General statistics
List of Youtube channels
Youtube commenter search
Distinguished comments
About
Anders Juel Jensen
Asianometry
comments
Comments by "Anders Juel Jensen" (@andersjjensen) on "Asianometry" channel.
Previous
2
Next
...
All
Your coverage of the silicon tech-sphere is really damn good! I've gotten SO much more gems here than I've been able to scrounge up any where else combined. Thanks a lot! :D
25
@dokkiro Leaving a critical comment before watching the video is precisely why there are flat-earthers, anti-vaxxers and other kinds of tin foil hats all over the internet. Please don't do it.
24
@clintcowan9424 Moore's law is an observation that turned into a cadence target. Moore actually revised it twice in the early years. And, depending on how exactly you interpret it, it tapered off a decade ago. But it is precisely quantum phenomena that starts screwing things up. Since electrons have this uncanny habit of "teleporting" themselves through barriers classical physics say they can't penetrate, at a certain point the thickness of the oxide insulator, needed for ever smaller gates, becomes a problem. We're already having massive leakage problems. Once we start reaching those thicknesses that is going to get exponentially worse.
24
"I'm not sure what worries more. The fact that we just lost a nuke or the fact that it has happened so often we have an established protocol for it..."
24
You have no idea how happy it makes me that you acknowledge Jon's work! Ever since I found his channel I've been wondering like crazy why it hasn't completely blown up in subscriber count yet.
24
Same. The Danish army insist on them, so after 17 years in the service, and never having one fail on me, it rubbed off on me and it's one of the few areas where I'm actually brand loyal.
23
@arthurswanson3285 It works well for people whith well structured communication... the rambly types aught to shut up and just press "next" on the slide when everyone has raised their hand to signal they're done reading! :P
22
Yet another interesting and informative episode Jon. Thanks! Your cadence/pacing is always on point and your level of high level vs low level detail is just right.
22
The American union story, in general, is a sad protracted saga of continuous absurd self sabotage.
22
Dude is Taiwanese, not Chinese. And he is fair and balanced in his videos on any topic regardless which country he is talking about. Go watch his video on how Japan temporarily got a prominent lead in the semiconductor race.
21
LOL... my brain read that as "Meat grader conference", so I was rather befuddled that it wasn't a butcher show of some sort.
21
17:12 Sorry, but Nvidia does not have any stacked die products yet. They took out a patent on their flavour of it in late 2021, but has not brought it to market yet. HBM is the only industry standard component that uses die stacking, and Nvidia does use those (just the same as AMD and Intel), but they're produced by Micron, SK Hynix or Samsung. The only other product that uses die stacking is the AMD chips that end in "X3D". I felt that was a needed clarification as the wording "die stacked chips like our modern Nvidia AI accelerators" was unclear to the point of being misleading.
20
Excellent structure and pacing to an excellent topic... as always :)
20
3D-stacked variants are already out for Zen 4 as well.
20
Intel shipped the original Pentum with a serious calculation bug. As in, you could trigger it in a simple spread sheet...
19
@10001000101 Room temperature superconductors won't help us unless we also find lower leakage gates and a much faster switching semiconductor. And even then we eventually run into the problem that signals propagate at the speed of light (in that medium), so if we want faster... there is only smaller distances to help us, as we cant speed up light.
19
@brodriguez11000 I have a few minor contributions in the Linux kernel. Hardware is outright hostile, and debugging is frustration as an olympic disciplin. Software, while as error prone as math, has incredibly powerful development and debugging tools. The problem just is that if something is comparatively easy, humans push the envelope until it becomes hard.
18
He didn't "ignore" it... he probably just didn't know, or didn't particularly feel like listing every place where it's popular. Reading the comments it seems to be a bit of a world phenomenon. I'm Danish, and don't even eat noodles unless I'm at an Asian restaurant, and I recognized the brand...
17
@phuang3 It could be done. That is precisely how their 10nm works. But it's a more expensive node today than TSMC N3 is. Intel 10nm uses SO many double and quadruple patterning steps that any gain in not shelling out for EUV machines gets lost many times over in wash and etch steps.
