The Death of the Modern Transistor and Moore's Law

- Start your free trial at Videoblocks by clicking today! bit.ly/2znDE4W
--Our Studio Gear on Amazon--
Panasonic Lumix G85 (Primary): amzn.to/2zoJ1AX
StudioPRO 1000W Softboxes: amzn.to/2zCl5e5
Audio-Technica AT2035: amzn.to/2yblmk1
Yamaha MG10XU Mixer: amzn.to/2zCP2e6
Panasonic FZ1000 (Secondary): amzn.to/2iBoM8U
Neewer Microphone Arm: amzn.to/2h9FEn3
Alphacool Build Mat: amzn.to/2zmXXzu
PATREON: / sciencestudioyt
TWITCH: / sciencestudioyt
FACEBOOK: / sciencestudioyt
TWITTER: / sciencestudioyt
INSTAGRAM: / sciencestudioyt
Subscribe to the channel!
MUSIC:
'Viridian' by Shahin & Bench
Provided by Silk Music
/ silkmusic
DISCLOSURES: All Genius links are tied to our Amazon Associate account, from which we earn a small sales commission. Links containing a 'bit.ly' reference forwarding to Newegg are tied to our CJ account, from which we earn a small sales commission. All sponsored links and comments will contain the word "SPONSOR" or "AD." Any additional revenue stream will be disclosed with similar verbiage.

[@ccmaster86]

"A single atom was used as a transistor.." It's like all the R/D guys were like you think that's small..hold my beer.

[@vnyggi621]

ccmaster86 So... Quark transistors?

[@99stem]

icndjfk nsbdbd I think he was referring to something else...

[@Trident_Euclid]

icndjfk nsbdbd Fermions

[@lodbldbol]

R & D guys don't drink beer, they drink knowledge juice.

[@jayteknica1175]

Basically lol nerds are competitive.

[@m15anthr0pe]

Awesome video! Back in 2006, I interned with a professor whose research focused on Quantum-Well Intermixing. I had no idea what he was talking about, but he explained how his reasearch could benefit the push for Quantum computing and how powerful those machines would be. That was a very interesting summer.

[@pj491]

m15anthr0pe those good old days mate :)

[@92juanreus]

I would like to say that not only the content is great, but the visual quality of your videos has increased A LOT! It's awesome to see the way you are learning and growing this fantastic channel.

[@DIYTech21]

This video made me think like some 70 year old. Ahh how far we have come. We use to have Processors as big as a room back then.

[@Demontorman]

Do you remember Pentium4? :D

[@DIYTech21]

If m not mistaken there were 2 type of Pentium 4 one had pins too. While the other was pin less. Then came the p 4 HT and Pentium D and the dualcores and xeons and core 2 duos and God knows what :D

[@_Daio_]

I started gaming on a binatone tv master and now i play on a 3440 x 1440 ultrawide monitor,1080ti driving it. I often stop in a game and take in the sights and think Ahh look how far we have come. I'm 48 so no need to go back that far:)

[@Chuckiele]

My first computer had an AMD K7 500MHz CPU that was a card inserted into a slot on the motherboard instead of a socket. The stock cooler was the case of the CPU itself with two fans attached to it.

[@DIYTech21]

Chuckiele i remember something like that too. I think pentium 3 were those slot like GPU like processor too with fan on them and appearance just like a 64 bit GPU

[@jdtechgear]

I think the next big thing with processors isn't quantum computing since it's still far ahead of us, but instead change the medium that the electrons pass through. Electrons are slow to begin with, but there are proposed ideas to replace silicon with graphene (once it's fully developed), so that the electron can travel much faster. I think there a few different mediums that are being looked at, but if we're able to crack the case for mass production of graphene, then it will be a much more feasible transition to a new age of computing (currently) compared to quantum processors.

[@Machinens]

JDTechGear Everyone make this comment higher!!

[@Gamevet]

2,200 Kilometers per second isn't what I'd call slow.

[@julian23561]

Well, it ain't slow in the natural sense, but compare that to graphene, which is a 100 times faster. lol

[@garrettgrosso48]

How would that speed processing up? A CPU is so small, electron speed would make such a tiny difference.

[@sirnathanieltayloresq7800]

that isnt necessarily true. Look up processors using light. In theory they will be much faster and use much less energy than our silicon based processors (lots of issues with using light unfortunately). The idea with graphene is that since the entire platform will allow electrons to flow faster with almost 0 resistance, it will use less energy to transfer electrons over a large distance (between cores), as well as making the entire flow faster. i am really tired when i typed this so i hope it made sense

[@JamesV1326]

5:49 did somebody say... CARBON NANOTUUUUUBES?!

[@paddydoestech]

JamesTheSoundGuy holy shit you watch netlinked

[@Mastrhix]

RIP netlinked

[@Aerith17]

Every time I hear Carbon Nanotubes I need to act it out

[@hellblazer3781]

Ahh.. Good ol' netlinked daily.. Shame they abandoned the channel

[@mikesama5458]

RIP Netlinked...we all loved that show right after that morning coffee...oh good old days...

[@alexandramiller9887]

i am wondering if we are going to get *stacked cpus* .... somewhat similar to HBM module stacking..

[@bruh2595]

Cooling will be a problem but we might see it in the future

[@TheDz1991]

Friedrich Miller that seems like a hot fix for now but I guess it works!

[@snrd]

That would be possible with Fredkin gates since they wouldn't draw any power and wouldn't emit any heat. At least this applies for theoretical perfect Fredkin gates.

