so basically he is saying that it does not matter because AMD is doing it, but let Intel do it and he will do a party celebrating how Intel is amazing. This guy is biased and he is paid by intel.

I like that the ad before this video was an Intel ad saying what's inside matters.

Linus is the kid that actually has his presentation prepared.

Mentioning carbon nano tubes without Riley around just doesn't feel right.

Intel says node size isnt everything.... I imagine it would be very important if they had 7nm and AMD had only 10nm.

“Smaller one is better”

When are we going to start measuring performance by how much RGB we have in the system?

I've worked on one of the 7nm nodes while it was in development. Fins were 8nm wide, fin pitch was 26nm, diffusion/poly layers were over 20x50nm and if transistors were not in an array they had to skip at least 3-4 fins. I think the only thing at 7nm was some spacing between layers...

"we asked Intel if lower nm matters" Then we asked Chrystler if we should buy a Tesla

Intel: Don't talk about benchmarks, transistor size doesn't matter, Apple still loves me... I need help!

If Intel had 7nm and AMD 10nm they'd be screaming out loud how important it's the fabrication node, as they don't they compare a 90watts gpu to a 65watts GPU and say it's a cpu comparation

Hmmm the words used exactly match Intel's marketing slides with "real world performance", "applications you actually use"

correction at 3:32 increasing surface area of the gate around the channel in FinFET was not to allow more electrons to pass through the channel but to combat short-channel effects and carrier leakage i.e. clamping down on the channel to turn it off when you want the device off. Without FinFET, at these sizes, the field effect would not be able to be controlled by the gate and would be Always-On instead of switching (on/off).

"Does transistor size matter?" *talks to intel about it* HMMMMM....

Linus: To learn more about process node size we reached out to intel... Intel: 14NM+++++ MORE PLUSES IS MORE PERFORMANCE GUYS.

The whole nm measurement is a hot mess between Samsung, TSMC, and Intel. One will claim 5nm, another 7nm +, and another 10nm etc. As so correctly stated, you just have to go with performance and architecture as ultimately that is what it is all about.

I can't stress enough on how important to me this video was on explaining to me the number of nanometers I am always hearing or seeing on looking up a cpu. Thank you so much!

This video feels like intel's last solution because they can't get into 7nm

Drinking game: Take a shot everytime Linus says "seven".

Ok but missed opportunity to have Riley come and say carbon nanotubes at the end

Uhm... Could you add TSMC engineer instead of intel?

Whatever happens we get a FPS boost in gaming and that’s what I need.

Alternate timeline: Intel sponsored this video to explain why its smaller node is EVERYTHING.

Six and a half minutes of damage control.

missed opportunity to have Riley deliver the "CARBON NANOTUBES!!!!" line.

Of course Intel would say it doesn't matter. They can't keep up with AMD and TSMC. 😂🤣😂🤣

Well, if you go and ask Intel if the number of nano meters matter, of course they'll say no.

This was amazingly useful, thank you very much for the invaluable info. Much appreciated.

I really enjoy the deep dives into tech like this!

Back in my day, CPUs had just 5 transistors and that's the way we liked it.

3:08 - If we're living in a simulation, then quantum tunneling is a clipping error.

thanks much for these informative vids!

Really, enjoyed this deeper dive.

5:30 _"Just remember that the processor node isn't the be all and end all when you're shopping for a CPU anyway"_ Oh and real world performance, but at what cost in efficiency? Check your process node for answers.

Gotta love subtle intel promotion with their "real world performance" bull...

This should've been put on "$h!t Manufacturers Say" instead of techquickie edit: this was intended to be a joke, the main point of this video is just "processor architecture (XX nm) are defined differently among manufacturers thus you might not compare the figures or the numbers side by side"

Great video, thanks a lot!

can you do a techquickie about AMD/Intel vs ARM and etc processors

I'm disappointed that you didn't added Riley when saying " carbon nanotubes" 😂

This is mindblowing, thank you so much!

Video idea for you, Ram timings - how and why they matter. keep up the great work, love your content

'Carbon Nanoooo Tuuuuuuuuube" - Said Riley

I would have liked a comparison of how TSMC and Intel count a new process

Have you done a video explaining all the M.2 variations? I get very confused when trying to get compatible parts from the limited information in motherboard listings!

Just saw your rant on WAN, wanted to jump in here and thank you for the deeper dive. I very much appreciated the info, and perspective you offered. The idea you got flack for it is surprising to me. Gotta hate that internet toxicity. Thanks again, more please!

A very "tactful" footnote about "real world" benchmarking Linus. Gotta keep 'em happy I know. 😉

You need to make more scrapyard wars! As well as more super cool factory tours. I loved your tour of the fission reactor video.

That ear and nose hair trimmer sponsership came very surprising after such an interesting professional high tech video :D

Great video, I liked the performance matters the most part at the end.