17
The entire point of mechanical watches is exclusivity. It is, strictly speaking, like owning a steam powered vehicle. It makes no technological sense but some will think it's pretty cool. Especially if it's very well made for what it is. So to answer your question: The quality of a (good) mechanical watch is that it is actually very hard to make them keep time to within a second a week, and to make them durable they use (artificial) rubies as bearings, as that's the only solution to not constantly having to relubricate the watch. So you buy something that something that is hard to make for the sake of being hard to make. This is all by design. There is nothing being muddied here. You just happen to find it pointless. If you like the look of a nice watch, go get quartz mechanism that looks nice to you. If doesn't have to be particularly "up market" to last for decades with only a battery every 3-4 years, and a new watch glass once in a while because scratches are a thing.
17
@achangyw The quote given about how quiet the room went when he asked for actual commitment is pretty close to saying "hell naw" in my book :P
16
"The chip hipsters" is the phrase of the week. Thanks for the chuckle! :D
16
@magfal Honestly I think we get our uniforms made in the same factory.
16
Samsung's best high volume node is about on par with Intel's (don't let the marketing names fool you). It's TSMC that's the hard nut to crack for Intel.
16
@0xdeadbabe240 The point being that developing the tools to make DUV lithography is a dead end. Both Canon, Nikon and ASML makes DUV steppers and can deliver on request. Building a factory to build home grown steppers with a sub-par throughput is just bonkers. These are steppers that are going to be in less and less demand. I can certainly get behind the idea of competing for next-gen steppers that will ramp up in production over the next decade... But investing so heavily in something that is on the way out is... weird. Looks more like a pride thing than something with financial thinking behind it. Also I'm fairly certain that the majority of those 350 billion worth of imported chips are made on the advanced nodes. It's not like SMIC is a small fry with insignificant throughput. And both TSMC and GloFo has foundries in China too. So there is plenty of domestic production. There is plenty of opportunity to hold shares in domestic production.... but it's that pride thing, isn't it? It needs to be "purely Chinese", no matter if it makes sense or not, doesn't it?
16
This was a really great analysis. I'm very interested in Taiwan (since I'm an IT Tech and self proclaimed computer nerd since the early 90s) and this perfectly summed up why "made in Taiwan" has always had an entirely different ring to it than "made in China". Not to dunk on Chinese goods in general, but goods from Taiwan has just always been more... well.. competitive.. and somehow expresses a better understanding of the needs of their target audience. Calling modern Taiwanese culture "to some extend a mix of Japanese and Chinese" drove it home for me. Also: Your granddad sounds like he was everything a granddad should be!
16
So my ears didn't deceive me? What do they use cats for in the oil refinery business? Like they used canaries in the coal mines? "If the cat dies you RUN!" :P
16
Because you're a nerd like the rest of us.
15
@Baulder13 I agree! Jon is in a league of his on in this field.
15
@Gameboygenius Team Blue has gotten themselves a CEO who's a silicon designer and not a bean counter. Patrick Gelsinger knows that Intel lives and dies with their ability to fabricate. Personally I think they should've stuck with Bob Swan for another two-three years to make sure they really got down on one knee before they get up again... but alas, it is what it is, and Intel is looking to monopolize the market again... That they'll have the 5200 before TSMC is actually kind of worrisome...
14
China is reaching that because the government pumps in an ungodly amount of money. The same government also gets ripped off by scam after scam in the high tech field... it's hard to figure out whom to trust apparently.
14
Thanks John :) This was really interesting!
14
@ralanham76 I don't have a protocol for what I'll do if my house suddenly explodes. I served 17 years in the army and we didn't have a protocol ALL the food we were provided was suddenly bad. The manual just stipulated that if a food item was bad it was to be replaced with a good one, inventory inspected for further cases and HQ be notified in case it was a wider problem. Nobody considered it plausible that it would ever happen that ALL of it had gone bad. My point is: Protocols are developed when the seemingly unlikely becomes reality. I guarantee you that the first time a nuke was lost everyone went "uh.. NOW what?".
14
As a Dane who's country was completely occupied by German forces very early in the war I'm just incredibly grateful that Denmark, at the time, had little strategic importance. We did have a few stray bombs the British were aiming at the SS head quarters in Copenhagen that unfortunately hit a girls school in the day time, but in the grand scheme of what war time can mean we only got a proverbial scraped knee, where others broke limbs and lost lives.
