No video
Breakthrough in Chip Manufacturing - x40 times faster 🔥
In this Video I discuss New Breakthrough in Chip Manufacturing which speeds up Computational Lithography by 40 times!
#TSMC #ASML #breakthrough
Support me on Patreon: / anastasiintech
Пікірлер: 485
Let me know what you think!
@firemedic5365
Жыл бұрын
In a past video you stated that the smaller the chip, the more problems arise due to them being more delicate. The wires connecting them must be correspondingly small, and are therefore destroyed by heat at a much faster pace. What is being done to address this problem? I need a computer that is more robust, not less so.
@og3suso978
Жыл бұрын
Since a german mathematics discovered a method of compressing data without loosing any called fractal compressing and he died from a very suspect case and his invention got banned for ever, i believe absolutely 0 of what modern industry try to sell my. Its all useless garbage anyway to collect as much money from you as possible for absolutely nothing you need....
@smarthalayla6397
Жыл бұрын
The disk you hold in your hand is the biggest chip I ever saw... Still, with no "real life" performance tests this chip is worth nothing and that because many times companies in many areas comes out with declarations and they do not deliver. I'm still looking for the 4k global shutter camera that costs less then $1000 when different companies declared so in the last decade.
@aprilgeneric8027
Жыл бұрын
how did a start up who had this idea get beat to the punch by a big corporation? did they tell nvidia about their idea when asking for GPU's or GPU technical advice for implementation of their idea and NVIDIA stole it?
@aprilgeneric8027
Жыл бұрын
@@smarthalayla6397 that's not just one chip. that is about 100 chips back when Pentiums were new
I worked on the worlds first EUV Reticle Inspection Microscope and it pretty much killed the company I worked for. Much like the computational techniques you describe here, we were attempting to image the effects of EUV masks rather than image the masks themselves, and it was almost impossible on the budget we were given. We didn't have a bright enough 13nm source, we were running right up against the limits of the optical, vacuum and stability systems which could be manufactured at the time, and we were attempting to characterise process windows smaller than anyone had ever attempted before. We had a fraction of the budget that industry leaders like ASML would have expected, for the same research and it's amazing to think of many times our budget being spent on supercomputers to do what we were attempting, using computational techniques rather than using photons. This has provided a fascinating insight into the current state of the industry. Thanks,
Thank you Anastasi. Sometimes the info from various sources are confusing. Its a blessing to have a person that reads the papers and post the pure synopsis output
@questmarq7901
Жыл бұрын
Are you from Greece btw? you look Greek
@ocoolwow
Жыл бұрын
@QuestMarq so basically you are admitting you are too dumb to read articles written at an average of a 5th grade level? That's sad bro I wouldn't openly admit to that.
I am a retired Electrical Engineer and I really appreciate your videos that keep me up to date on various technologies. Keep up your fantastic videos.
@autohmae
Жыл бұрын
I recommend: asianometry on KZread as well 🙂
@davidrandell2224
Жыл бұрын
“The Final Theory: Rethinking Our Scientific Legacy “, Mark McCutcheon. Electrical engineer.
@thewishmastur
Жыл бұрын
she is AI
Yet again another insightful and interesting look into the blindingly fast advances taking place in this realm of tech. Thanks again for bringing it to us. Get ready for the 100K party.
@AnastasiInTech
Жыл бұрын
Thank you 😅
Awesome videos!! Turning 60 tomorrow. Still super keen on keeping up with all this cool tech 😎
@MrChiangching
Жыл бұрын
You're still a kid!😅
@AndOgre
Жыл бұрын
Just turned 44 too. :)
@Knowyourbody
Жыл бұрын
Happy belated!! Wish you long and prosper life 😊!
This is awesome, since the presentation at GTC I was shocked, basically designing noise to create perfect patterns, but with you Miss Engineer, I always learn more, I didn't know there was another company that tried the same thing much earlier and that Nvidia beat them by getting to market first. Thanks for the video.
