Intel: We're Replacing PCBs with Glass Core
Vložit
- čas přidán 13. 06. 2024
- Ever wondered what the green stuff is when you hold up a processor? It's like the motherboard, right? Actually it does more than that, and when we look to the future, it doesn't look like the technology will scale to big chips in the future. So insert Glass Core Substrates - the next wave of technology, coming to a CPU near you.
[This video is not sponsored by Intel]
-----------------------
Need POTATO merch? There's a chip for that!
merch.techtechpotato.com
more-moore.com : Sign up to the More Than Moore Newsletter
/ techtechpotato : Patreon gets you access to the TTP Discord server!
Follow Ian on Twitter at / iancutress
Follow TechTechPotato on Twitter at / techtechpotato
If you're in the market for something from Amazon, please use the following links. TTP may receive a commission if you purchase anything through these links.
Amazon USA : geni.us/AmazonUS-TTP
Amazon UK : geni.us/AmazonUK-TTP
Amazon CAN : geni.us/AmazonCAN-TTP
Amazon GER : geni.us/AmazonDE-TTP
Amazon Other : geni.us/TTPAmazonOther
Ending music: • An Jone - Night Run Away
-----------------------
Welcome to the TechTechPotato (c) Dr. Ian Cutress
Ramblings about things related to Technology from an analyst for More Than Moore
#techtechpotato #intel #glass
------------
More Than Moore, as with other research and analyst firms, provides or has provided paid research, analysis, advising, or consulting to many high-tech companies in the industry, which may include advertising on TTP. The companies that fall under this banner include AMD, Armari, Facebook, IBM, Infineon, Intel, Lattice Semi, Linode, MediaTek, NordPass, ProteanTecs, Qualcomm, SiFive, Tenstorrent. - Věda a technologie
We asked Intel for more transparency and we got this
omg underrated comment
we sure got more transparent from intel
Computer nerd dad joke
I love hearing how old future technologies really are. It never ceases to amaze me that the cutting edge ten years from now was being worked on ten years ago.
Its not cutting edge. Not even close. Nothing sold to the masses is.
Same, you never know what big shit is being worked on
@@cumbob Sure you do.
Just wait till the 10+ y/o diamond chip wafer becomes mainstream 😉
more like some tech are 20 years ago
Glasses (and ceramics too) have some *great* physical properties!
Great to see Intel exploring this area!
Hey Intel. Invite the man so he can inform your future customers about the cool and exciting things you guys are working on!
Up
This comment makes much more sense at the end of the video. I concur.
Up
Dude I swear us engineers are shit at talking to people, and the marking guys don't speak engineer so it's hard to convey how exciting some stuff is.
HEY INTEL! INVITE IAN!!!!!!
The content is so engaging and informative that I wouldn't have noticed the background noise if you hadn't pointed it out. But, thank you for being kind enough to a acknowledge it.
Man, Glasscore has been my favorite music genre for a while and I'm just happy that more people are getting into it now~
Musicians pour their blood, sweat and tears shredding on the glass harmonica.
Literally.
New materials are coming down the pipe for printed circuit boards as well. Some of them are pretty mindblowing. Similar timelines are expected in some applications.
name some
@@_xXx_YTlee richey at speeding edge has done some work on this. His website has documentation but it's paywalled. Im not aware of any public domain documentation at this point
“things are happening. it will affect other things. some of the things will be a thing soon”
@@anieldev we live in a society
@@louiscrasher haha nice. you passed the turing test with that one
Good to see intel coming back and pursuing other materials to drive innovation. I'd just ask what are the benefits in terms of die size, frequency, power consumption, etc? Is glass necessary to support a robust chiplet ecosystem with UCIe? What about cooling performance, is glass a better material for more effective cooling? So many questions. Thanks again!
Signal integrity seems to be the biggest thing
I think the biggest deal would be wire size. If you saw AMD talk about RDNA3's MCDs requiring an order of magnitude more interconnects than their CCDs, and splitting into multiple GCDs would require an order of magnitude more than that, this is where that comes into play. Since they can make the wires and contacts smaller, you can have higher bandwidth between chiplets at lower power and lower latency. I wonder if this will replace passive silicon interposers.
Packaging is also a thing they have done a lot of research on. Back in I think the late 90s/early 2000s, they worked a lot with Ajinomoto (yes, the company that also makes instant ramen) to work on a new, increased density packaging system. Moving away from the fully woven types of substrates (with cores and prepregs) they moved to a core with what some call ABF (Ajinomoto build-up film) to get higher density and thinner substrates (which in turn allows for smaller vias and tracks, and thus higher density). These became the defacto standard for anything high density up to recently, where CoWos and such have been taking up a lot of the demand.
the benefits are probably going to be something like cheaper HBM3 without the huge silicon interposer, but probably a bit worse performance then the interposer, the larger benefits is its going to make chiplets that work without silicon interposers a hell of a lot better, right now on organic substrates IF can only reach pcie4.0 speeds and this limits memory clocks too, and the interface between ccxs on dual ccx parts is literally slower then dram, AMD already made a special PCB for RDNA3 chiplets, but Glasscore could allow Chiplet CPUs to reach pcie5/6 speeds, and who knows maybe its cheaper then the thing they used for RDNA3, or maybe its faster.
