Stacking Dies on Glass Panels

[@Cs13762]

"we can already make perfect silicon substrates, and the cost and manufacturing ability is practically unbelievable, but what if we could make it even cheaper by switching to literal sand?"

[@kevinnivek8907]

we just need intelligent sentient sand…..

[@BastyTHz]

the thing is huang did absurd thing like put 2 600w chip on a single package without LN2 thermal solution

[@Nib_Nob-t7x]

⁠​⁠@@BastyTHzln2 is never needed as a thermal solution outside extreme overclocking. Multiple kilowatts can be cooled with just water and in die watercooling.

[@squirlmy]

silicon is no less "sand" than glass is. Silicon for chips is highly purified sand. Glass is made of not so pure sand, and technically it can be described as either a highly viscous liquid or an amorphous solid. Silicon crystal is a solid, but a metalloid -halfway between a metal and non-metal.

[@pudy2487]

@@squirlmy Sand and glass are specifically silicon dioxide. Chips use elemental silicon, for the most part. They are different chemical compositions.

[@TardMan1]

These videos are always so unbelievably well done, especially considering how often you crank these out. The work is much appreciated, I can always count on your videos to help keep me updated on what's going on.

[@MDUD777]

Asianometry doing gods work

[@ThootenTootinTabootin]

Difference between researching for a video and just telling people the cool stuff you already know about.

[@makisekurisu4674]

John is employed in the Industry right?

[@squirlmy]

@@makisekurisu4674 Nope! He studied business in San Francisco, then moved to Taiwan. It's probably why he's so adept at economics and history, and it also explains why descriptions of engineering in such an accessible manor. I think very few in the Taiwan chip industry are so educationally well-rounded, most are engineers who specialize in one field.

[@elodiefiorella4750]

@@squirlmy Out of curiosity, where did you learn about John's background? I've seen snippets here and there in the comments, like your comment just now. Knowing the story behind Asianometry sounds just as interesting as the videos themselves.

[@omerkaya545]

My condolences to the passing of Mr. Stacking, I personally warned him of the these Glass Panels.

[@richcole157]

Sure you warned him, but what was he to do?

[@rollinwithunclepete824]

I am kept on the leading edge with Asianomtry videos! Thanks, Jon!

[@catsspat]

背 (se : back; spine) 割れ (wa re : crack, broken piece)

[@recklessroges]

Thank you.

[@MichaelOfRohan]

Now straw, and camel.

[@BertLensch]

When you started talking about organic warping due to heat it reminded me of some of the research that was being done at Georgia Tech's Package Research Center when I was providing IT support there in the early 2000's. And then low and behold you mentioned them and the research they have been doing since I left in 2008. I think if Corning and chip manufacturers can master the glass substrate it will be a benefit to not just being able to package chips more efficiently, but also help make glass stronger. I know Corning's Gorilla Glass has some of the best tinsel strength out there, but I would think they would be able to evolve it to be even stronger than it currently is to be able to stand up to the cutting and drilling of the panels for use as a die.

[@weedlordbonerhitlerii3862]

gorilla glass is more of a surface treatment akin to surface carburized steel, they do a chemical bath on the finished product to cram bigger potassium ions into the matrix, creating a surface layer of enormous compression stresses. Materials Science is a crazy field, you can go as crazy deep as you want into any one little thing and never get to the end of the hole

[@mytech6779]

Strong glass does not mean more machinable. Often quite the opposite due to the balanced internal strain that makes it strong.

[@AsbestosMuffins]

almost seems like a return to the 1980s silicon on saphire tech

[@cv990a4]

Corning Glass Museum in Corning NY is pretty cool. If you're ever near Corning (or Cornell or Ithaca) it's worth a trip. GLW is pretty special company that's managed to stay at the forefront.

[@lukeeisenbraun5428]

Katherine Bourzac has a writeup in the October 2024 issue of IEEE Spectrum (pg 8) that describes how a team in Switzerland is using x-rays (PyXL) to image 3D chips down to 4nm resolution. This could give new life to 3D design testing that is very difficult to do right now as this excellent video describes. Glass recipes are highly guarded trade secrets to whom the spoils go.