I remember many years ago when photo lithography was hitting a wall and everyone in that industry was worried on how to resolve smaller features. Many thought it couldn't be done using conventional optical equipment yet they keep managing to come up with clever tricks to get around the problems. I see that innovation is still alive!
@bighands69
Жыл бұрын
Manufacturing transistor design is going to need a change. Most people seem to actually believe that transistors are actually at 2nm in scale when in fact they are about 18000 nm square. And are about 90000nm cubic.
@TonkarzOfSolSystem
Жыл бұрын
They've blown way past using conventional optical equipment. I'm not sure exactly when but 15 years ago they weren't.
@soylentgreenb
Жыл бұрын
One of the most clever ways is to simply lie and call a node ”3 nm” when there is no measurement anywhere even close to 3 nm.
@MaxWindshear
Жыл бұрын
@@bighands69 I believe that measurement is the thickness or width of the transistors gate. Where did you get your measurements?
@lucasRem-ku6eb
Жыл бұрын
The wavelength of electron beam. that was always the issue. We invested in EUV, and now xEUV replaced that, we always kept innovating ! The mask is not needed ? How you think we do that ? CRT TV skills needed, not an electron beam ! Mores low, did you remembered it correctly ?
This breakthrough couldn't have come at a better time ! pretty awesome video ! 💯
@TheBlackManMythLegend
Жыл бұрын
exactly.
Nobody else is talking about this, which is insane. This is a big deal. You deserve many more subscribers. I hope you get there. You deserve it.
@kazedcat
Жыл бұрын
It is not really that much of a big deal. It will speed up prototyping but will not affect fully ramp up production. Once you have your mask design you can use it to print over and over again until the microchips end of life. So with this technology you can cut down a 6 month bring up into 5 month and 3 weeks.
@user-wq9mw2xz3j
Жыл бұрын
It's big, but not as much as it sounds and you might think. No reason to hype it up. It helps, nothing revolutionary in itself
@Wr41thgu4rd
Жыл бұрын
@@kazedcat agreed it is an overblown nothing burger, a 40x in chip performance speed would be different.
@Muiuiuky
Жыл бұрын
The news is a couple weeks old, so most media outlets already reported it.. that's why nobody else is talking about this, they already did
@emilen2
Жыл бұрын
I'm sure Asianometry has covered it.
Such a wonderful and straightforward explanation of this significant breakthrough!
I really enjoy your videos, thank you for taking the time to make them for us :)
@AnastasiInTech
Жыл бұрын
Thank you
Your gaze is like a quantum qubit, always in a state of superposition, and my love for you is like a quantum algorithm, always present and never resolved in its complexity.
@michaeljorgensen790
Жыл бұрын
Keep your quantum qubit in your pants.
That is amazing! Everything seems to be accelerating. I am also looking forward to the 100K party! 🎉
Excellent information. As a microscopy researcher I endeavour to do the inverse - i.e. use combined optical and mathematical modelling of diffraction effects to be able to 'see' nm structures in my specimens using visible light (without fluorescence). This sort of thing is a lot more interesting to me than the latest advert/investor prospectus 'research paper' from the competing generative AI modelling companies. Thanks for making this video.
@questmarq7901
Жыл бұрын
I love the scientific method. One branch of science is levitating another. This is a prime example
@beautifulsmall
Жыл бұрын
The Puma 3D printed research microscope looks excellent after a quick look at your channel.
Thanks for the explanation! You provided really great details about how ASMR Lytho works.
I’ve just found your channel I happily say. Love your accent & your joy at what you’re saying adding a “wow” I’ve been going through your videos enjoying everyone. I need to stop as I’ll run out of your videos to watch. Keep up the good work stick to it you have a talent you truly do, which is refreshing to see in someone so young & the field in which you speak about. I wish you all the best (was gonna put luck but you don’t need luck you have talent!). Stafe Marc xxx
Thank you once again for updating us on the newest tech news and breaktroughs!! 1 Love!
Ur channel is so amazing. I’ve not found one like it at all. I’m sure you’ll blow up! Sending love, much thanks!