[@andrewrobinson2985]

What are the implications on heat dissipation?

[@crazed357]

Im sure they will figure out a cooling solution

[@justuslm]

Great video! It's also nice to finally see a person who realizes (and spreads the knowledge) that quantum computers aren't really useful for most of today's everyday computing tasks.

[@sever9268]

AMD navi will be skipping 10nm and going directly to 7nm

[@sever9268]

Kap they said 7nm on the roadmap. It's been said for quite some time now. I think I have heard it's because 10 nm was difficult to develop so it would have been better to simply skip it and develop 7nm. It would also make sense since also intel is been using the same 14 nm since skylake which is quite old now. This makes it 3 generations, and i believe their roadmap was to do only 2 generations with the same size

[@eliadbu]

biropop their definition of 7 nm would probably differ from intel 10 nm or TSMC or samsung, for instance samsung 10 nm is a lot closer to intel 14 nm in terms of density and size than intel 10 nm. and global foundries in last generation were much behind than all other so i doubt their 7nm is that advanced.

[@devindykstra]

biropop AMD time... remember...Vega was supposed to launch late 2016, Navi won't be here for quite some time.

[@sever9268]

Devin Dykstra true

[@malteroeper3723]

vega was never supposed to launch 2016 some idiots claimed it would do and people believed it but amd never said vega will launch in 2016

[@brianazeri]

"DEATH" is this video get unmonetized? :0

[@MrDeeb8]

surprise yes

[@lontongtepungroti2777]

toxic

[@chemp231]

Because of this comment the channel has had to claim bankruptcy and no more videos will be uploaded (jokes jokes)

[@HueghMungus]

I hope this is unmonetized, fucking misleading and just fearmongering and clickbait title. In the video he said that this won't be the end of Transistors opposed to what he wrote in title, then what the fuck was the point with the title? He just talks about what people already can find out on the internet themselves, looking up Moors Law. And by logical thinking traditional computers and the transistors in them won't die out.

[@Shadow77999]

Youre comment will be unmonetized

[@snrd]

It would be interesting if you could check out some non-standard circuit designs like reversible and asynchronous circuits that Moore's Law essentially prevented from being used in large scale.

[@nonayirbusnez3811]

I sometimes think the same. Sometimes I think that since money runs the world, the cheapest computational power should prevail. Eg. neural networks, everyone knew about them since before I was born, but only recently became useful thanks to your everyday cheap transistors. Having said that, a lot of great/better ideas don't eventuate (is that not a word?) due to long standing standards/traditions, qwerty for example.

[@whocares_yes]

Quantum computers actually can't manipulate data like we do with transistors. It uses and gives probabilities instead of giving full answers and using 1's and 0's. Also a full quantum computer can't exist and would be ridiculous, the only quantum part would be the processor ;) The other thing is that QC needs a LOT of cooling, it is really really close to absolute zero. Would you imagine that in a laptop?

[@vnyggi621]

Who Cares Wasnt the catch that higher temperature means more 'noise' and less exact answers? Or something like that

[@abhishekgourav6144]

Who Cares yeah Linus covered this quantum computer topic...the cooling system was enormous...I think the temp that was required was 0.15 K ..not sure..

[@whocares_yes]

icndjfk nsbdbd: Exactly, as the Qbits inside the processor would start to gain some unwanted properties like moving unexpectedly.

[@kalebbruwer]

Can you imagine a future where a quantum computer only takes up HALF a room?

[@vincenthodges647]

Who Cares who Cares

[@surajsuresh8945]

As a non engineering student, this video made me wish I took an engineering course.

[@ThexXxXxOLOxXxXx]

Suraj Surash You dont need to be an engneer to be a student of engineering.

[@redsquirrel3893]

All the information is avalible for free online for you to learn if your intersted.

[@anandsuralkar2947]

Nice

[@anandsuralkar2947]

As suraj been indian i can see that even though he can study by himself..he cant match knowledge of engineering bcz education in india is very strict whatever he is studying he will need to give his fullest..work and time so he would probably cant study engineering..i m being engineer..i litereally have no time and space to study anything else 10hrs college everyday..exams every week and all u have left with no time for other works..they dont even provide..summer leave or leave between semesters .its just college whole year for all four yeat only one month leave for every year which in tern..gets accupied by pending college exams and projects

[@steffeeH]

Quantum computers will probably only be used as servers aimed at complex calculations for a very long time ahead. Not only are they insanely expensive, but due to the extreme magnetic shielding and low operating temperature (-273.135 C / -459.643 F) required for them to even function, they're far off from the average consumer, unless the consumer don't mind a computer the size of a bathroom. And no, this one is not easy to shrink in size. And as stated in the video, they're not faster, just more powerful.

[@Machinens]

steffeeH Well the history will repeat it itself no doubt

[@BurcoGames]

GREG WHY ARE YOUR VIDEOS SO GOOD?! I can't stop watching these. Honestly. So much learning!

[@abhishekgourav6144]

Sh0kGaming yeah..his videos has got a standard...

[@GregSalazar]

I greatly appreciate it! Glad you're enjoying them. What would you like to see discussed in a future video?

[@FRIENDPIRA]

why you dont decide to shave

[@GregSalazar]

I let it grow and shave all the time. Some like it with, some without. Can't please everyone.