[@ycplum7062]

"Recipe" is the right term. This is more of a Black Art than a science. Much like a complicated recipe, trace elements can make a huge difference. Even if you know all the ingredients and their exact proportions, the processing may be complex and exact. Slight deviations in timing, pressure, or temperature can ruin the batch.

[@michaelmoorrees3585]

Kinda new, but reminds me of an older chip/display tech. Think of the active matrix LCD display. There they grow the driver circuitry directly the glass. TFT, or thin film transistors, located close to, and individually driving each pixel, making the pixel speed much faster. A Japanese company, just for fun and yucks, built a Z80 uC right on the glass, but it was large. About 2 inches by 2 inches (50x50mm). This was in the 1990s.

[@SeanBZA]

Yes, that company was Sharp, who made the glass Z80 processor, and it worked. You can probably use this with the glass substrate as well, growing vapour deposited silicon active circuitry, then a thick passivation layer that has terminations through it, which then is ground flat for bonding the chiplets on top. bottom has the connections for the PCB, then through holes to the top, and you could even have active layers both sides.

[@Csupenner]

Very nice coverage of advanced packaging! One possibility of through glass vias is laser induced deep etching (LIDE), a really cool technique where quartz is first modified by a laser and afterwards etched away. Seems to produce some strain-free and very steep vias.

[@Keiseru]

I love how fast chip technology is being pushed right now. It's scary how high the barrier of entry is though. If there's any area where subsidies of more companies is actually a good idea its in fab technology. We can't just rely on a couple of companies or technology development will eventually stall.

[@frankstrawnation]

Bad idea, subsidies are poison to innovation.

[@grizwoldphantasia5005]

Subsidies don't reduce the cost. They actually raise it by adding layers of bureaucracy and tax collection. All they do is redistribute the cost from those actually benefit from the process to everybody, including the poor who pay for something they don't use.

[@ttb1513]

The subsidies for chip manufacturing influence where chips are produced more than they change what companies produce them. The barriers to entry are indeed tremendous, and subsidies being deployed are not changing that, creating new producers. IMO, anyway.

[@Keiseru]

@@ttb1513 i agree. Which is why I said subsidies of more companies. I dont necessarily agree with just enticing production to move or that current subsidy goals is good.

[@makisekurisu4674]

I don't think so, they've learnt the lesson from the 14nm++++ ordeal of Intel. Look at Nvidia, it's pushing the envelope but Jacking up prices instead.

[@joshieecs]

the corning achievement of packaging tech

[@yxyk-fr]

13:47 we have TGVs in France, they are pretty fast ;-)

[@lucasrem]

Bullet Trains in Japan, old electro tech, not real high tech.

[@fnordist]

Yes, it's about time they start using glass panels; especially since heat transfer is much more efficient.

[@Indrid__Cold]

Fascinating the way repurposing an older technology can suddenly restore or even enhance its relevance. Glass substrates certainly look compelling from a stability and material compatibility perspective. If I were looking at a fascinating career possibility, I think material science would be a compelling option. The way we make "stuff" and the new and varied ways we can use it just seem to be on the verge of a Renaissance. At 66, I'm too old to embark on such a career, but I'm enthusiastic about the opportunities it offers curious young minds. Anyway, thanks for the relevant and informative content.

[@LS-mg4lz]

It would be really cool to see a video about thermal paste / pads that are used on high performance chips. I love your work, and look forward to all of your videos!

[@peterweller8583]

I love your blog it is a bit overwhelming of corse, but like Bob Ross, a classic.

[@muskepticsometimes9133]

about 15 years ago I saw excellent presentation by Georgia Tech prof on glass for packaging. it was more to replace PCBs 1) much better dimensional stability than pcb 2) fairly low cost 3) leverage high volume roll-to-roll glass production for flat panel 4) flatter, takes hotter temperature 5) Thermal expansion closer to that of Silicon 6) cheaper than Is 7) glass roll-2-roll you can get huge areas - see flat screen TVs

[@lucasrem]

muskepticsometimes9133 We all saw that, now we can use algoritmes to develop any mask.