@AnastasiInTech
Жыл бұрын
Thank you so much!
🎉🎉 this is very exciting. First of all your channel is awesome. And the speed with which you share information is so amazing. You are one of my top 10 resources to keep up with the breakthroughs in field of Chip manufacturing and AI. Keep up the good work
@AnastasiInTech
Жыл бұрын
Thank you 😊
Great video as always I like to see more videos on chip manufacturing.
(We need to get Anastasi over 100K subs.) Seriously, increasingly, I only have to watch this channel to stay current at the level of detail I need to for my work (largely in economics and geostrategy). Partly because Ana doesn't 'color, load, frame, or propagandize' the material. And partly because she knows what she's talking about (which isn't true for a lot of other providers.) Then I just search for whatever she mentions that I need more information on.
Everytime you say "another breakthrough", I fall in love.
I love listening to your videos hearing you giggle occasionally tough information made fun & light hearted approach love the kind vibes yay kindness
My favorite technology news channel. Thank you for breaking down these wonders for us layman out here.
Super interesting, super well explained & presented. Fascinating to see how a bottle neck is engineered away by using graphical processing rather than cpu, creating such huge speed of production gains & much less power consumption: Win Win! Thanks for sharing!
you always keep me up to date on the latest great news in tech, and of course, your trademark giggle of approval. thank you again!
Your tech fascination is a bliss and contageious, and your work put into your vids is much appreciated 🙏☀️🌱
Thanks for the video! I'm currently choosing between electrical and electronics engineering at college, and these videos are giving me the motivation to follow my passion in digital electronics/ computers despite living around electrical-heavy industries...
Excellent to know that there still is space for less space and that we slowly become faster. That said, a big thank you for your understandable recaps and explanations!
Can't wait to see the results! I LOVE all this
I'm glad someone understands this wizardry. Thanks for explaining it to the rest of us.
It will become to the point, nanites will be manufacturers to add more transistors to the surface at a smaller scale. Excellent video. May want to podcast the future of WiFi 7 and mesh cellular.
i just realized i wasn't subscribed to your channel ! love the content of your videos. keep going !
Thank you so much for taking the time to explain these complex things to us 🥰
awesome stuff as always ana =] thnx for sharing the news with us ; it is certainly an exciting time to be alive
this video was one of the most interesting tech vids ive seen in a long time. watched it several times, so much good info.
@AnastasiInTech
Жыл бұрын
Thanks!
Si tous les vulgarisateurs pouvaient être aussi clairs et posséder cette pointe d'humour et de finesse cela rendrait la technologie et la science beaucoup plus sympathique. Merci pour vos vidéos qui sont de très bonne qualité. Thanks.
Well done Anastasi, many thanks.
your videos are brilliant, dont stop anytime soon
Thanks and great to know this. I think it can also be used for computing other physics simulation problem, such as designing airplane wings or rocket controls.
Sounds like magic. Thanks for giving an insight of what's happening. Very well explained!
Very interesting and informative video, thanks 🙏🏼
Excellent! Thanks for sharing and the very best of luck!
I'm not into hardware engineering like this, thanks for these lessons and insight.
In answer to your question I always learn something new from your videos thank you.
You are awesome Anastasi
Thanks for the interesting video. It's amazing how they can make things so small and so fast. Cheers from Canada :-)
Amazing! What a wonderful quick overview of an incredibly complex process. 👍👍👍
That's great! I love your videos. You can translate very complex subjects in simple and cristal clear language. Ever with a nice smile and the sweetest voice among all youtubers!
Thank you, very interesting! I appreciate particulars like Maxwell equations applied to masks and litography. Would you go deeper in tech details about chip manufacturing? Video series with several chapters would be cool. Thanks!
This should be headline news! Massive breakthrough, thanks for covering this for us!
Incredibly fascinating how far they've come since I finished my studies in December 2001! 🤩
Super cool I love these machines they are the pinnacle of human engineering
Very interesting video and good explanation. Thanks for sharing it with us.
Thank you! Looking forward to the future!