[@BurcoGames]

I personally like the beard look. But... maybe that's because I have a beard... lol

[@SaturnusDK]

Your assessment of quantum computers reminds me of the famous Thomas Watson, Chairman of IBM, quote: ""I think there is a world market for maybe five computers". While at the time he said it was probably accurate as a computer with less computing power than a microwave oven was the size a gymnasium. In a historical context it's hilarious though. And so will your statement about quantum computers be seen.

[@vnyggi621]

Samsonite Dove Maybe affordable QCs would create a usecase for normies, but right now its pretty useless, I mean what are you gonna do with a pc thats like: 1+1? Hmm thats 99.999...% =2 and 0.00000...% =3 and 0.0000...% =8264 and so on

[@SaturnusDK]

lol... same reasoning as good old Thomas... indeed, what possible use can increased processing power have for any one? The same was said about the invention of the laser. Neat but entirely useless for anything, right?

[@GregSalazar]

100 years into the future from any point in time, then-modern assertions are almost always seen as silly in retrospect. You're literally stating the obvious (and to be frank, so was I).

[@Ridiric]

the biggest problem with quantum computing becoming smaller is that it requires ridiculously low temperature to work ( minimum is about -272.35 Celsius for something like the D-Wave quantum computer but the closer to absolute zero you can get the better) and there is no way to make a cooling solution that needs to go to such low temperatures small (i mean think of the size of a modern air conditioner, fridge or freezer system and none of those even approach the type of temperatures needed and they are already quite large), so it doesn't really compare to the Thomas Watson quote, because the cooling technology used in modern computers is still just as simple as it was back then and doesn't need to become much more complex, heat-sinks and water-cooling (which have been around for hundreds of years btw) are room temperature + type cooling when you start needing to go into bellow room temperature cooling the size and complexity of the equipment needed to achieve those results goes up and up and the lower the temperature you need to achieve also the higher the energy cost to achieve that temperature.

[@SaturnusDK]

@webbysaurus. At the moment maybe. However, there's nothing in the operation of a quantum computer that dictates that advances in material sciences can't in the future make it possible to build one functioning at room temperature.

[@johnjohnson5818]

I am a 70yo. We have come very far. Very far.

[@linagee]

Young people talk to their phones, have graphic masks added to their video through neural networks, are advertised to through target advertising neural networks, have their location constantly recorded... Most people have trouble seeing forward or backward through time more than like a decade.

[@R.-.]

It's worth mentioning that recent shrinks in process size haven't yielded the benefits they once did, but the R&D costs are still very expensive. CPU clock speeds no longer increase notably. Power consumption reduces, but not by as much as in the past due to increased current leakage. The main performance improvements are due to design changes that use more transistors, but even there the potential to increase single-thread performance is approaching a limit. If technology improves to the point where individual atoms can be manipulated, we will enter the age of nanotechnology. That is moving mechanical computers constructed in 3 dimensions. Once that is attained they will be able to self-replicate and create any physical object from other atoms.

[@Psychx_]

Nice that you mentioned carbon nanotubes, but for me a mention of "graphene transistors" was missing. Wouldn't these allow for significantly higher clockspeeds within the same thermal envelope and same transistor count/die size?

[@GregSalazar]

There are dozens of solutions. Not gonna mention every single one of them.

[@TUTOSANDROIDtutorialesymas]

love this type of videos , they are perfectly planed and well scripted , and pretty well edited in a simple but fully understadable , thanks for the info!.

[@Andras889]

There is actually a limit in phisics for the current way of computing. A phisicist calculated, that with the current chip size, at a limit of about 2^50 flops, the calculation would require energy to be so dense, that it would create a black hole. Currently we can achive 2^19 flops so there is quite a long way there.

[@cj09beira]

the thing is that we now have the capability to produce cpus with multiple dies, and doing so we increase the max area of the chip a lot while also making it cheaper :)

[@vnyggi621]

That blackhole part sounds fishy

[@saladwithsalad]

Black hole? Seems legit.

[@tanzimmiah2357]

Flamerking1 Yeah, all we need is to contain the energy released by a supernova to use that processor...

[@Andras889]

source: arxiv.org/abs/quant-ph/9908043 you can download it in PDF

[@The13thSword]

I NEVER "thumbs up" ANY videos on YT, but just had to on this one. A very good, straight-forward explanation of the subject, which even an old & techno out of touch guy like me could understand. Well done sir!

[@rossq9432]

Good video, but there is another alternative other than Quantum computers, and that is Spintronics. Which involves the spin of an atom is used to represent 1s and 0s. Plus it is uses a lot less power than the traditional transistors.

[@anandsuralkar2947]

How does that work though

[@dogodogo5891]

thats quantum subpart

[@NeilRieck]

Speculation: I have always assumed that electrons were able to tunnel because of their wave nature (eg. tunneling electrons are out of phase with the thin material they are traversing). If correct, then electrons should be thought of a radio waves (now think antenna design) so all you would need to do is tune the electron wavelength (via voltage) as well as chip geometry so the the electron wave was always "in phase" anywhere in the boundary . This calculation of the de Broglie wavelength of an electron: ecee.colorado.edu/~bart/book/ex009.htm shows that 2eV yields 0.87 nm. Ignoring the half-wave length because it would be impractical, the length of 1.5 waves would be 1.3 nm while the length of 2 waves would be 1.7 nm. I suspect that all the big chip manufactures will (eventually) get down to 2.6 nm. Of course, lower voltages will translate into smaller features (but you can only go so far with band-gap tuning)

[@MrCykko]

Greg can we get married? I work at a chinese food buffet place and get a sick discount.

