you-my life might have been very different (not that its bad, afterall-I do have the world in my hands)
had to dislike because of 'Moore's Law' which should be stated, isnt a law, and has been entirely wrong for the past 5 years if not longer. also he is neither a scientist nor a law maker, so it shouldnt be a law. it was only ever a guideline or goal.
LMAO Intel is still on the 14nm node. We already have widespread 7nm node chips. I cant believe all of the vulnerabilities discovered in Intel chips have put them back nearly a whole decade.
Despite the never ending stream of bad decisions by UA-cam, it's nice to see at least one algorithm output is an improvement. Great channel, and a real joy to watch.
I work at a pretty well known university and have got to tell you, we don't have many people who can deliver a lecture so concisely
Have you never tried weed?
@@fss1704 No. Do you recommend it?
@@simon199418 Damn, not exactly, but if you want to have deep well meaned and thoughtfull conversations you might wanna hang up with smokers, it`s 1000% easier to focus deep on subjects.
You can`t find the intelligent people because they`re all smoking weed out of the class, obviously there`s some dumb people but even dumb people smoking weed are smarter than the average joe.
@@fss1704 you sound like you're trying to recruit for a cult
I've watched enough videos he's in that I can tell what he's writing down by listening to the squeaking.
Your content is fantastic. Thanks!
FINALLY. A professor that's figured out how to use a slideshow/presentation without fumbling around for the first 5 minutes trying to figure out why it's not working
Moor from Intel back then: 11nm in 2015
Intel in 2019: 14++++++++++++++
Intel also needs to be remembered about that 'same price' thing
@@cleitonfelipe2092 As do AMD, their prices have jumped +$50 this generation.
It used to be that the glass pattern mask was the same size as the silicon wafer but as time moved on the computer chips got smaller and the wafers got bigger. Now the pattern masks do only one chip and the pattern is stepped across the wafer until all the chips on the wafer are exposed. The precision needed to step across all the chips in a wafer and then repeat it for the 40 or so different patterns masks while keeping all the lines aligned is extraordinarily complex.
A big thank you from a retired chip designer! In a remarkable short time, one becomes
far behind the current technology. Thanks for the update. I am still in awe of the results
achieved by the process development people...............
10nm is 10^-8 not -9
Yea I caught that one right away too.
@paul snor "On September 10, 2019, Apple announced their A13 Bionic chip used in iPhone 11 built using TSMC's 2nd gen 7nm process." Try to keep up. 5 nm coming in 2020.
@paul snor www.eetimes.com/document.asp?doc_id=1333881# I don't know enough about advanced lithography to argue this any further. Press releases indicate that 7 nm currently in production uses 13.5nm EUV.
@paul snor You can, if you get clever with the masks and leverage diffraction effects. The light sources and optics get really expensive down in the EUV range.
I've just found this channel an hour ago. These lectures are so good and ther are so many of them I literally want to cry!!!
oh my goodness, I sure hope this is not just a wet dream!
0.000000001 is 1nm not 10nm. 10^9 is 1nm not 10nm.
Not to be pedantic with a pedant, but 10^9 is a billion, not a billionth. Prefix giga. And neither dimensionless number is a unit of length...
Sorry. Did I say "not to be pedantic?" I meant to say "to be totally pedantic."
I would love to see a similar lecture on the evolution (and perceived limitations) of network switches, from the single wire of the telegram to current optic fiber technologies.
In all cases, thank you for an enjoyable and informative lecture
I wish my pay check doubled every two years.
All old people would be rich driving up inflation like mad then.
What an amazing professor! It's rare that a great researcher is also a great educator. Really enjoying these.
Working in dry etch was my favorite job, we made the optane memory chips for Intel and crucial, with some of our tools hitting 12nm and the aspect ratio of our tools and process was etching at 40:1, another interesting aspect is that the wavelength of light including uv used to expose the photo lithography mask is to large to exposes the design and uses water to refract the light to expose the mask clearly.
The teacher I never had 😞😞😞
I work in the semi conductor industry myself and I find this video to be the best possible explanation of where the production comes from and where it is going to
IMO the key phrase in the entire video is a bit tucked away: "at the same price".
Your contribution to the world is unparalleled in the world of cell phone belt clips.
Love how you flip education on it's head. FIRST you explain a complex system in simple language so a non-technical person gets excited (and not intimidated) by the subject......then they have the motivations to put in the effort to learn (and really understand) the math, physics, chemistry, material science, etc. needed to create and improve the system & subject that excites them. Current education system requires someone to master math, physics, chemistry, material science, etc. BEFORE, they really get a chance to play the various system and then hope they find an area that really excites them. As a semi-retired engineer w an MBA, it's still the "show me" (don't just tell me) professors & TAs I remember most! I also love how you tie everything back to the economics of the marketplace for the device or system. It takes money to improve technology, and investors to provide the money......but they need to be paid well for taking the risk. The economic circle of innovation.