Thank you, Anastasi. Valuable technology to accelerate the transition to a prosperity economy.
I always get to know the technological advances from you. You are amazing!
@AnastasiInTech
Жыл бұрын
Thank you
I worked for ASML in Austin, TX from 1998 to 2010. ASML is an awesome company and It is good to see that they are still at the forefront of photolithography.
Fantastic ! Seriously good video !
@dchdch8290
Жыл бұрын
And it won’t stop
Love the videos love the accent!
Anastasi you have an amazing voice and I love your laugh. Your vids are very informative about chip manufacturing and a.i. keep them coming.❤
@therealavolpe
Жыл бұрын
She can hardly speak English
Kind of a head scratcher that they didn't use GPUs for this before now. It seems like an obvious application. I guess it was a case of they were so used to the old way of doing it they couldn't think outside the box, or maybe the thought of re-writing all their software was too overwhelming.
@AnastasiInTech
Жыл бұрын
The same here I was surprised
Thank you for your content!
Absolutely incredible. Had no idea that mega data centers were needed to solve Maxwell Equations just to create the photo masks.
Wafer fab has come a long way from the days of simply printing a large mask and using lenses to reduce to the wafer size. It would be a dense episode, but I would like to see a general overview of the process of creating a mask nowadays. Is the large amount of computation to compensate for the image distortion of the optics? I get the need for multiple masks because of the many layers of the chips. Has the entire process evolved to the point where the source of the UV light is a version of an LCD screen where there is no actual mask involved?
@markbooth3066
Жыл бұрын
Once upon a time, it was a mask per lithographic layer, but every mask has flaws, and as the features got smaller, each flaw had a more pronounced effect. Reworking masks to remove these flaws was incredibly expensive and time consuming (imagine using a scanning tunnelling electron microscope to move atoms around and you probably won't be far off at current process nodes). So, lithographers started imaging the same layer multiple times, with two or more masks (double and multi-patterning), so that the flaws 'averaged out'. Back when I was in the industry all anyone ever seem to talk about was LER (Line edge roughness) and this was the solution they had to endure, even when teh industry went from 193nm to 13.5nm. It also explains why modern chips require so many masks. It's not just the layers, it's the multi-patterning per layer. I suspect that the computational techniques used now are about calculating the total lithographic effects of combining multiple masks, and possibly even the whole stack of masks, to minimise the process window for every one of the billions of features, taking into account the flaws in every mask simultaneously. It's pretty damn impressive whichever way you look at it.
Amazing results, I can see some other companys racing to compete with this LSI on Wafers
Oh cool :) Nice to see that huge advances are still possible here
Thanks very informative and well done
I learned something, maybe, around 4:45, when the guy told about arranging the pattern on the mask (reticle) to make the pattern you want on the wafer. I am a former IC engineer. All I learned from Anastasia was that Maxwell's equations were involved. And that she thinks she is cute when she giggles, with which I agree, but I didn't come here to see a pretty woman giggle. I thought one needed to compute diffraction of light around a barrier, and how it interacts with the other light coming off all those other barriers. It is amazing that one may use a light source of 300 nm to make a feature of 7 nm, but I didn't learn about it here. And those dizzy-pans through the server racks were less than useless.
When I was nine years old I found an adding machine, a manual, handle turning machine that I thought was wonderful.
There are still many steps to improve and innovate. I approve.
NVidia GTC Conference 20 thru 23 2023 was fun thank you for the heads up Ana. I Learned something about what this Conference highlights were.
Very good and informative presentation, Thank you
acceleration continues
Thank you Anastasi. Groet, Arjan
Loving you and your presentation
@Anastasi in Tech - I have to admit, you could probably discuss the finer nuances of painting a toilet wall and still come across as thoroughly entertaining. 😊
Great video! I wasn't aware that state-of-the-art lithography was currently computationally restricted but you explained the reasons for it nicely. PS. I think you should try applying de-esser to your audio processing chain. It sounds to my ears that you have too strong sibilants in the recorded audio. This typically results from combination of your speaking style, room echo and mic positioning.