[@therealb888]

MrCykko I don't think he's gay...

[@cakemnstr42]

As a computer science student those videos are giving a nice reference of what the average person know about computer-internals: not much^^ And I'm speaking of the comments like "learning soo much", not about the video. So it's really good that you make these videos!

[@junwhang6293]

but net improvement of cpu technology (and net end benefit to consumers) has been deminishing rapidly. Olden days, decreasing lithography size meant huge jump in capability - think MOS 6502, Intel 8088, 80286, 386, 486, pentium etc. 10 years used to mean huge jump in cpu technology, and net benefits to end users. Not any more. Decreasing lithography size means very little if you can't produce quantum leap in capability and features! In 2018, I can still use core 2 duo from 2008 perfectly fine for nearly everything! New doesn't necessarily mean (much) better any more.

[@TimbavatiLion]

When a single atom is used as a transistor, you still have the problem of connecting it. Data has to go in and out somehow, so i don't think we'll ever get down to that size. My guess would be 0.1-0.2nm transistors to be the end, where we could potentially shrink the size of the transistors further, but not the data lanes connecting everything.

[@nosirrahx]

I am starting to wonder if we will see a stopgap technology emerge. I picture a dual CPU system where 1 CPU (4 cores) is designed with only speed in mind (maybe 6 ghz or higher). The other CPU would be clocked far lower but with upwards of 64 cores. If something like this existed legacy software would get a massive boost while newer software focused on threads would also perform quite well. This would give scientists time to develop a complete replacement while we drain the last drops of performance from our current technology.

[@JB52520]

What you're describing is heterogeneous computing, which is something we essentially already have, except for the extreme single core clock speed, which we can't do. The power draw required to achieve it would immediately melt the chip (unless it's cooled with liquid helium or nitrogen). CPUs generally handle serial operations, and GPUs handle massively parallel ones. On the GPU, you define a problem in terms of a 1d, 2d, or 3d grid, and each cell gets its own thread. The number of threads running at any one time depends on the number of GPU cores and how well they can be kept busy (which is quite a challenge). There can literally be 10s of thousands of threads running at any one time. Each thread is far slower than on a CPU, but when added up, the performance gain can be astonishing. With integrated graphics, you can gain the parallelism of a GPU with none of the massive latency of traveling across the motherboard. It's the whole idea behind AMD's APUs, which are used in modern game consoles. Heterogeneous computing is a great idea. Check out CUDA and OpenCL to get an idea of what's possible. In the future, we may also see FPGAs and neural network accelerators integrated with CPUs as well.

[@nosirrahx]

Its pretty obvious that things are stagnating badly. Anyone that has taken an oldish CPU with decent clock speed and paired it with a SSD and new graphics card knows the score. SSD and GPU improvements have heavily obscured just how bad CPU stagnation is.

[@MyAsdfqwe]

these faster cores would probably be done with GaAs. but first we have to squeeze the ipc, which means to the extend of 1-2nm cpus.

[@BrawlerDoge]

@@nosirrahx you were right!!! Intel 12th gen is exactly that

[@r.d.5882]

One of the more interesting videos I have seen. Great explanation of the intricacies of transistor leakage like quantum tunneling and the options to mitigate the problem. Thanks for the video!

[@albertkropp]

There are other options besides typical electronic transistors and quantum transistors. For example, there is ongoing development to create an Optical transistor which uses light rather than electricity. This would greatly increase the speed of the transistor, and would allow roughly a 200% increase in speed before you even have the same heat output that a traditional electronic transistor outputs. If they are ever able to perfect this new technology, everything we currently have on the consumer market will effectively be outdated. While they have created sample processors using this new technology, the next problem they face is shrinking the transistor size, and engineering some new way to filter the light through the transistors so that they don't bleed light and cause interference between transistors.

[@Machinens]

Paul Bäumer This comment musn't be lost in twisting nether

[@atomm7316]

Paul Bäumer do you have link for the source? with light they will be dealing with problems that arise with photons. But yeah sounds like a great solution.

[@albertkropp]

For the potential of the new transistors, look for optical transistors on the IEEE Spectrum website. Here's an article that explains the transistors themselves in more depth: news.mit.edu/2013/computing-with-light-0704 I'm aware of the problems they face with this new technology, but the fact that they have actually made transistors using light rather than electricity is a huge step in the right direction.

[@redsquirrel3893]

Optical logic gates would be pretty good asuming they still work on nano scales.

[@rednafi]

If you are familiar with the concept of De Brogly wave, then you know that every electron has a wave of it's own. Now when an electron is within a potential barrier, (here the gate provides the potential barrier for the electrons in source and drain) a part the wave always tunnels through the barrier and go to the other side. The thicker the barrier is, the less significant the wave function becomes. Once the physical barrier dips down to 5nm, the potential barrier drops enough for the wave function of the electron to pass through and quantum unpredictability kicks in.

[@parsastrife6629]

I believe I read somewhere that TSMC has the capability of producing 3nm process

[@pcsproshop8972]

Great way to express this process. Real science, or at least the search for real answers, is sorely lacking (in general) on this medium. While "antics" are quite enjoyable to watch, it's quite refreshing when you delve deeper. Thank you for your work, keep it up!!!

[@doudoud82]

if you can't make it any smaller than make the chip bigger with more cores ! at the smaller transistor nodes size and opt for more parallelism workload.

