How AMD Left GlobalFoundries for TSMC
ฝัง
- เผยแพร่เมื่อ 11 ก.ย. 2024
- AMD began as an integrated design manufacturer competing directly with Intel - meaning they designed and made their own chips. When the company's internal manufacturing operations started to lag behind Intel, the company transitioned into being a fabless manufacturer.
Even after the transition, its foundry partner GlobalFoundries struggled to keep up with the company's technical and yield demands. Eventually the company would turn to the Taiwanese giant TSMC for its leading edge processes. Today, TSMC is AMD's closest foundry partner and the manufacturer of some of its best products.
In the video, we will look at the chip designer's move away from its manufacturing roots, its up-and-down relationship with GlobalFoundries, and how it came to fab its most advanced products at TSMC.
Links:
- The Asianometry Newsletter: asianometry.com
- Patreon: / asianometry
Check out other video profiles on this channel with this playlist: th-cam.com/play/PLKtxx9TnH76Qod2z94xcDNV95_ItzIM-S.html
If you want to support the channel, check out the Patreon: www.patreon.com/Asianometry
Friendship ended with GlobalFoundries. TSMC is now my best friend.
Primary school best friend didn’t made it to the college ;( But there are other cool guys!
Dead 😂
GlobalFoundries is parasite to AMD
*Chinese Communist Party has entered the chat*
AMD didn't start as a competitor to Intel. They started long before they did anything in the same product areas as Intel. AMD was originally famous for things like the 25xx and 29xx series of high performance bipolar logic devices. Later Intel were really keen to collaborate with AMD when Intel was struggling with advanced MOS processes that AMD had running smoothly. AMD got rights to the early x86 designs in return for helping Intel's fabs. In those days Intel needed a solid second source to penetrate things like the telecoms market, and the company was struggling in a number of ways, so they didn't mind sharing their x86 designs. When Intel got what it wanted, the demand for PCs made their fortunes soar, and the demand for fully second sourced parts became less significant, they tried to welch on the deal. AMD then took Intel to court.
You have more historic details than this video.
imagine having today single core cpu at 500$ and dual core cpu at 1k $ if it wasnt amd damt
@LunarVVolf Before intel can go anywhere they are going to either have good supply and pricing or hope demand for CPUs increases as current shortages make sticking with current cpu and waiting for gpu upgrades basically a necessity. So at least for Desktop and HEDT it is unrealistic for intel to take a clear upper hand anytime soon (5-ish years) That is if AMDs upcoming LGA chips are to be trusted.
Theoretically their chios should be a big jump in performance, I don't know how efficiency will turn out.
Do you perhaps have information I don't?
@@sayhongtan8993 his name is american, probably know the industry history better. School doesn't teach everything, only the profitable knowledge.
IBM actually told Intel to second source their chips back when the IBM PC came out. A ton of companies second sourced the 8086. AMD did second source the 286 as well but the reversed engineered the 386 and 486
Yes, that's a good summary of the AMD & GloFo relationship over the years. AMD's technological advantage actually started as a result of getting rid of its fabs, whereas Intel's previous tech advantage was as a result of keeping its fabs. Crazy how what was once the best solution is now the worst solution, and vice-versa.
but now Intel is trying to either buy Global Foundries or create it's own fab with the US govt subsidies
@@soup100
Yes it is an ironic world we live in. 🙄
Actually what I find even more fascinating is that tsmc created a new business model that didn't exist.
@@mdogzino That's not entirely true, Fairchild, IBM and Japanese companies worked with independent chip designers but the business agreements usually went the other way - the fabs licensed the design, built and sold the chips.
There were exceptions, though. Acorn, Commodore and Zilog were some of the first companies that designed logic without having a fab, and sold the product themselves. But there were lots more in the 16bit home computer generation. It started with glue logic chips and then custom controllers, acelerators, risc cpus and gpus.
The complexity of designing and supporting cpus with software was becoming as vast as manufacturing the things, so the split in industries was unavoidable. Intel was just lucky, ruthless and dumb enough not to follow in this path till now.
This channel has become my GoTo source for information on the semiconductor industry. Amazing quality 👍
Mine too.. please a video on who are ASML's pratners like ziess n other 12diffrent other USA n EU companies..n who could compeite with ASML a permutation combination..
Me too!
His Taiwan content is really good too
Check out TechTechPotato too if you're interested in wafer production etc.
It hedge fund quality info
Ugh... I thought I knew the GloFo AMD split pretty well... but I learned a good handful of things today! :D
I remember back then, everyone is roasting AMD for going fabless. Now, the tables have turned.