Moore Transistors is very Noyce.... I'll see myself out. Moore's Law has actually been considered dead for a while in the Computer Engineering community. We need to move on from planer 2D lithography and focus on monolithic 3D lithography with better thermals even at a higher die size. EUV is also prohibitively expensive for any manufacturer that isn't TSMC- or GlobalFoundaries-level. CNTFETs are pretty great though, but we need to have better package cooling especially with hybrid systems within packages. Kaby Lake G had an AMD graphics die within the "CPU" packaging connected with interconnects to the actual CPU die. I can't wait until ReRAM gets placed in an M3D configuration (as in ILV not TSV) on HPC or consumer chips. GAAFETs are also definitely a game-changer.

4:36 10nm ~mid 2017 He he he

Thanks...nice recap.

0:18 subtitle "And why node". Didn't think we would notice, but we did

I'd love to see 1nm processors in coming years. It would be damn interesting

Everyone 2020 cannot get any worse! 2020: Linus saying apple rocks

"that's a lot of nose hair to trim" I feel attacked.

when learning about how they made processors pre-3d i always wondered why they couldnt go 3d and "stand that on end" or "stand that on edge" "lay that on its side" after talking to a friends uncle whos a samsung eng, they had the same thought and, had been talking about it at the time, mostly as a way to boost both ram and nand memory capacities on the same size die, or make the same capacities smaller in the case of microsd type chips... he actually took some ideas he got from talking with us back then and apparently used them, not sure whos, like 8 of us came over to play DND and watch movies for the weekend, and he decided to join us rather then hang with the other relations that had come into town for their family reunion, the uncle spent 1/2 his time in the US and 1/2 in korea/tw/etc even going directly to the lithography plants and looking into issues/bugs and, how they could be fixed, since sometimes, they have to tweak the design from what "should work" to what will work with the lith tech and how it behaves in actual use... one reason he was so successful i think, was, the man never assumed anything would work as expected, and has always been willing to try something, even if it seems stupid "This cant work"... more then once, over the years he came back with "that shouldnt work, but, it does, now we just need to figure out why it works this way, and, what that means for future designs, and, if theres some way we can exploit or improve on this..." he was building a, and i kid not, electric-steam powered remote control plane, the batteries or capacitor was used to eat a closed loop steam engine he had built in his spare time, by built.. he machined the blocks and built over a dozen prototypes out of some high grade heat resistant plastics, before flushing the design out and making it out of aluminum, it was actually amazingly functional.. the only real issue was.. with the tech of the day, he had a hell of a time getting the temp control and throttle setup to work quite as he desired.. (im told now that hes semi-retired, he got that sorted and, the things got an old smartphone in it as a "Brain" that can report back all sorts of shit depending on what hes added, it can also fly for longer then the vast majority of remote planes.. im told, hes now talking about a design idea one of his old work buddies came up with, thats sort of like a modern drone... but, using mini turbines, that are actually powered by steam, rather then electrically driven.. the whole thing came about when one of his friends won an auction for the contents of several storage lockers, and found an old steam car in it, in perfect condition, i mean, the thing just needed somebody fimiliar with the tech to give it a once over, add some lube a few places, filled with water and fuel, and no joke took them 45min to get it working and have a cop offer to escorte them to the place a few blocks away that could get it both appraised officially and get the thing setup as street legal and the title issues taken care of.. turns out the old fellow who owned it, was still alive, but in bad physical shape, and had just let the storage units laps because he was never going to be able to do what he planned with any of it.. apparently, they befriended him and... well he lived longer then the doctors expected, and, got to ride in the steam car again.. any time he asked... im told though that a few of them started getting ideas after looking at how the steam engine worked... an electric-steam hybrid could actually, im told, possibly could be more efficient and powerful then current electric or combustion engines, and, in some ways simpler then those techs as well.. since, in many cases a very simple closed loop near zero loss system for engine, with electric used to heat the liquid, could genuinely give more power for the same or less energy used, and, a simple steam engine...isnt all that complex... and if you are mechanically minded, you can watch how it functions and figure out what the majority of the parts are doing just from deduction alone, but, another advantage im told, they could actually use hybrid metal/plastic blocks and such with current plastics/carbon-fiber/etc type materials, there are a good number of ways they could build a block that was light and durible, able to take the pressure thats required to operate the unit and, possible as much pressure as the system can produce before other components will fail.. (emergency pressure valve to discharge if it hits a pressure above what its rated for/set at. anyway.. the idea is neat.. and so is the idea of 3d transistors, carbon nanotubes, graphine..... they have stuck with silicon as the base in no smallpart because its duriable and a substance thats well understood, other materials they have tested that give superior results have serious flaws im told, one being, they can be quite fragile and even the pressure of an ihs under a cooler could damage them, the flex of bga packages expanding/contracting from heat... etc... the most promising seems to be silicon and carbon nanotube...