[@anandsuralkar2947]

Apple does that apple's processors are double in size compared to qualcomm ..so its not just apple chips are architechturally faster but they just have double the size so they just do tasks in half time..no innovation here..if u just make apple chip the size of qualcomm apple wont even stand qulcomms performence

[@add3685]

Most UNDERRATED science-tech channel ever!

[@rajarajput3483]

why dont they just make bigger chips, like increase the size of the die

[@kyleherbig]

I'm responding to both your comments. A temporary fix. 2 SOCKETS increases core count, not core size or power. The more cores you try to use for tasks, the harder it is for them to coordinate with each other. And Moblie devices can't really get bigger either.

[@nickmattox6027]

I love these videos you do! Topic ideas for future minute science/crash course - quantum computing, dark matter, or the recent gravitational wave observations over the past year in space

[@allan1447]

nice video man..... good stuff

[@GregSalazar]

Thanks for watching!

[@thunderloong]

Lithography IS NOT metalurgy. 5nm is an accuracy of lithography that we CAN actually beat, there are small area lithographic laboratory equipment that can do 0.5nm focus.

[@1pcfred]

That's some serious black magic there. Being as UV light runs out at 10nm wavelength. Industrially they can't even use that frequency of light either. They've no way of making it.

[@randomdudewalkingdownthest8158]

what are your thoughts about AMD & INTEL teaming up?

[@vnyggi621]

RandomDudeWalkingDownTheStreet lol they could price stuff however they want xD

[@Smorfar]

That would be pretty bad for us

[@ToriRocksAmos]

Yeah they are going to segment the market... AMD for low end laptops, Intel for high end, AMD for a lot of cores on the desktop, Intel for fewer but faster cores. I don't know the specific details of the deal, so I don't know how this will play out for AMD Intel and Nividia. For the customer though... I hope that we won't go back to the 2012 - 2016 era where Intel did little to nothing for the desktop market while keeping the prices high.

[@danielkoontz6732]

Bro... hothardware.com/news/intel-amd-combine-core-processors-radeon-graphics-single-chip

[@randomdudewalkingdownthest8158]

nah it won't, at least i highly doubt it.

[@gumbilicious1]

From what I️ have read on quantum computing, there is no “replacing” transistors with qbits. Qbits still suffer from the uncertainty issues of really small transistors, but qbits use this as a processing feature rather than as a detriment. Thinking quantum computers somehow “fix” the uncertainty problem reveals a certain lack of understanding of computing

[@JackBauerwashere]

Well AMD will skip the 10 nm process and go to 7nm immediately. It will be the first time in a long time that AMD will have the node-lead over Intel . Since Ryzen is already very competetive with Intel on the same node Ryzen 7nm vs canonlake 10 nm will be very interesting. Ryzen 7nm 5Ghz Clocks with 10-20% IPC gain and more power efficiency ? A man can dream.

[@aboalthoq]

you are dreaming / 8700k able to be 5ghz easliy

[@JackBauerwashere]

No dude ryzen 7nm will reach 5Ghz.Wait and see.

[@JackBauerwashere]

Doesnt change my point,because Icelake will also be on the 10 nm process.

[@Malus1531]

JackBauerwashere Did you even watch the video? They just make those numbers up. "7nm" could just be marketing bs, we don't know if it will be more advanced than Intel 10nm.

[@88oscuro]

Amds 7 nm is equal to Intels 10 nm.

[@arthurvin2937]

Quantum tunneling is not about jumping over the barrier, or go trough the barrier, it's all about teleporting on the other side of the barrier. Not much can be done to overcome this phenomena except may be utilize this effect for good which effectively turns it into quantum computing anyway. So, transistor won't get smaller, and major computations will be delegated to cloud computing via network where formfactor is not an issue.

[@jesse-dg8yx]

lower the voltage and then its less likely to jump the switch

[@hjalmarnedin2949]

I doubt it's that simple, sadly

[@atomm7316]

an electron will still be able to "jump". what a misleading term. it's more like it will pass through the barrier, not "jump". Welcome to quantum mechanics.

[@jesse-dg8yx]

Atom M well jump through the air from the switch to the contacts

[@DerkNotDirk]

We literally talked about this today in my Electronic and Computer Engineering module "Embedded Systems and App Programming"

[@shadowdance4666]

Eventually they will produce organic computers… oh wait we already have those they are called brains

[@linagee]

To be fair, there is no evolutionary pressure for intelligence. (Dumb people can have babies just as easily as smart people and they don't breed any more or less because of it.) It's possible we can design neurons in wetware that perform faster than human brains. For one example, a microprocessor can do math instructions a heck of a lot faster than you could in your brain. But if you were laying out neurons in an artificial brain, you might be able to have them in such a configuration that is optimized for mathematical operations moreso than a human brain is.

[@shadowdance4666]

linagee the failure of the human computer as always is deficient programming or it would be faster. A human properly programmed would outpace any synthetic system however it might require a hybrid systems for programming

[@mnoxman]

Technically speaking.... Long ago we harnessed the use of Quantum Tunneling to make high speed triggering circuits for pulse generators and oscilloscopes. These semiconductors are no longer made. Tektronix used to use them extensively in their products. You can kind of simulate their use with Neon bulbs these days but Tunnel Diodes were at one time the only way to get 1-2ns rise times (or less) from a pulse.

[@bliglum]

Nope, you can continue halving distance endlessly yet still never reach. Thus Moore's law will live on for all time!