But Goflo needs to step up its node.
GloFo will, once they finish expanding their older node capacity. The
Small nit: AMD started by making minicomputer components, that is, the chips that made CPUs before they were all single chips. Mini computers went from the 1960's into the 1980s and were still more powerful than single chip computers until the late 1980's/early 1990's. AMD also second sourced the 8080, then Zilog's Z800, and finally the 8086 series that started their long patent wars with Intel.
AMD has come and gone as a competitor to Intel, more often than not a process generation behind that they "made up for" by being more efficient technologically. It was nice to see AMD finally pull ahead in both technology and process than Intel, but the game is not over. Intel is known for falling behind, then waking up and charging ahead again.
We'll see.
PS I designed hardware with AMD "bitslice" components in the early 1980's. IMHO AMD always was a step ahead of Intel in design, which Intel overcame by superior process and by sheer number of designers. I am an ex-Intel employee, but I bought Ryzen recently when I upgraded my home system.
Yes, but Scott just remember what you are seeing is not Capitalism. It really Socialism, with out government support TSMC would not exist.
@@marctemura2017 Sorry what had that to do with my reply?
@@scottfranco1962 Everything, because what you are seeing a mirage. A lot these companies can't exist without government support period. And another thing semi conductors are not a good business. Every year you must pour billions in research and development just to stay relevant. Do you know how much Coke spends? "Zero" in R & D.
@@marctemura2017 how much coke have you had
It was Dr. Lisa Su and her team that made the AMD to lead with Zen.
AMD wanted to stay american.. Dr Su is obvious chinese and trust chinese people(taiwanese are ethnic chinese)
@@destinilund4771and what, taiwan are entirely different country from china.
@@raisofahri5797Most informed American.
They literally call themselves the Republic of China.
@@WaterZer0Still they are completely seperate from mainland China.
@@WaterZer0我們不這樣自稱 中國共產黨 會威脅 台灣 發動戰爭 台灣名稱 比中國 好多了 我是台灣人
Free Taiwan isn't CCP China
TSMS is world's factory
Taiwan and countries in the world that share the same beliefs in freedom and democracy, together in pursuit of a better future for mankind
However the world needs a plan to survive the death of Free Taiwan, which may come from PRC or natural disaster. Because many of us remember the industry impact of Japan and California earthquakes hitting major chip factories at the time. If TSMC had major factories and engineering offices outside Asia, it could make their company dominance survivable.
@@johndododoe1411 Taiwan TSMC has set up branch factories in the United States, Japan, and China, and European branches are also being sought. It may be in the Czech Republic, but Germany is not very considerate.
The example of Ukraine...Taiwan will keep the most advanced technology in Taiwan
@@user-fi7js2zx5g Keeping the key source of essential supplies in the place most likely to be destroyed is really bad for business stability. Having the rich company bosses in that place is less of a problem because the company can grow a new head.
@@johndododoe1411 Europe has Infineon, ST (STMicroelectronics) and NXP (NXP), you can make the chips that Europe needs, you can get rid of Taiwan's gunpowder arsenal, congratulations
@@user-fi7js2zx5g Those 3 companies are not major VVVLSI chip fabricators like TSMC. TSMC could continue in exile if the home island is lost, unless some essential part is buried on that island.
Quality content as always. Big love from the UK
so manufacturing took down AMD in the past and Intel right now. Seems like TSMC should be thanked more than the Zen architecture for AMD's comeback
It's both, takes two to tango.
08:15 tsmc made ontario apu and it sucked
they could have a 2nm node, but if the zen design was bad it would have failed anyway
@@pinkipromise Ontario was competitive Intel was just giving away a billion dollars of Atom processors in a “contra revenue” strategy to break into the mobile/tablet space.
Yes because AMD was never far behind with Intel in chip design, only the process node. No because Intel was far ahead than everyone else in manufacturing until the first time they delayed 10nm since why bother with 10nm if people kept buying the 14nm products. in other words, weren't for Intel's greed, they should be on par or slightly ahead of TSMC today, another way, Intel chips remain slightly superior than AMD's chips on the same process node.
i worked at AMD around this time and the only thing i would add is GF20. not mentioned here was a complete frailer. yield was around 0.1% and AMD was only able to get 2 working chips after a year. i believe that is what pushed GF to sign a deal with Samsung for S14.
Love the farewell message. Was a pleasant surprise to hear a little caring message at the end.
Excellent! When you are bored, would you consider putting out a video on AMD's join venture with China and China's Hygon CPU? Or, more broadly, consider to make a video on China's adventure in x86 CPUs, including the development of ZhaoXin as well?