Totally. It's really hard to keep the big picture simple and in context before and during a technical lesson like this. I've seen a few of EnergyProf's lectures and he seems remarkably good at it.
I have spent a lot of time researching how chips are made, I find it fascinating how small everything is. I have taken the covers off chips and put them under a microscope to look at the features. This was a great video, this technology has advanced more than I was aware of. New subscriber.
1:48 to point EEEE EEEE EEEE EEEE EEEE EEEE EEEE EEEE EEEE 1 meters
He teaches classes on radiation and keeps his phone on a holder at the side of his hips.
This guy knows something...
makes me feel horrible that I've ever dropped my cell phone.
The graph at 06:17 has a glaring error. That spiked axis was accidentally labeled zero! Of course, you cannot pass through zero while asymptotically approaching zero. It is indeed 1 (one), or ten to the zero power. It must be fixed. It triggers my OCD.
John Bond and he said half as small when he should've said twice as small! It's appalling.
*Professor* "Your CPU is hitting the level of mouse brain for computational power...isn't that impressive"
*People* "Yes it is...Now lets watch TikTok on that machine!"
AMD has already released 7nm CPUs and they're trying for 5nm in the coming years...
This was recorded several years ago, only uploaded recently.
@@JohnMaxGriffin I know, I watched the video.
@@zedg7473 Oh alright I misinterpreted you. Sorry about that!
@@JohnMaxGriffin hey no worries, it's all good 😅
The nm scale is not an industry standard. In general the Xnm scale refers to the smallest unit of measurement that can be addressed not the size of a transistor. Marketing has required chip makers to lower their nm quotes to denote a new node. A 14 to 7 nm shift doesn't always work out to double the capacity. AMD is using TSMC's. 7nm node for the CCX but are running all of the I/O on Global Foundries' 12 and 14 nm node. A Frankenstein monster that is a very effective means of producing high core count microprocessor.
What Engineering Mathemagical devices really are, in applied techniques.
When human DNA was finally deciphered, the real discovery was how much further understanding there was to go, so I'm not worried about robot take over, more like the work required to usefully apply the simpler devices this far.
Must be another word another order of intensity beyond awe, for each advancement for this stuff? Awe to the nth +1...
In the seventys I wish I had fuel injection instead of carburetors thinking fuel milage would be better I had a 350 chevy and could tune it to get 20 miles per gallon, now I have a 6 ltr and fuel injection the best I can do is 14.4 mpg. Very Disappointed
We went from 56Kb (modem) to 1.5 Mb (DSL) in under 5 years. That’s the biggest jump in the smallest time frame I’ve heard of.
Even the air change constantly. Deepholes must drilled into ground inorder to prevent vibrations from earth. Nearly all chips are designed in India. They dont however make any chip except for military purposes. Same method used for anti radar coated air craft cockpits i.e. sputtering.
I really like his "clips". I wish I was one of his students cause I really would've enjoyed attending his classes.
we should apart from improving the computations per unit time per unit volume of circuit, try to decrease the ultimate need for large quantities of simple arithmetic computations per unit time, or lots of arithmetics, per user or per mankind.
better mathematics and physics should in theory reduce the dependence on lots of arithmetics. better formulae should reduce the need for large quantities of simple arithmetic operations.
there is after all a finite limit to miniaturisation of physical devices. human progress is not likely to have any limits. therefore the ultimate future of human progress lies not in lots of arithmetics but in freedom from lots of arithmetics.
Kinda ironic, but the more complex something is, the better it gets with time.
But it also makes sense, since more complex things have way more room for improvement.
Absolutely fascinating!
The sophistication level of that collection of technological innovations is a mega achievement.
When you think about what most electronic circuits involve ie: resistors, capacitors, transistors, inductors, and you think back to just a few decades ago where all of these components were placed onto a circuit board that had visible metal traces connecting them all with holes drilled through contact points for soldering on the components. Now it's at the atomic level using photon energy to install the circuits at microscopic scale.
What Plasmas Have to Do with Computer Chips? Well I should know in about 23 minutes...
The secret is not that you can build one chip, but it's possible to make thousands with high yield and make a profit.
Thanks lot Sir, the almighty creator have given you a best brain and you also do better for us. May the almighty give u best ability and oportunity to do better something for the world. May the creator keep feet u every time.