[@bliglum]

Quantum computing will give Moore's law a second life! He will live on!!!

[@thearmyofiron]

Walther Penne humankind has survived for thousands and millions of years mere 50 years is guaranteed nvn just realized that some deadly virus can spread

[@darkshadowsx5949]

I haven't read anywhere where morse law specifies transistor shrinkage. "Moore's law is the observation that the number of transistors in a dense integrated circuit doubles approximately every two years. ... Moore's law is an observation or projection and not a physical or natural law." NOTE the "observation or projection and not a PHYSICAL or natural law" it doesn't specify a specific area in which the transistors have to fit. just a "dense area". meaning the dense area is free to upscale so long as the transistors remain densely packed. I think so many people misinterpret mores words. Quantum computers most likely won't be a consumer product at all or at least for another 50 years or more. they function entirely different and are best for number crunching work and not social media updating or gaming... they are based on "guess and test as fast as possible". at least right now. problem > random tests > result. it's extremely difficult to persuade a qbit to change to a specific value without destroying it. you still cant observe the result until its finished or see how it came to that result. whereas in modern CPUs the bit value is always there, observable, and changeable at any time. if you wanted to change a qbit value you would have to recalculate the result. like re-rendering an image because you didn't like the position or color of an item. that's at least what I got from my in-depth research on quantum computers. I would like to see them scaled down to the size of a modern computer and sold commercially but until they figure out a much more compact cooling system that is unlikely. fans just dont cut it when you need tempratures below liquid nitrogen.

[@freshkd2172]

Holy balls the PC in the backround looks awesome, especially because of the S shaped hardline tubes. Could someone give me a link to the build video/link to the parts?

[@GregSalazar]

It's only a few videos back on this channel.

[@MrRiver-dj7us]

Blind Bob There's nothing wrong with personalisation.

[@XX-121]

+Blind Bob - and sounds like you got your rocks off watching the video. must be why there's so much shame and anger in your comment.

[@narcis4none]

One of the best videos of Science Studio KZbin channel Good Work!

[@ecophreak1]

I seem to remember people saying Moore's 'law' was coming to an end over a decade ago, and here we are still going - refinements in manufacturing, engineering and materials have kept extending it, although at some point we will run into serious quantum tunnelling issues I'm sure some engineer will find a way round, or even to utilise the effect.

[@isaackarjala7916]

SteelZenith at some point you will run into a hard stop.

[@isaackarjala7916]

Brad S continuous periodic doubling would require that the number of op/s per cubic meter of processor trend towards infinity. At some point further progress will either be technically impossible or economically unfeasable.

[@digitalnomad9985]

It has been slowing down for a few years now. The miniaturization/speed increase curve is transitioning from geometric to asymptotic as hard limits are approached. That has been the impetus for multiple CPUs, the only way forward is to go wide.

[@michietn5391]

Alternate pathways to explore: transition out of binary into trinary logic, with components that may have 3 (or more: tetranary etc.) stable states. DNA uses 3 base codons to build amino acids. Also look for "spintronics" which detect spin states of bottled atoms.

[@universemaster100]

Being an engineer myself....I love you videos.

[@GregSalazar]

Glad to have ya on board!

[@reisshep]

Same here. Extremely good content.

[@GregSalazar]

I appreciate it, Petar.

[@filipkralj2618]

3. year engineering here, I agree keep posting this it really makes you stand out from other tech youtubers

[@visitthisprofiletodieinsta9173]

Game Changer He videos love you too

[@elimalinsky7069]

7nm chips are set to arrive in 2019 and there are talks of commercial 5nm chips coming up in the early 20s.

[@shinevisionsv]

Nice!

[@cj09beira]

gate all around transistors will also improve perf very soon (higher clocks), ultra violet is entering production soon which basically means we will have a much better resolution when making the dies, (like going from painting monaliza with a roller to painting with a small brush) this will make this smaller process cheaper, (it took time for us to be able to make a light source powerful enough to do this)

[@GregSalazar]

GAAFET looks very interesting!

[@niclasohr]

Love this vid ! :D

[@GregSalazar]

Glad you enjoyed it!

[@jordinio28]

I love this style of video, keep it up mate. Along side the normal PC builds and reviews of course.

[@jackapc4727]

:D I'm earl woops, early*

[@vnyggi621]

Hey earl

[@artkel737]

A technology node doesn't say anything about the dimensions of the gate. When TMSC is talking about 10 nm process they mean that the smallest dimension they can process is 10 nm and it's normally done by some atomic layer deposition and some weired etching techniques. The end of shrinking transistors is at the half wavelength of electrons. That's around 5nm and the effect is called quantum tunneling. But like I said, the gate dimensions are much bigger than the technolgy node name indicates. Current CPU's are not effected by quantum tunneling other parasitc effects are the problem. I'm a German engineer specialised on nanoelectronics. When you want a deeper discussion on that or some papers about this topics feel free to ask me.

[@kyleherbig]

Just read up on FINFET. I came to the conclusion that many in the industry consider the switch from planar mosfets to mult-gate 3D ones like FINFET as like halfing the die size, hence the same as a half-dimension planar design. AND gates used to take two transistors at least, now they take one of the same size.

[@Kappi1997]

specially the quantum computer needs to be cooled down to around -170degrees/0kelvin

[@bulatante870]

Kappi1997 lol

[@Kappi1997]

Unity Member not used to imperial metrics

[@jakobygames]

So basically 09 AMD?