That's a very interesting topic! I Hope It can be considered for a video.
I’ll add it to the list. Thanks for the suggestion.
Just because of money and "competition", AMD has become a traitor for US. Unlike Intel that never gives the technology outside the US, AMD has spread the technology to whatever country that can give them money.
China knows this very very well and ready to get the technology from AMD to beat the US forever and forever more.
@@jacwi8719 Did the US punish Intel heavily when they bribed OEMs to not use AMD CPU's? AMD is doing everything to survive. They don't have Intel's market share thanks to Intel's very dirty business practices and AMD's own bad decision to focus multicore CPU's like Bulldozer architecture.
And Intel does give US technology to outside the US. They have Factories in Israel. The difference however is China is an enemy and Israel for now is not.
@@grimgoreironhide9985 I don't know about that. Does China do that?
The supply issue means Globalfoundries cannot deliver technology in time and AMD keeps losing its market. That's not a good thing if AMD wants to survive. I think that's why AMD changes it foundry partner from Globalfoundries to TSMC. Until now, it has been proved that the decision is right and it makes AMD return its honor.
Sometimes when you are in second place all you got to do is stick close enough to first place so that first place feels the pressure continuously. Almost surely first place will have a misstep sooner or later. Often winners aren’t used to adversity and have a problem learning how to get back up, run the race, and put the fall out of mind. Intel demonstrated this when they missed the tick for the first time. They didn’t get right back on track. They rolled around a bit giving others who had kept competitive a chance to set the pace. Intel was not used to chasing so it took them a bit to gather themselves, decide on a course of action, and then executing.
Good article, but some depth is missing. AMD acquiring (and overpaying for) ATI was Plan B for failed AMD-NVIDIA merger. At the same time, silicon bugs in ATI silicon (R200) and AMD silicon (65nm quad-core) caused issues with sales. The investment was not $700, $300 million etc, but rather an $8.6 billion investment into AMD and Chartered to create GlobalFoundries. Without that investment, AMD would not exist today (Chapter 11 was looming on five separate events, saved by selling tools from Dresden to Russians, which caused a lot of issues to GlobalFoundries, as the investor bought factories with less-than-correct tooling).
AMD overpaying for ATI just before the stock market crashed ate their cash reserves which would have been used to build new fabs. Hector Ruiz got sued for his shenanigans around that time including shady dealings with Intel Capital. Also, AMD couldn't switch to TSMC any faster because of the patent cross-license agreement they had with Intel stated that AMD couldn't outsource or license production to external fabs for x86 processors. AMD had to jump through a lot of legal loops to go around that.
It’s a 16 minute video. Something’s gotta give.
It's worth mentioning _why_ the acquisition of nVidia (not merger - AMD was far larger at the time) failed. It's because ego-maniac Jen-Hsun Huang insisted on becoming the CEO of the resulting company.
@@TrueThanny Jensen is not an egomaniac, he's an extremely capable conductor of an orchestra that went from producing gaming chips with register combiners (brilliantly branded pixel shaders) to a company with a market cap higher than AMD and Intel combined (or Intel and IBM combined). AMD had a disastrous CEO who robbed the company in 10 figure range (and was sacked when SEC came knocking), and a CFO that made deals - bad or good - which robbed company of another 1.3-1.7B circa. Jensen's remote relative, Lisa took over AMD and turned it into a beautiful story with almost 100B market cap.
Jensen - egomaniac? Sure, it can be categorized like that - I'd prefer the term "focused on survival in piranha club called silicon valley, when every VC says you will fail because you're in hardware." But certainly not a... thief (like the two honchos, with a remote stubborn engineer becoming a follow-on CEO that decided twice to ship buggy hardware - and redirected funds from tools for the Fabs into bonuses for executives...)?
@@TrueThanny huang is a super villain. He will be slain some day.
Thanks for summarizing and analyzing all this semiconductor industry history! I've tried to follow the history of AMD in the past, but your video helped shine some light and improve my understanding of a lot of events and details. Keep up the great work on these videos! 🙂
TSMC is the future of all man kind
AMD was such a poorly run company before Lisa Su.
Before even watching I can say that GF decided to drop 7nm node development for the reason it would not be economically viable (they were not profitable at that time). AMD needed more advanced process for future processors, since they can not get stuck on 12nm for good so they went to TSMC.
It might be inconvenient to consider that prior to the first Quad-core release, AMD led Intel with native Dual-Core release and copper interconnects. Additionally, just because Quad-Core are not on the open market does not mean that they are not being manufactured in high volumes and sold directly to specific customers. This is a common misunderstanding.