[@tommihommi1]

8) possible to make much finer traces in it than in normal PCB substrate due to mechanical and dielectric properties

[@delahaije25]

thanks for your high wualty and information packed videos. i really enjoy listening to them on a long drive. ❤

[@pietroferlita]

Keep up the good work Jon!

[@TheBlackBuddha17]

I look forward to these videos every day. I love the asianometry youtube channel!!! Keep up the amazing work Jon!!

[@tomas5695]

Love these videos. Very interesting even though I know very little about the tech subjects. Fall asleep to them and in the morning watch them again.

[@XSpImmaLion]

Interesting development... glass substrates. Have to wonder if the recent EU anti-trust lawsuit against Corning had something to do with this future development. If this switch happens, it's certainly going to change quite a few things in chip development and production.

[@saultube44]

Maybe sound waves can help, making vibrations for cutting paths, so the cutting will propagate through the vibrating thin stream, or something like that; then a laser could heat up the line, and finally a small saw can small cuts to propagate a extremely fine breakage that won deviate

[@joaovitorsilvagohl682]

with those new MEMS cooling microdevices couldn't a company use them to cool the stack? i mean embedding a MEMS inside the stack to remove the heat from inside?

[@FloydMaxwell]

This seems like the way forward, for sure

[@mytech6779]

I'm curious if they have tried a pre-cutting and scoring process, doing most of the material removal at the start with raw glass, then an intermediate chemical/heat treatment to repair and smooth any micro-scratches that would later cause cracks. Then doing the fancy substrate creation, and finally finishing the last little bit of cutting.

[@douro20]

Chips on glass are definitely nothing new- most modern active-matrix LCD and OLED displays have one and Sharp makes them with deposited silicon on glass, a technology they call CGSilicon (Continuous-Grain Silicon). As for cutting those glass interposers, ultrafast pulsed lasers may work. Disco has already successfully marketed this technique for cutting silicon, calling it "stealth dicing" as the lines are so thin that they are hardly noticeable when the dies are inspected.

[@Leicht_Sinn]

I find it funny to see you already pointing out what technology could be in the near futhure -> which most likely will have an impact even at my company in the long run

[@elon6131]

11:55 - did someone just mention high speed trains?!

[@youcantata]

How about polysilicon wafer for interposer or chip base? Not monocrystal silicon wafer used in semiconductor. Polysilicon wafer is way cheaper and easier to work with than monocrystal silicon wafer. It was rather expensive material (but way cheaper than monocrystal wafer) in old days. But is widely used in photovoltaic industry, so now dirt cheap like $1 USD per wafer. Very similar electric characteristic to silicon wafer and comes in square shape. Good material for interposer or chip die base.

[@LaserFur]

Isn't polysilicon slightly conductive due to the grain boundaries?

[@calvinhobbes1617]

I would think due to its anisotropic etching behavior it is not suitable for the semi process.

[@jpjude68]

dunno if you saw, but the IEEE has had project going on for a few years now called "heterogeneous integration roadmap", there's freely downloadable papers on industry trends and research in semiconductors, manafuacturing techniques, and in the scope specific contexts (like automotive, defense, etc...) it's really interesing, many papers

[@keyboard_g]

Hewlett Packard was building silicon on sapphire chips in the 1970’s. Sapphire is extremely strong, heat, and radiation resistant. We’re coming full circle.

[@lucasrem]

keyboard_g We still apply them on special utilities, in orbit etc, through the radiation belt. We’re coming full circle ? What is it you do, who you call we ? All Keyboards ?