13
Someone with a bit of status said it wrong and then everyone just went with it. That's why half of all Americans say "nookeelar" instead of nuclear.
13
Ahem... didn't you forget the obligatory "I'll see myself out" at the end of that barage? :P
13
Yup. Jon doesn't "milk the content", which is why his videos are almost always my first pick of the day. He only gets to "yield the throne" for Clickspring and Bad Obsession Motorsport, but since I'm lucky if I get 4-5 episodes a year from them he generally sits quite comfortably on the royal chair :P
13
"This theory takes for granted a force that works instantaneously over infinite distances. Only a madman could ever belive such a thing" -- Sir Issac Newton..... About his own theory of gravity that stood unchallenged for 228 before Einstein published general relativity, which can be reduced to Newtons formular with a "+ G" added at the end. I think that still holds the record for that particular millennium, as far as understating the importance of your own work goes.
12
@goa141no6 I don't know where you get your info but they certainly don't match neither Intel's nor AMDs earnings reports.
12
@Longlius That is a short term benefit but a long term disaster. Chip lithography is a brutal business where you start off making high value products but 6-8 years later you're making crap for toys and appliances. If you do not constantly push for more advanced nodes you're basically eating the company from the inside.
12
The next 30 years is where it gets REALLY interesting :D
12
Yeah. "Fact milking" makes me absolutely furious.
11
When a company buys back it's own stocks the outstanding stocks just become that much more valuable. It's the opposite with stock splits. You own one, the company splits, and now you own two at the same combined value. For some utterly strange reason humans don't want to trade things that costs a few cents, despite being able to trade millions of them at a time. And quite obviously if a single stock costs millions a lot of traders can't participate. So you want your stock price "to look nice" to as wide an audience as possible. Buybacks and splits are mechanisms to do that, but neither alter the actual value of the company.
11
Next up: How AMD is changing TSMC. Since Apple leaves behind massive amounts of almost-cutting-edge capacity each year TSMC would like a costumer who do not just take market share from other ARM players. They want someone who competes in markets TSMCs existing customers are not in.... But TSMCs processes have long been focused solely on density and power efficiency. Not so much raw clock speed. Now where AMD have proven themselves thoroughly and gobbled up nearly all the capacity HiSilicon left behind, TSMC have a vested interest in reducing leakage at higher voltages to accommodate AMDs needs. AMD needs processes that function equally well at the ultra portable level, the high power demand of desktop chips as well as the ever so finely tuned balance between power and performance in data center chips. That N3 does not use GAA may or may not be indicative of this, but critical voices have raised concern that GAA, while producing faster switching transistors, has unsatisfactory leakage levels at higher voltages.
11
Kikilcom 88 TSMC spends around $4.2B per fully equipped leading edge node fab if they have to break ground, and about $3.8B if they just clear out a legacy fab and furnish it with the latest goodies from ASML, Tokyo Electric and Applied Materials. So no matter how you slice it $15B is a healthy sum. TSMC is building 5 fabs in 22 and 23 combined for reference. If the worlds absolute honking giant expects to be able to keep up with demand with 5 new fabs over two years then the almost 4 fabs you can get for $15B isn't just "a peanut in a zoo". That's a sizeable start.
11
Apparently cats are an integral part of oil refinery. They show up many weird places. In medical, for instance, they apparently use cats for scanners. No idea how that works either....
11
I expected him to go make some kind of statement about how a bomb differs from a cow. Now people might confuse the two...
11
@Syomiti Nope. It's similar in my own language as well though. Danes pronounce the candy bar brand Snickers as "sneakers". Most people here are fluent and literary in English, yet this cultural little gem (or turd if you will) remains. The reason I know about the theory behind it is because a psychologist gave a lecture on it. They had made a fake scenario where people thought they were at a job interview and "the important person" pronounced something wrong... and more than 75% just started pronouncing it that way for the duration of the interview.
11
@phuang3 They aimed for 2.7x but reached 2.3x. Since Intel does foundry service now so the density of their 10nm is publicly known. They have an ungodly amount of capacity for it, that they're moving off themselves now, and still no customers because it is so expensive. It was a recurring topic at yesterdays investor call......
11
STANDING OVATION! Absolutely stellar episode! And 10 thumbs up for the use of applicable meme material! :D
11
Previous
2
Next
...
All