Intel is going to want this kind of thing for Tiled CPUs too, the whole industry is slowly moving to chiplets, i think it makes sense to have all the power hungry chips under 1 good cooler, but the whole gluing together a bunch of smaller chips idea is mostly just a temporary cost saving thing, today we have CCXs and IODs, and Intels going to make a GPU chiplet, probably AI acellerators or something, and we have off chip memory, but in the future when EUV lithograpthy gets cheaper it will probably just be a big fat integrated die with some memory chips around it, a lot like Vega.
But the million dollars question is: is it edible?
Not from the nutrition stand point but if you just want something to chew on, than yes.
@@zperdek a special type of gum
I mean why else would they be called chips?
"Ask not if it is edible, but does it burn on the way out!"
I'm pretty sure that's how the quote goes
@@user-cz9jf1ec8sthey arent called chips, thats what leeks call them..🥹
+1 for an invite for Ian. Make it happen Intel! I want to know more about this glass wizardry.
"the laws of physics get in the way" is such a raw line
Thank you for explaining it in a good way. I got this news in my news feed and didn't know what it meant. Your video is the first one I checked to explain it and you did a good job.
Cheers.
The advancements as of late with this glass based substrate and the photonics data communication looks very promising. Please Intel send Dr Ian Cutress an invitation so that he can explain all these highly interesting tech you are developing.
Feels like Intel has now gotten a bit of a fresh air again as of late, very exiting!
I'd love to know how it's significantly different than a ceramic substrate of yesteryear. They had that figured out decades ago and moved to the organic substrates because they were somehow better. What's old is new again, I guess.
The difference is the way you manufacture these glass substrates and the integrated Through hole vias, and future more complex, interconnects. Being transparent, you can use laser technology to make complex 3D structures inside the glass. Intel have about 2 years ago purchased a small scottish SME who have the technology to achieve this aim.
@@robertrjm8115 That's an interesting idea regarding the lasers, but I've seen/read nothing about lasers being involved with this tech to date. I'll have to do some more digging.
Reminds me of all the weird transparent circuit board looking things in Star Trek
Subscribed! I love contents like these!!
Super excited!! Hope you get to do a deep dive on this eventually
This is fascinating. I thought of this idea years ago but not in the context of substrate. Thermal stability. Big yes.
5:03 "You've likely seen that package size is growing."
It is always amazing to see even a brief peek at how modern semiconductor tech is fabbed. I am very eager and excited every time someone on youtube has a fab visit / tour video. +1 let's get Intel to invite you!! This stuff is incredibly fascinating.
Really excited for this innovation and progress,,hope to see more about this with upcoming videos.
Is there a full version of your outro music? Absolute banger everytime lol
Fascinating mate. More please.
Excited for the future, thanks for the info.
During your explanation my first question was what about thermal dynamics? And Yes, this kind of a break through gets me excited about technology again. the thought that it could produce transparent substrates in the future is an inviting thought with liquid cooling. Lets face it PC building nerds love our esthetics as much as new technology.
I'm a lot more interested in what's replacing silicon. Gallium Nitride (GaN) and Carbon Nanotube (CNT) chips are needed more so than replacing the substrate die packaging. Sub
Better get that invite my man cant wait to hear more about this tech
Great to know, thx a lot Doc Ian!😁
Yes Intel pls invite TechTechPotato. I would love hear more information on the glass substrate.
Wow, that's really interesting! I want to see more about those Glass Cores!
Hey Intel! I'm very excited about your development of glass substrate. We really you to give our favourite teacher Dr. tech tech potato an extensive tour right away. 🤘😎👍 New chips please.
This is really cool and I really hope Intel can give you more info to share. I am full of questions as well. The pictures look like they expect normal copper connections on both sides of the glass, but is that really necessary? If the flip-chip has VCSELs on the surface, and they take a similar amount of die area and power as a pad for connecting a pin, maybe that's how Intel plans to get by with fewer layers. Wild guess though.
Intel, please invite the Dr. to your R&D labs, this is very exciting and we are all watching!
A visit to the fab would be great content imho. Thanks!
It's still going to be a composite, with copper or tungsten wiring embedded in a glass matrix
Love it! The via aspect ratio and the smoothness of the surfaces really are game-changing. As for the optics path, wow! Amazinq! : )
I Would love to see more about this topic. At one point pcb's need to have linear guides as well, i think this is a technology that could make its way to pcb's as well.