This is awesome. It gets me thinking, as someone who's studied physics, though. Essentially an integral can always be brut forced with compute, this technique being an improvement on that. But most of the mathematical calulations for exact results rely on symmetry arguments, without which the calculations were not possible by hand. Like calculating what the case and result would be for a sphere in a way that generalises to the universal solution. What if we trained large language models on a bunch of key mathematical loop holes, so to speak, and see if it can go on to find exact universal solutions for problems we currently brut force with compute?
@the.other.ian.
Жыл бұрын
Considering that GPT-4 as-is is already proving capable of performing decently on some relatively high level math tasks, and can improve with self reflection learning, it might be possible soon without training a unique LLM for the task specifically. But yeah I was thinking the same thing. When I read up on the recent research on GPT-4 the other day it struck me that think it won't be particularly long before we start seeing significant human-understandable math proofs of new and old problems achieved using LLMs. Machine learning has been used in the past to make a proof for something but the proof was so insanely long that it couldn't possibly be checked by humans. I can actually see LLMs being used collaboratively with humans in this regard when it comes to advancing math in new directions (not just finding proofs to theorems and stuff) - they will be able to riff off of human mathematicians' insight and proposals presented as input. Kind of like a synergistic brainstorm aid.
I'm an elderly Canadian man born the same year that the invention of the point-contact transistor was announced by Bell Labs. As a pre-teen boy I brought discarded 1930s-vintage vacuum-tube radios home from the town garbage dump {later called a "landfill"} & got them working again. That was the start of my electronics career. Today I watch KZread videos about 2 nm features on huge wafers that will be cut up into individual computer chips each containing not millions but billions of MOS transistors. Our species is very inventive... Now waiting for Elon Musk's "Spacex" to begin launching giant rockets to Mars ferrying people to that planet to create human colonies. What next? We live in very interesting times!
wow wow wow - amazing!
It might have been worth mentioning that the computed mask must interact with the light beam and cause interference pattern that "happens" to result in correct details on the chip. And computing the required mask pattern that results in correct interference pattern is the hard part. (And the reason for doing it this way is that wave length is already too long to do this with regular masks.)
Nice explanations of this interesting thecnics
❤❤❤❤ люблю ваше содержание и люблю вас.😘😘😘😘🌹🌹🌹🌹🌹🌹
I love her “GIGGLE” . . It ‘cements’ the issue the intellectually because she makes my brain “LAUGH” too Ha Ha !
Thank you!
That can come in handy for all the AI accelerators that are going to be made the next few years.
Excellent! Thanks so much, Anastasi.
@AnastasiInTech
Жыл бұрын
Thank you !
@RalphDratman
Жыл бұрын
@@AnastasiInTech I did not know we now have 2 nm transistors! What is the memory cell size associated with that figure?
Am I the only one that likes the way she talks? I find it very calming
"I used the GPUs to create more GPUs" -Jensen Huang.
Wow!! Cool news. Heard it here first.
Thanks for the video!
I love it , when u make hardware related videos ☺️
You don't want to know what I'm thinking right now. Wow! We're witnessing the dawn of computer & AI designed chips, progress is accelerating us into a bold new future. I hope we can adapt without too much pain.
Awesome video. Thank you.
4:12 glad you did it for me
In 15 years we’ll have comercial viable robots
I hope this reduces the cost of masks sets. Deep sub-micron process nodes have such high NRE, they can only be afforded by companies guaranteed a very large return on their investment, either through having very high value chips (meaning not suitable for most consumer electronics), or having extremely high volumes. As far as I know, there is no one imaging actual 2nm features yet. The 2nm designation refers to the equivalent gate length that a planar transistor would need to have to achieve the same performance. And even modern planar gates rely on lateral diffusion to achieve their final gate length using larger gates that mask the source/drain diffusions (this is a self-aligned transistor).
I had a computer game made by synopsys. It was made for ATARI 65XE and the title of the game is "DIMENSION X"