My phone is 7nm.
2:00 Hmmm ... is it me or did he say 0.1m is 1 cm ... shouldn’t it be 0.01m is 1cm?
Cool videos but you need some different markers. lol
19:54 "The future is now" Maaan. That sounds so cool )
ive been using the same cell for last 4 years and same with PCs & laptops.
android 9 runs terrible because it is a full blown dragnet surveillance OS. same for windows 10.
Same Samsung R451C 'slider'phone since 2010.
Same battery too. Leaving it connected to the charger a lot helps ...
very intersting...thanku very much
EUV - he is actually describing what intel is having trouble with the last 5 years plus ... and moore's law is dead. At 3nm or less it will be very very hard to manufacture (down to single atoms)... every gen 18 months or so, new CPUs are only faster by 10-20% and the improvements are getting slower and slower and taking longer and longer. Without a huge breakthrough by 2030 computers will last 10 years plus.
That's what the extra cores are for. Going faster (smaller) is becoming harder, instead thread parallelism shares out the workload. Core to core todays CPUs aren't much faster than 5 years ago but when running software that scales over multiple cores they can be several times faster.
Meanwhile the Cdyn is out of control.
More technophile's dream world..
9 people want larger transistors.
3:32 .. "a gold mine for consumers" ... A gold mine for someone for sure...
was surprised he said consumers XD
Outstanding lecture.
Amazing the drive that capitalism produces.
If only there were a way for each individual person to have to buy their own personal energy source for everything they do. Just imagine the results!
Right. Some people would buy from wind and solar, paying fifty cents a kilowatt hour. Some people would buy from nuclear plants, and pay eight cents a kilowatt hour.
6:44 thats a bullshit graph.
At 1:05, the inventor was Jack Kilby (not Kirby), who subsequently won the Nobel Prize in Physics for the discovery. IIRC, he held a Master's Degree, not a PhD.
I searched for this video son 😝
This was filmed at a somewhat perfect time because Moore's law pretty much petered out around 2015 2016 it's now 2021 and Intel's 10nm is only now a viable node at scale.
TSMC is arguably continuing Moore's law as they're now upto 5nm which is utilised by Apple but the costs of wafers on new nodes is growing massively. I think 5nm at TSMC is now above 10K Us a wafer whereas 16nm was less than half of that I think when new.
I used to write control software for Plasma Deposition and Sputtering machines. It was great fun making all the devices work together to process wafers. Turbo vacuum pumps that ran at 33,000 RPM, Cryo Pumps that cooled down to a few Kelvin in order to scrape a few more gas molecules from the chambers. All the way to rotating assemblies weighing several tons spinning at 60 RPM with enough energy to throw the machine through the wall if they suddenly stopped. Oh, and we also had plenty of flashing LEDs ;)
I've always thought the biggest leap in computing would be going from binary computation to omninary (my crappy made up word). In other words moving away from on off bits to multi-value bits. Like a bit not just equaling on or off but say 256 levels of on or off. The entire software industry would have to be reworked of course. But you could move a variable of 256 (or whatever number) in the same time as on or off. It would be huge.
We can make chips with 10 nm features, but we still can't spell "transistors" reliably, and we still write with markers (although on a transparent surface, and "backward"-got to admit, that's pretty cool).
Fantastic lecture
"10cm to 10 nano-meters" Unless you're Intel amirite?
14 nm for 6 years kekw
I like this guy's style. I mean his delivery but also his fat ties.
His delivery reminds me of a beloved statistics prof at my university, he had a very similar way of speaking and delivering the content.
4': Good point: Moore's Law doesn't happen because of any hard law of physics. Physics simply allows it, but what really makes it happen is economics: Businesses and other organizations make plans to cause those increases to happen.
Hui, First i thought about plasma From our sun. I Hope all the Important Server on Earth will protected From sun system plasmas
This is a great lecture. There is a minor error though: The exposure performed by scanner-steppers is performed one dye (circuit) at a time (=stepping). Each dye is exposed by scanning a slit shape image across the dye (=scanning).
en.wikipedia.org/wiki/Stepper#Scanners. I cannot grasp how it is still possible to expose 100 wafers per hour with a precision in the nm range.
I believe the economics of computer chips is very simple, the smarter the chip, the smarter it _can_ make you and it's always economical to be smarter.
SO DOES PLASMA ETCHING PRODUCE GAMMA/BETA RADIATION?
No, it produces uv radiation. But it does use a lot of rf radiation to generate the plasma though.