[@AlexRoivas]

7:40 In the early 2010s, there were prototypes of 450 mm waffers but TSMC at the time couldnt fund them so out of all the semiconductor companies it was them, that you could say, sabotaged the plans. Now it seems we are stuck at 300 mm waffers

[@aware2action]

The issue is dis-similar thermal coefficients of materials between the conducting and semiconducting layers. Anything planar is also suseptible to expansion/warping due to non uniform thermal expansion. What we need is a graphene impregnated glass sub layers, to wick away the heat efficiently, avoiding localized heat buildup. Also we can assemble all these into a dodecahedron structure for the most flexible interconnect🤔. Are we trying to re-create structures similar to what is in our 🧠🤯. Just some 💭❤👍

[@ModernOddity728]

Was not aware the glass came from Corning, I've been to the Museum multiple times throughout my life, always an interesting place to visit!

[@LaserFur]

We used to make large glass headlights. so that ribbon of glass could be pressed with hot presses into squares with tiny holes.

[@douro20]

The image at 2:50 is that of the multi-chip module out of an IBM POWER 595 super-minicomputer. It has four processor dies and four cache dies.

[@GigAHerZ64]

It's not "warpage"! It is "pringlesation", as it when it warps on one axis, it forces the perpendicular axis to warp "the other way" because the heat is on a spot. ;)

[@sync232]

Why are chips not round?

[@karehaqt]

Will this glass have scratches at level 6 with deeper grooves at level 7?

[@Inceptionxg]

When I said we should focus on semiconductor, many Saas entrepreneurs said that is so complex. We have to improve this for the future of computing. Please, we must have to talk about this.

[@michaelsnow3]

Guess I should've expected the mention of GATech's PRC, but it still took me by surprise lol

[@recklessroges]

Glass chips, glass ROM storage. It makes sense with so much Si on this planet.

[@williamhoodtn]

These PCB and silicon Via are pronounced "Vee-a".

[@recklessroges]

It's vhi'ya in my dielect of english. It's like "Router" (US) and "Router" (GB).

[@sjoerdvandriel7201]

Great video once again! Can you maybe follow up going deeper into optical interposers and light sources? Interested in AMD and Samsung investing in Celestial AI and POET Technologies working with Mitsubishi for using 400G EML.

[@jobidi99]

On the topic of the organic substrate layers, have you heard of ajimoto build up film before? It's one of the materials used and sometimes a bottleneck, as it's produced as a side product from soy sauce production. Thuis was hilarious to me and I think it would make a funny tangent in the future

[@WolfmanDude]

I wish they would use more ceramics in electronics, its a wonderful material!

[@Gersberms]

I really love your videos and I have a request: do you know how the dead pixel problem was solved in LCD panels? I remember from a while back that it was said Samsung had solved the problem. I've basically not seen any dead pixels ever since.

[@andymackie2896]

We need to differentiate glass panel level packaging (GPLP) from polymer-based (PPLP)

[@soiltek2015]

Great video!

[@Kikay0n]

What about micro water jets?

[@llothsedai3989]

I think what you're actually looking for is 4d circuitry, where it self reconfigures due to the software update.

[@Toksyuryel]

In the future, an AI-managed FPGA could pull this off.

[@afc8981]

An FPGA?

[@Clos93]

​@@afc8981FPGA (field programmable gate array), an integrated circuit that can be repeatedly reprogrammed after manufacturing. Think of an Intel chip that can turn into an AMD chip, then an old Pentium, and then a Super Nintendo.

[@AlexAyoub_]

en.m.wikipedia.org/wiki/Field-programmable_gate_array

[@dtibor5903]

I have never thinked about the FPGA as a 4D object

[@TT-lf5hi]

I am not sure why these are considered problems. Cracks can be annealed out after processing if required: The vias can also be made with etching if it is a requirment (feasibility depends on the thickness) KOH & NaOH work for etching glass, hydroxide is not used for silicon etching because it also etches silicon.

[@Asianometry]

I’ve never heard of cracks being annealed out of glass. Is this something you’ve done before?

[@TT-lf5hi]

@@Asianometry Experience: no galss working experience Glass however undergoes viscous flow and at high temperature. This allows for healing of cracks and bridging of gaps (reference paper mentioned in seperate comment). Hairline cracks are likely to immediatly undergo healing due to the short distance and lack of trapped air.