I’m excited and appreciate this info!
Would love to heaar about this new application of materials and even the chance to see clear chips in the future.
Can't wait to see more about this tbh
In the mid-1980s there was a startup called "Trilogy" which was founded by Gene Amdahl. They were going to build "wafer scale computers." The big problem with doing this was clock and signal distributions, skew across the wafer. The proposed solution to this problem was to do the signal distribution optically, but that technology did not exist then. Off and on, optical signal distribution has been the subject of significant research. At one time, Intel was anticipating implementing this by 2017. Rather than the subject being the focus of research for the last decade, it has actually been on the table since at least 1985.
From the glass tubes to ceramic to organic and now back to glass, electronics making full circle
I want to see a video from you at an Intel chip visit! I also really want see through chips, that would be super cool. Would glass core be more conducive for integrated water channels on the chip itself?
Sorry if I missed this, but I would really like to better understand the challenges they are facing with the glass. Basically, why haven't they been using glass all along?
I wonder if they are also thinking about incorporating nanoscale heat syphoning or redistribution (eg., by incorporating graphene channels into the glass).
Oooh, can you imagine transparent substrate? So you could see the chip, the logic (see it. Not…being able to see individual transistors and wires). That would be awesome.
Fascinating. I sure hope to learn more about this and also if it will lower the cost of current manufacturing? Here is hoping for a Factory Tour!
This technology definitely excites me. When you already have a use for such technology then it interests you. Waiting is excruciating because I have already been waiting decades as it is. 😊
I’m interested in seeing how long before they start doing build outs with circuitry in the cores.
Get this man in your fab Intel!
How the quantum effect behavior of electron is adressed in the new substrate? do the glass core just reduce the effect or it is still the same but the conductive layer can be construct in a other manner so to circumvent th effect? I'm not sure to comprehend fully how the quantum tunneling is the "wall" limit of silicon tpday.
What will be next in motherboard PCB design ? Creating PCIe 5 is expensive and there is signal degradation problems will there be something before optical pcb ?
Using glass seems like it could be a new half step in between a silicon interposer and organic substrate. The benefits of a "si substrate light" as the base package seem obvious but we'll have to see if cost comes down enough to use it in a lot of products. It will be clutch for very large multi die products which would otherwise be size limited by the cte mismatch with the organic material so it should have some market even if it doesn't become cost competitive.
fab visit are like the best content on silicon you can get, they are so cool
Mark my words: the semiconductor industry will (in the midterm) have to replace most both high-bandwidth & low-latency electrical data interconnects with optical ones for both economical & physical reasons - including inter-device, inter-die to possible inter-core/ip connects. Think short-range fiber optics across dies/inter posers & rigid silicon wave guides built on die using various methods coming out of optical computing.
Someone give this guy a medal… medal in shape of an invite to the fab where he can get us more info on this.
Cool stuff, keep it up.
I am wondering if this will allow chips to work in higher maximum temp, which would give headroom when 3d stacking
Very interesting. Get him a tour already Intel.
A tour (and the possibility for an interview) of fabrication (R&D?) would be an amazing form of marketing for Intel. Consumers that are interested in Intel's future technology (i.e. your audience) will be sitting on the edge of their seats seeing the innovation. Hope you are able to make it in for a tour!
I wonder what dielectric/isolation they apply between the layers
would a glass core make it possable to layer opitical transmission channels? Could we see the chip and the package become one?
Someone listen to this guy
I wonder how this new substrate will affect the overall coefficient of thermal expansion of the package and minimum bump pitch. It's easy to imagine it'll behave much more like silicon than organic polymers, but to what extent?
More give us more. And great video as always.
Dear Mr Gelsinger, please take Dr Cutress to your lab and show him the cool things, thanks, Henrik.
I'm keen to see how it pans out. One immediate thing comes to my mind though and it'd be an interesting question for Intel to answer. The organic substrates have some flex in them naturally, how would a glass substrate cope with the high compression forces placed on it by a socket clam and heat sink? what's their approach for dealing with the mechanical stress a chip might face during installation, would the glass break or is it strong enough to hold up to mechanical stress?
Dude I pray we some day get see through packaging not only for the CPU but for the mother board substrate too. Just imagine all that RGB going on around the case. Daaaamn
Intel invite this man to your facility!
Well, I expected this to be an understandable explanation but instead at about 4:57, I felt the same way about glass-core as I do about the retro-encabulator.
This is soo cool, I have always been a huge fan of ceramic and glass in electronics. Very underrated materials imo. If they get this working I would switch back to intel
It seems like more and more power goes towards computation but what about memory? Either way great video as always!
GDDR7 and DDR6 are on their way.