There is an additional Cleanroom in Dresden since those photos were taken. The current Wikipedia photos do show this. However all the nm fun and games were occurring in Upstate NY.
There are many segments in the semiconductor industry and they don’t all benefit from being on the leading technology even if they could afford it and there was enough capacity. Some are concerned with low power and low temperature and have little need for high performance. It is not a question of “just good enough” at all.
The Tick-Tock slide included Intel’s Itanium massive misstep, which basically created the gap in the market that AMD’s x86-64 dutifully filled. Without this strategic failure on Intel’s part, AMD would probably not have survived to purchase ATI or for Dr. Su to take the reigns.
At its heart, AMD's multi-chiplet architecture shows that not everything needs to be at the leading edge. And you can benefit from using the technology that best fits the task, rather than trying shoehorn everything onto the leading technology and it is certainly not settling for "just good enough".
What's not being mentioned in the tech news is that there are many highly skilled Taiwanese working in the tech industry, both in Taiwan and the US, and they are very patriotic and willing to help each other out. It's simply natural that Lisa Su, who is Taiwanese, would gravitate towards TSMC, which essentially are her "homies" back home, instead of Samsung due to language and culture.
tsmc had the best process
Yes, essentially the founder of NVIDIA is also Taiwanese
Samsung nodes is inferior to tsmc nodes
I guess Zen2 (rome) was planned to be TSMC right from the start - I still remember the reports about AMD having working 7nm silicon in the labs, and this was in April 2018. There is no way this was done in response to GF announcing halting 7nm, keeping in mind it takes months-year to port a designt of this complexity to another progress. Either AMD developerd the part for both GF and TSMC (unlikely), or they bet GF would be unable to provide a high yielding 7nm process at all (or so late it wouldn't matter).
i worked at AMD at the time and GF 7 was originally there for trial. it died so really on that they didn't decide which one to use before GF dropped out of the race.
@@slave2234 wasn't the Apu part delayed quite a bit? Was the CPU-only chiplet planned for TSMC 7nm and the APU gf7?
3:48 *Nice.*
You need to do a video about Tesla and their chip manufacturer
Tesla buys both off the shelf (AMD, among others) and from TSMC. The system on Wafer (25x Dojo+RAM+interconnects).
What a crazy bright idea of selecting a Engineer to head a Engineering works instead of big school MBA's and letting them do there work without "correct corporate lingo practiceses".. who can thought that??
As a CEO, much of your job is dealing with politics and client relations. It's the highest level of sales and marketing, 24/7.
A good engineer makes a poor CEO.
A good businessman makes a mediocre CEO.
You need someone with both backgrounds.
I knew she was promoted from within but didn't know Lisa Su was only 44 when she took the wheel. It seems like, and this is a laymen's take, that an engineer CEO who cut their teeth during the 90s boom is a big advantage.
Taiwan is a high tech country which is really impressing
Well explained, great job. Good luck to all of them.
It is Taiwanese connection. AMD ceo is a Taiwanese and tsmc is owned by Taiwanese. Business connection and partnership by cultural affinity and country origins.
Thank you for a wonderful sum-up of a long situation.
AMD is a standard US company saved by Taiwanese, Lisa Su & TSMC.
By that logic, TSMC and AMD were saved by the european companies and their UV tech that allowed them to make chips under the 10nm process.
Emm... Great video but anyone noticed the high frequency noise in the background?
Cheers bro, awesome video... Always great content, neutral and it's just so informative.
Thank you
Thank you for such a thorough summary on the AMD and semiconductor industry. I was wondering if you know the reasons why Intel fell behind TSMC and Samsung? Where there many components as to Intel falling from grace? Thanks.
just my 2 cents. Intel had been ahead of the curve in process node for decades until 10nm. The "official story" from intel was they were too ambitious with the node. In 14nm, they targeted a 2.4x transistor shrink instead of a typical 2x. They did pull it off after overcoming some tough challenges and it went on to become one of intel's most successful nodes. Perhaps due to the momentum and confidence they gain at 14, they targeted a even more ambitious 2.7x shrink for 10nm. This time the technical challenges were way bigger than 14nm. Without a backup plan, they delayed the roll out of 10nm from 2016 all the way to 2020. It seems Intel's ambitious could not be realized on the last gen lithography tools, as ASML CEO once said that the 10 nm delay could have been avoided had EUV machine been available to them.
wrong bets by intel
The 6:52 article highlights an interesting detail. The Llano APU featured a custom CPU design that was specific to the GlobalFoundries process. So it could not be made by TSMC, even if AMD and GlobalFoundries agreed to do so.