[@Kappi1997]

why should i have a metric unit for everything i use in my job as an electronicengineer so i don't care about imperial

[@TheNuclearPhysician]

Kappi1997 -273°C

[@alexandreesquenet3736]

You can use lower silicon node by using a silicon wire to avoid electron migration (surrounded with the gate).

[@jjaurrgui]

You made me ruin No Nut November ):

[@Carvin0]

There is a difference beyond just marketing hype between 20 nm (using traditional planar fets) and 14 nm (using FinFets). FinFets have a much lower subthreshold leakage (leakage of "off" transistors) and therefore much better static power performance (better battery life). Parasitic capacitances are also less for FinFets, so gate delays are less, increasing speed.

[@fnglert]

"Intel is releasing Cannon Lake here soon based on the 10nm process" - that quote did not age well.

[@dkkempion8744]

Eventually, there won't be any gates at all because processors will work by directly exchanging charges. At that point length will mean far less than density. The more dense the field conduits are the more charges they can exchange between them.

[@stupidburp]

There will be a practical barrier somewhere roughly around 5nm because of cost in addition to the physics, lithography, and material limits. Even using experimental methods of improving yields the costs are already rising rapidly moving from 12 nm to 7nm. 10-12 nm is about the limit of cost effectiveness close to what we are accustomed to. To what degree these increasing costs will passed on to consumers remains to be seen. If some technological improvements are successful on a large scale then this might be pushed down to 7nm. Getting to 5nm is feasible but the costs will be considerably more. Limited runs below that can be done but will probably be impractical on a large scale. There is already significant differentiation between the sizes of features between different manufacturers that are at the same marketed process node. These various features will improve until all reach close to practical limits. Those limits are not far away. We are currently shifting to widespread production at 12nm as a half step improvement over 14nm but the differences in feature sizes between manufacturers in what they call a 12nm process is significant. As we near the end of the road these multiple complex differences will be more relevant than so called die process node size which eventually may not change.

[@1pcfred]

Lithography doesn't work past 24nm. Which is why Moore's Law died then. Since then the industry has just been blowing smoke up everyone's asses.

[@rexlupis]

While there is no reason, currently, for consumer level quantum computing, by the time it reaches the consumer market, I don't doubt there will be a use for it. Consider multiple cores in the early 90ies... No consumer applications were using multiple threads and coding for parallel operations is far more complex than procedural coding. People thought there would never be any use on the consumer level for parallel computing with multiple cores, but ~20 years later it is the norm. We certainly won't see quantum computing at a consumer level in the immediate future, but as the technology, and, more importantly, use cases for the technology expands and improves, a consumer quantum computer will likely become viable economically.

[@petroldevo9934]

Couldn't dry nitrogen be used inside the chip package to stop the tunneling ? Or would this be an expensive direction to go ?

[@Exachad]

Gallium Arscenide, possibly even better semi-conductors and also trinary, hex and Octal. Moore's law might be temporarily dead but once the next new thing comes, it will restart!

[@TheDuckofDoom.]

I have read Mr. Moore's papers, the quote used is a little misleading. His main focus and predictions were not about complexity or speed of a single chip they were economic, specifically the COST per unit of computation for mainstream [high production] bare unpackaged silcon integrated circuits. (In modern terms a single die of a core-i5 rather than the latest release flagship 48 core Xeon platinum complete CPU package and ) The advances so far have been enabled by the economies of scale that come with increasing complexity, speed, and shrinking circuit component sizes; but this technology implication was secondary to the economic intent. One big factor early on was silicon wafer cost, larger chips have exponentially more material cost. Along with the obvious fact that bigger chip contains more material, raw wafers are round and chips are square, and a tiny defect in the raw wafer will trash whatever chip is etched over the location of the defect. So if you average one defect every other wafer and have huge chips that only fit one to a wafer you will toss half your product in the trash, at the same defect rate with 20 chips per wafer you only lose 1 in 40 and in addition you may be able to arrange the grid of rectangles to use more of the circle's area. At the other end the material cost gains from fitting more on a wafer flatten out and a chip with only 20 transistors is going to require the same handling costs as one with 2 billion(up to the point where they are pin bonded and packaged).

[@JonathanS89]

This video was so interesting that I watched it twice :) All of your science videos are great. Thank you.

[@GregSalazar]

Much appreciated, Jonathan.

[@jameswu7850]

But the Moore’s law is ending for traditional silicon semiconductor industry. The 5nm, and other smaller node you mentioned, is either based on silicon nanowire or carbon nanotube, which is inherently incompatible with current planar processes. A few years earlier, the electron beam lithography could easily carve with few nm resolution when DUV lithographer only get ~45nm. Yet today it’s still the DUV lithography we use, electron beam is unfit for a reason, a good one. So are nanowire and nanotube.

[@willimo8565]

Thanks a lot Heisenberg.

[@ejectkid26027]

I was on board with what you were saying, then.... a marshall stack in the background caught my eye! Lol great vid sir, keep it up!

[@john-bs8pt]

Thank you very much sir for your videos! I learn a lot and I appreciate your work!