@@vyor8837oh I know memory will always keep ratcheting up but from what I understand DDR has some latency limitations that aren't getting better. So if the CPU can do 10,000 cycles in the latency gap you're losing quite a bit of performance. (all hypothetical)
Feels like memory is very much make the same technology faster for when the data arrives but I could be totally off the mark.
@@Zaf9670 The fix is going to be on package very fast low latency memory with cheaper DDR as the bulk storage memory pool. Many of the limitations are imposed just by the distance from the CPU to the memory, having 4gb or 8gb of HBM on your cpu package would go a long way, AMD is already doing this with their instinct CPU/GPU combo package.
15 years after graduating in photonics, photonic integrated circuits start to show up. I did a master thesis about photonic crystals.
I saw a video covering how beneficial modern Analog processes are at running A.I. facial recognition, self driving awareness programs compared to.. on modern CPU's that use a branching algorithm web that demands a lot more effort & isn't as effectively accurate as analog processes. (I forget the name of the term for modern algorithm deduction search processes?)
Transparent would be cool but I guess it'd still be covered by the cooler anyway so not a big deal. Great to hear technological roadblocks getting cleared though!
It would be able to handle and transport heat to the cooler much faster!
If you've ever seen PCB's being made its a chemical plating process that creates the copper via's ... i picture it will be a multi-material solution using chemical plating on glass with solder-mask which will not make it look like a transparent piece of glass but the same as fiberglass PCB's
Just a curiosity question, what material is D? I have no memory of that or is the group D in the periodic table?
Not much longer until we have the crystals of the ‘60s sci-fi.
It’s interesting how anisotropic materials are often a pain and people try to make them approach being isotropic, but can also be exploited. The classic material is of course wood: anisotropic properties are exploited with beams and joists, but suppressed by making it into plywood.
Here, an isotropic material just seems easier to use.
The simple fact is that a lot of structural materials are asked to share all types of loads for simplicity's sake. Isotropic materials are easier to design with.
yes cool, invite Dr. Ian for a tour, and also for the research into CPU memory, I heard that Intel was doing research into memory in the CPU, excited for sure!
Very interesting tech, looking forwards to hearing more about this.
I sincerely hope Intel will invite you for a deeper look and chats with the real nerds. :)
This is the best form of marketing there is, and I hope Intel agrees.
I think with the decreased illumination area through High-NA EUV chip packaging will play a much more vital role in near future. Intel seems to be going the right way by investing in Glass Core Substrates because they enable more complex chiplet designs.
first what i tought was Ahh SCI FI had it right that chips will be in glass, but not yet what i understand :) I am all for bug leaps in tech like we seen with the ASML extreme ultraviolet lithography for example. Still a long way to go to see it in my desktop PC but would like to hear and see progress in this tech for sure!
Dang, transparency was the first thing I wanted as well! Give this guy an invite Intel
Does this also mean it will be more fragile then? If they put in on laptops or phones for example, will it break from a fall?
Are we talking about silica glass or is it any amorphous networked solid that fits the design requirements?
Maybe I missed something but how does this help Moore’s Law? You can pack more dies together and therefore more transistors? The idea being that you can just fab many (hundreds) of cheap dies and then just put them all on a glass substrate?
I expect to see the glass used as an integrated heat spreader with high-speed optical data/address buses to motherboards with fiber optic interconnects.
How would this accommodate force from mounting? My number one fear would be cracking the substrate.
Hey intel! Glass him and show us what's in our future. Lets see a sample!
Does it taste different to silicon chips and organic substrates?
Oh wow great, addresses upcycling also. A see-through chip would be nice though.....:-)
Does this mean we could get 12 cm CPU's? aka a HPC cpu 12 CM in diameter ? with liquid immersion cooling?
Old IBM CPUs in the s/390 used ceramic substrate.
I've heard that just being around this substrate will improve your mood, digestion, and metabolism.
i think its cool that we still have room for improvement after how far weve come
and when you drop it will it break like a glass?
I wonder if they'll use a machineable glass like macor, so they can drill and mill it?
There are properties of materials that are indicators of how efficient a material will transfer signals. For PCBs the main ones are Dielectric Constant and Dielectric Loss. Using these properties, glass is not as good as many other materials used such as Astra so I don't understand how glass can have less signal loss. I do understand that the fabric weave is a problem and glass could eliminate weave issues also thermal conductivity for glass is likely better.
One thing that helps with decreasing high frequency signal loss is to use a better conductor. Silver is often used as a coding on coper wires to reduce high frequency signal loss. Since high frequency signal tend to travel on the skin of the conductor silver plating can help. I have not seen silver used on PCB or substrates yet but my visibility is very limited so maybe someone out there is using it.
I wonder if this will be available to other manufacturers.
I would love for @Intel to give you open access to their in development projects