[@TT-lf5hi]

@@Asianometry youtube immediatly removing DOI. Title: Thermal healing of cracks in glass Publication Date: 1988 Journal: Journal of Non-Crystalline Solids Shows the crack healing and includes images of the sample.

[@stephengrey1880]

Super interesting and well produced, as usual. Love your content, thanks!

[@davidsimmons8990]

ROTFL😂 cte… via studying advanced packaging terminologies!!! Awesome brother… keep on keeping on…

[@LanceMcCarthy]

Lol, CPE from studying packaging tech terminology

[@peterweber79]

A synthetic, flawless diamond would be perfect, except... ... 🤑

[@tulsatrash]

Now this is exciting.

[@SaccoBelmonte]

What about cutting dies with a nanometer scale beam?

[@michalchik]

Diamond interposer and packaging are going to be the next big step. We've already gotten pretty good at thin film coatings of diamonds and we can thicken it. It's going to be better than glass in every way.

[@ec2_alcatraz]

Sounds a little bit like you forgot cost in your way(s). I only know of small diamonds being produced. Or Was that maybe a joke I did not get? 😂

[@adr2t]

@@ec2_alcatraz Cost will be there, but we can make diamonds pretty cheap. The hard part will be scale of larger sheets than it will be to really produce them at a cost. There is already a way to produce them in sheets, but I havent really seen heard much past that. I assume its possible though, but last I heard there was issue with the uniformity of the crystal.

[@michalchik]

@ec2_alcatraz it's not immediate but it's coming. As I mentioned we can already diamond coat most substrates. Yes it will cost more but it will start as high end high performance devices and then come down. I was joking a little bit, but I really think we can make this happen in the not so distant future.

[@3jims]

Most definitely sounds cool

[@ModernOddity728]

I'm sure you heard about the huge gaff of Pat Gelsinger bad-mouthing Taiwanese Semiconductor industry leading to TSMC dropping the discount Intel had on their chips. OOF. Hope they are able to stay independent and start making the right moves.. So far they haven't given much hope of either, rumors are they may merge with Samsung. Regardless, I thoroughly enjoyed the video, thank you!

[@ec2_alcatraz]

Did i here that right - blackwell is so huge they only get 4 chips out of a big 300mm wafer??

[@brandonv8721]

Isola recently EOLd FR4

[@BastyTHz]

cant we just use photonic chip then ? i doubt it would be much heat enough to melt

[@diegoantoniorosariopalomin2206]

Could you talk about cryogenic computing? Is an alternative to traditional scaling that doesnt reinvent the wheel. Cryogenic cmos performs better without fundamental changes

[@paper_airplane]

There was a video "Why IBM's Superconductor Computer Failed" on this channel. Nevertheless I also would like to see a video about more recent research on that matter.

[@diegoantoniorosariopalomin2206]

@paper_airplane cryogenic cmos isnt about superconducting computing

[@belstar1128]

12:35 why don't they just smaller moulds like 1 mould for one chip instead of cutting one big piece of glass

[@BetterLifeCreations]

Grat video again I learned something despite I’m in the industry 😂❤👍

[@SFVYachtClub]

So sad to hear Stacking died

[@ivanalaskevich4736]

Normal people: mmm technologies Internet: solder's balls

[@lucasrem]

Algoritmes can translate any mask design, developing chips on glass and other silicone alternatives.

[@tonomoto]

damn man! your videos as really awesome!

[@adamk.7177]

"glass cracks" Me: gotta use water and a diamond saw, bro

[@LaserFur]

Or use a die and punch the holes while cutting it to size. we used to make large headlights out of glass and I could imagine having that ribbon of glass going threw hot rollers.

[@cooperkramer2985]

Albany Nanotech Next!

[@erictayet]

Ah, AMD RDNA3 organic substrate seems to be the limiting factor in their GPU's infinity fabric. The latency was too high and affected gaming performance. I hope Intel finds a way out of its current funk. Is Intel's Foveros based on this new glass substrate or some sort of TSMC CoWoS tech? Intel has so much technology, even their own product managers are lost if they don't talk to the research fellows.