[@NeilRieck]

Hey, back when I was young (collective eye rolling goes here) we did everything with bipolar junction transistors rather than the FET you just described

[@BrianSu]

Thanks for clarifying this perceived limit. By the way the Apple A11 SoC is already 10 nm. It’s has 6 cores, faster than an Intel Sandy Bridge i5 in both single and multi core performance but hardly produces any heat and consumes so little energy. It’s really mind boggling what 10 nm can do. On that note, will we reach a point where home PCs no longer need to be cooled? I know passive cooling already exists but those are just large heat sinks on steroids rather than a true cooling-less system like on smartphones. On a side note, maybe if Nintendo went with Pascal rather than Maxwell in the Switch then it wouldn’t require a cooling fan. Funny having a cooling fan in a portable gaming console.

[@10THPROPHET]

I like that little bulb lamp to the left of your monitor

[@wheelie.productions4805]

I absolutely love that light bulb lamp on your desk, could you post the link to get it????

[@Graystripe000]

I think there's going to come a point at which the transistors stop getting smaller, where CPUs just get bigger or just internally rearranged with each new generation until we have an entirely different breakthrough somewhere else.

[@PhilippBlum]

I believe we will see more in the direction of RISC. Improving the Instruction Set instead of shrinking the transistors even more. Hightly parallel computing is also a big topic, which we see, especially in servers more and more. 8 cores are totally normal now. There are also more an more 64 cores CPUs. Another topic, which is really interesting, ternary processors. Which makes it possible to use larger numbers. Could be also an improvement. So, transistors can't get smaller. That's totally fine. We have still much other options to improve CPUs.

[@LeoH3L1]

Jumping a gap is not quantum tunnelling, quantum tunnelling is where the electrons wave function causes it to have a high enough chance that it will simply appear on the other side of an insulator between two conductors.

[@unsacredgamer5594]

The biggest Problem about Q bits (Quantum transistors) is that they need to be kept at almost 0 degrees Kelvin to function properly. This is because the quantum effects are so minuscule that any heat will cause interference.

[@ToddMulhern]

Came here for the slick builds, stayed for the science.

[@sidewaysfcs0718]

a single-atom transistor IS a quantum-transistor, but a quantum-computer is defined based on the way information is processed exploiting superpositions of spin states instead of just two states. A single paramagnetic atom can still function as a transistor, you can simplu exploit two states of it's spin alignment with respect to a magnetic field and not bother playing with its superpositions.

[@zwz.zdenek]

Moore's law is more natural than you give it credit for. The DNA's complexity is also growing exponentially.

[@dylan-kt7kd]

zwz • zdenek very insightful

[@MatrixJockey]

GREAT VIDEO...I've been waiting for a computer science video from you.

[@lorenzofourpop]

So if they making smaller transistors to fit more of them in a CPU, why not doing bigger CPUs? Would that be too expensive?

[@jackofallmasterofnone8024]

you are absolutely correct, the smaller your chip is the less money you spend on it's fabrication and thus keep the competitive pricing. A large die size is just not economical for companies. And i'm talking in the scale of Billions of dollars in either saving or profit if they make their chip shrink by mm^2.

[@allmycircuits8850]

So far making smaller transistors allowed to reduce power consumption 'per bit processed' because of smaller capacitances which must be recharged every time state changed from 0 to 1 (and vice-versa), that also allowed higher clock speed because there is just less distance the signal should travel: with big processors you just get various parts out-of-sync!That's why we have multi-core CPUs today instead of one extremely fast core.

[@hrnekbezucha]

And the problem with quantum computers isn't that they're expensive but we don't know how to make them at all. There are researches but it's still more of a nice idea than anything tangible.

[@GregoryCarnegie]

10nm was supposed to be the end, but it wasn't. Engineers will find a way, probably by stacking chips on top of one another. This will usher in a new Moore's law where the separation between layers will be the deciding factor of computational power. But this will only be a prelude to genuinely 3D circuits. Furthermore, there are developing technologies such as pinched hysteresis loop electrical components that may have the ability to supplement/simplify/surpass current designs.

[@JDPorter180]

Cpu and gpu die might start being made a physical part of the actual heat sink in the future to help combat the heat generated by these small, powerful architectures that can't get any smaller but are still gonna be getting boosts from redesigns. Heat generated from these parts wouldn't need a medium to travel through and would be dumped directly into the heatsink allowing for faster more efficient heat removal. The gpu boards in the future also might have pga dies that are gonna be upgradeable to better dies as they are released that are also gonna be made part of the heatsink. This is the way the lithography industry make shift after shrinking transistors down to that final, magical size.

[@Xibyth]

Magnetic flip transistors could potentially reach as small as 2nm. Laser plus magnet flipping the magnetic field and an interposed conduction layer to register the fields orientation.

[@jc1982discovery]

Just found channel by randomly looking at recommended vids.. not my usual haunt but very interesting I haven’t used many pc’s apart from work one which use basically tech from early 2000’s recently as the March of tablets continue. But My understanding was Ram was more important that processor and the speed of hard drive... bought an early ssd which lasted 6 weeks. Enjoyed watching thx.

[@MisantropoFilantropo]

I heard that quantum computing is used for predictions and decryptions since quantum computing works in terms of probability, so it seems that it has no much to do with "normal" computing (everyday tasks, browsing, games, media edition, etc)... but for the time we reach the "0" nm, shouldn't we go for the software optimization instead of more hardware shrinking?

[@ronindebeatrice]

I remember, several years ago, the speculation was that we would never go smaller than 21nm.

[@eliasdargham]

I wonder if we reduce the current passing through the transistors would negate the effect of the the "Quantum Gate". However that would cause a significant power loss.

[@daytonasixty-eight1354]

I remember back in the day being excited about buying a 14 transistor radio because that was a lot.