I think this is the earliest I've ever been

BHuahahahaha, No thanks, i'll stick to my 7nm AMD that sips power

Great video!!!

Excellent video. More microelectronics topics please.

intel: we're not on the smallest advertised process node also intel: process node numbers done matter everyone else: you sure, as AMD are kicking your a$$ on a lower process node.

"We reached out to a company using a 7 year old process"

I expected Riley to scream when Linus mentioned carbon nanotubes

Can you please make a video about future CPUs? And about how the carbon nano tubs would work ? Thanks!

PC enthusiasts: Less nanometres is good! Intel, struggling to go below the 10nm threshold: Nah, it's just marketing BS, size does not matter THAT much!

Hmmm...when I read "Is CPU Marketing a LIE?" my first thought was how they like to use the boost speed to market their CPU's EG: 5.3GHz in the marketing trying to make people think it is actually gonna hit those speeds all of the time when in fact it might hit it not often at all and if it does only on 1 core and for about 1 millisecond. Oh another lie TDP numbers or at least how they are measured and actually mean nothing now days because the chip might only actually run at those TDP's under very very light loads and only if it stays at base clock rate. Everyone that makes CPU's now days should be forced to market their chips with full load TDP numbers since that is what the chips will be runnning at most of the time when actually being used by the consumer. This not only tells everyone what to expect from a CPU it also informs them just how much cooling they are gonna need to keep things all nice and cool. Instead we got some piece about talking about nodes. Yes we know different companies measure their nodes differently/ Most that can read have known this for 20 years now.

Hopefully this doesn’t sound dumb, but how does a computer shut down, and turn back on when restarting. If it is off, how does it know to wait a couple seconds and turn back on?

Thanks for finally cleaning up some loose bits of missing data.

How can someone dislike the video that was uploaded 1 minute ago yet the video is 6 minutes and thirty-seven seconds

We all know who really should’ve cried “CaRbOn NaNoTuBeS” in the video

Love the video, I learned about this around two months ago, and I wondered how come you never said anything about it.

Caption says "And why node?" instead of "why not" XD love it

“CPU node size doesn’t matter” “This video sponsored by Intel” Lol

lame attempt by Intel to divert attention from their competition, "the size isn't all the story hey look what intel has done..."

Thank you. Love the science videos

Also, how the transistors get used matters. An ARM gets more done per transistor than an X86 because the simpler instruction set means there are fewer transistors set aside for things that are rarely done in real software. All processors theses days have a "pipeline" with multiple instructions working their way through it. If another stage of pipeline is added a lot of extra work elsewhere in the chip also needs to happen for things like branch prediction. Deeper pipelining helps to let you move to a higher clock but comes at a cost.

"we reached out to people who work at intel for information on process nodes, when they've been on the same one for 6 years at this point. Surely this won't lead to biased answers saying that anything smaller than 14nm is a waste of time"

This video pop up after Intel Marketing launch their New Real World Performance benchmarks, Coincidence much ?

Thanks for the great vids. When will you do one on Koomey's law and other computer "laws"?

I am using this as a source for a college paper and I hoped that you would know that Linus thank you for explaining this to me because I had it super wrong you saved my butt.

Linus legit looks more like Luke everyday

You know, the "Intel inside" sticker, isn't that important anymore... We need new ones, like "AMD inside" and "ARM inside". 😉

excellent wording and video overall ;)

The process node size is one of the most important thing if you are shopping for some CPU's (FPGA's) as it may increase the speed with lower power without optimizing the HDL-code. (or you may add another pipeline with all the syncronisation problems, switch to parallel code, ...) As the process node size is no more the transistor size: Can I still use the old formula for calculating the maximum clock rate (on paper) or should I better use a computer simulation (restraints calculation)?

The title should have been "CPU process node as fast as possible", I feel like that question in that title hasn't been answered.

Intel: using nm cannot be measure number of transistors today. instead, we invented "+" measurement.

I really wanna see TECH T&A... make it happen!

When will we get RGB transistors in cpu?

I see Intel’s decided to start sending checks again.

Asking intel about node size when they are stuck at one size.

Watching this again after LMG Clip on today's Intel long term plans.

This was interesting thank you

You do such a busy schedule just to give us the content we subscribed for I have been your sub since 2016 and the journey from then to now, had been a great time. I really appreciate the work from second most famous Linus in world 😉💚 And make every channel intro RTX ON

yeaaaaaaaaaaaaaaaaaaaaaah

I’m not sure if you have done this video before, but it would be interesting to know what magnets do to electronics, it was always a big no no to have magnets next to your electronics but now it seems to be normal to have magnet phone cases and other bits

Perfect video !!

gotta love how the comments manages to devolve the topic to a fanboy dispute

Well explained.

Linus Media Group has filled the early 00’s TechTV hole in my life. Thanks dude