Professor, I have watched just about all of your videos. When will I be getting my degree in the mail? Zab
And Mr. Professor Ruzic i will say thank you for the best way of Education. You are great 👍🏼
i absolutely awestruck. i don't think there's any technology out there as cutting edge as chip production. or as vital.
How can you compare the computations a logic circuit does to the "computations" a brain does?
Chips are small, but what if we could make these same fast chips but in a larger NM. I could live with a physical size of s chip like a CPU if the chip is faster and cheaper. I’m a novice but does use of larger NM chip slow it down?
It comes down to economics. You can fit more smaller chips on a 300 mm wafer than larger chips. It would cost more money to scale up than to scale down.
It's about the speed of electricity. If the chip is larger, the electricity has to run more km's to do the same amount of work, which means it takes a longer time.
Intel actually reintroduced a 22nm node because their fabs have not been able to meet demand for 14 nm while they have been stuck trying to transfer to a 10 nm node. The main reason that smaller is better is because the biggest margins are from the data center where the cost of operation in electricity is as important as compute performance. Smaller nodes can traditionally run at higher efficiency.
@@puffthemagiclepton7534 is right, its economics a 300mm wafer where i used to work was about $20k from all the different processes, materials and energy. It would take over an hour and a half to run a single process in my area and could have over 800 processes to complete so realestate on the wafer is super important, you want to fit as many chips on it to get the biggest yield possible
2 nm (2021)
If this guy was my dad I would be successful. He has great dad energy.
If we make it in 10 years I hereby volunteer to be the first to upload.
wish i could subscribe to this channel twice, great stuff
Did you need a 700 watt power supply for your PC in 1995? No way! Surely we have much better chips nowadays but they consume more energy to operate.
Lol I recently built a new computer after having mine since the i7 920. And I was SHOCKED to see how little wattage PSUs were nowadays. My ATI 5870 + 920 required a 1500 watt PSU. Now I have 1000x the processing power on 650 watts.
Cant you use these methods to make Graphene type batteries?
This is very interesting very well made. Ty.
21:37 Its places like this where the future is made
I thought he teaches nuclear physics?
19:49 I am doing a FAB build now 2019. This is a clean room with people. This area is considered to dirty for the chip and will never see this room. Instead the chips live in a box that can be even more cleaner. The box itself is only open at inside the tool.
Until AMHS takes your FOUP across the fab with the door still stuck to the ATM lol. We had N2EFEM ATMs that would purge the FOUP and transfer module with nitrogen to protect coatings from oxidation as well.
i missed the part where you use the plasma.
I really like your lectures.
I think I found a goldmine.
Rejoice people. The EUV is here!
Someone could think that this video was created in the 80s. xD
..prof didn't drop the style since reagan .. he is not even putting the phone in his pocket !
Wonderful discussion.
Nice
Double, double toil and trouble;
Fire burn, and cauldron bubble.
"Shakespeare ... what a cornball!" -- Grandpa Munster.
Great channel.
Top Professor
Beautiful
This guy is amazing at writing in mirror-image.
Or his videos are mirrored?
I’m pretty sure that Gordon Moore said that the price per transistor would halve every 18 months, later revised to 2 years. Nothing about speed. Moore’s Law is often misrepresented or misquoted. The end result has been similar. Processor clock speeds have stuttered and the increases have gotten smaller, getting 5% to 10% faster each generation rather than twice as fast, all whilst chips are produced with multiple cores on a single die as transistor density have increased, delivering cheaper transistors, just as Moore predicted.
Apart from this little thing I found this tremendously informative. Subscribed.
That first IC, from 50's, what was it used for?
A doorstop
A transistor amplifier
1m=100cm
you sure?
@@DarthZackTheFirstI you're not?
how can i be with the quantum development. we always found something more tiny.
Thank you sir, I wish my educators were as clear and entertaining as you-my life might have been very different (not that its bad, afterall-I do have the world in my hands)
had to dislike because of 'Moore's Law' which should be stated, isnt a law, and has been entirely wrong for the past 5 years if not longer.
also he is neither a scientist nor a law maker, so it shouldnt be a law. it was only ever a guideline or goal.
LMAO Intel is still on the 14nm node. We already have widespread 7nm node chips. I cant believe all of the vulnerabilities discovered in Intel chips have put them back nearly a whole decade.
All for the sake of faster devices that allows us to watch porn and the latest celebrities mischifes. Way to go, Progress.
Speak for yourself...
Jim Lahey how are you still alive? Superposition?
@@alanmalcheski8882 it's the liquor talking
Yeah, speak for yourself. Some of us only watch porn.
Computer for single guys only has 3 buttons on the keypad, ON, OFF, and PORN. And that was introduced 20 years ago.