[@MarkoCloud]

I was wondering where stacking dies. 😊

[@chrisbampton5225]

Glass makes sense because we could use it with photonics

[@chinesesparrows]

No don't die Stacking!

[@dazombiekilla9990]

RIP

[@perguto]

Weirdly low view count compared to your other videos. It didn't show up in my feed even though I follow your channel religiously. Maybe it's because the algorithm doesn't like "dies" in the title 😅

[@hiyan1]

Glass is glass....

[@Toksyuryel]

I feel like the endgame for 3D chips is going to be laying the stack onto its side instead of keeping it vertical.

[@szurketaltos2693]

That would be neat, but how do you remove heat effciently? Needs some sort of through silicon heat pipes, which adds even more complexity. Even crazier would be if we could minaturize some method of shedding heat to neutrinos or other weakly interacting particles.

[@LaserFur]

@@szurketaltos2693 if we had room temp super conductors it could use electron cooling. Otherwise passivizing each thin wafer and liquid cooling would work, but if you ever have had to clean a skived heat sink you will see that it's not easy.

[@szurketaltos2693]

@@LaserFur hm, how does the former work? Is there a good resource -- I couldn't figure it out by wiki. As for liquid cooling, yeah microfluidics is not easy but it is possible. Can apply pumps from biomedical applications.

[@LaserFur]

@@szurketaltos2693 one research article mentions "Both electrons and phonons are present in a metal, and they can act as carriers for heat transport. Nonetheless, conduction through the free electrons within the metal lattice is typically considered to be the dominant mechanism of heat transfer in metals." ref electrons-phonons-promote-heat-transfer-material-systems

[@wololo10]

bring back monolithic

[@charlesmurphy5644]

9:55 Really enjoyed the joke Asionometry Man!😹

[@alexgishnock5409]

glass is terrible for thermal transfer, those chiplets are gonna smoke

[@LaserFur]

luckily the glass would just be on the PCB side. That leaves the other side for copper.

[@bgone5520]

that sounds a little fragile

[@taorente7438]

Taiwanese manufacturers have developed stone paper materials by grinding stones into powder and bonding them, along with exclusive production equipment, to mass-produce the world's only stone-based paper products. These products are used for food packaging, meal boxes, and medical transport and preservation. They have been strongly recommended by the European Union for use in food preservation and eco-friendly dining utensils. Stone paper products, sourced from stones, significantly reduce the amount of tree cutting in the global paper industry, protect forests, and help green the earth by reducing carbon emissions. Stone paper can be used for packaging and transporting frozen food and pharmaceuticals at -40 degrees, as well as for high temperatures exceeding 100 degrees, unlike regular paper products that can only be used at room temperature. After being recycled, stone paper products can be sorted and buried underground, naturally decomposing and returning to the earth as stone material. The innovative invention and mass production of stone paper are a point of pride for Taiwan!

[@MadawaskaObservatory]

1M subs in 2025

[@Vatsek]

You could get some of those chips, cut them, polish them, and take pictures.

[@jmirodg7094]

👍

[@larryslobster7881]

another win for si

[@fredinit]

Two things that glass substrates, or even all-around, can add 1) Easy access for optical interconnects 2) On-die diagnostic LEDs. Think of the die displaying a QR code as it runs. As always, excellent content.

[@Matt_The_Hugenot]

Every time Intel gets written off they come out with something new.

[@NobbsAndVagene]

'ganics are overrated, choom.

[@TheGreatAtario]

Oh no. Are LCD panels really being pushed out by burn-in-prone OLEDs? That sucks.

[@vitaliypro8441]

I use to love your videos, but it seems like you turned on aggressive advertising options or something. Way too many ads

[@ThroatSore]

Comments turned off on China V Vietnam. Poor.

[@ninupimps007]

When are you making video on trump election win to semiconductor industry?

[@AlexRoivas]

If tariffs are added, will everything that has to do with the semiconductor industry go up?

[@TheBackyardChemist]

afaik a common material used in the making of organik stubstrates is ABF, Ajinomoto Build-up Film