In contrast, the Ontario/Bobcat APU could be re-synthesized to be made on a different process. A Bobcat-family CPU was used in the PS4/XBoxOne APUs, and so they could be made by both TSMC and GlobalFoundries.
Ps4 and xbox one apus used jaguar cpu cores and gcn 2nd gen gpu cores
@@DaoistYeashikAli
Jaguar was the core, bobcat was the die.
In semiconductors, "just good enough" usually means "just right". The world of electronics extends far beyond phones and PCs.
the just good enough sector usually can’t utilize leading edge node because they need the reliability and durability factor that leading edge node will be too costly to offer
@@lansiman Yeah, I want my life and death chips made with large enough transistors to not get spooked by quantum glitches.
Yes which is why during the pandemic We were considered essential workers. Some of our products went into respirators and other medical equipment.
Worked with AMD in Penang
Respect the effort that you have put it!🙏
many new insight i got today.
and i just read that Intel now eyeing GloFo
wow, just wow, it's so dynamic!
I have worked for the company before AMD (ATI).
AMD is still using the same office in the city that I am living in.
Thornhill? I used to pass by daily.
@@gdfath3r yes
In case you haven't noticed. This video is flagged "Unlisted".. that's an oversight, isn't it?
It’s because it’s in Early Access. It will get released in a month or two.
Thanks for video keep going 🤠 greeting from Morocco*
Did they leave? Isn't the I/O die still made at Global Foundaries?
Noted in video how chiplet main chip is TSMC, side chips are GF.
Excellent job on these videos. I’m in. Hitting the subscribe button now.
How do you find the time to research and present such detailed analyses of complex issues? You do it very very well.
Sometimes, people just follow the industry.
TSMC is an exceptional company.
In 2017, TSMC provided the silicon for Oracle’s SPARC M8 CPU chip: 5 GHz, 64 bit, 32 core, 4 instructions per clock cycle.
It really screams, was the fastest general purpose chip for over a half decade, some suspect it surpasses & outperforms all current CPU’s on encrypted workloads even today moving into 2024!
TSMC has their stuff together.
World 🗺️ Class!
2021 TSMC is the best chip maker right now on earth, and it’s 10 years ahead others. They are working on 1 nm right now. In this section, Taiwan is # 1.
Everyone going to TSMC, even it's competitor at one time including AMD, Intel and Apple.
@@hemant3332 currently the latest AMD and Apple CPU are equipped the I/O module that is designed by TSMC. TSMC gave them for free and is faster than original design.
thank you for great videos, i posted to reddit
This is great stuff..but is it a video essay on Asia?
good information, bud
What would you say about thr new deal with gobal foundaries
Kudos to your correct pronunciation of Llano. Respect.
Points removed for pronunciation of niche :P
Is there any reason you're still uploading in 720p?
just listen, this is not a cartoon show.
@@abdiganiaden what is that supposed to mean?
Your vids are interesting. Well documented
Love this channel.
At 16:03, who are the remaining ones, and which one doesn't do work for other customers? My guess is TSMC who work for other companies, Samsung, and Intel. I just don't know which of those last two work for other companies?
Intel doesn't fab for others
Highly informative 🙏
the race to lower ¨nanometer measure¨ is really stressful to even listen to. LOL
But great for the consumers and for the tech itself.
More importantly, as usual, presented in a very interesting manner.
the cutting edge silicon industry is really interesting
Will you do an analysis on the impact of the new IBM 2nm chip
IBM will not produce any 2nm products. That 2nm chips were only for research and demonstration purposes for technologies that they'll lisence to other companies.
Its just like the one they published back in the day for 7nm too. Nothing much came of it.
@@tobiassteindl2308 tsmc already doing the 2nm in Taiwan. Tainan City.
@@veryinterestingpersonaliti8321 you mean researching. Also the nm companies use are pure marketing. We don't know if the IBM 2nm is better or worse than TSMCs.
Very informative, thank you so much
I really hope Globalfoundries does an IPO and competes again on 3nm. I bet they regret not going for it with current monster profits being made for 5nm.
Doubt it, they licensed their 14nm from Samsung, so their last own tech is 10 yo. If they are lucky they may license 7nm when it is obsolete in few years.
but also the monster R&D costs and risks
Mature technology based chips are in high demand. The profit is very good without billions of investment.
Nice video.
Great video. Keep up the good work.
Great job 👍 Thank you
Good video mate!
good one. i have nothing to add or correct here. no bias either. it's just sad that the industry stagnated because of Intel, where Intel should be on equal footing or slightly ahead of TSMC today
Wait, so Huawei gets priority over AMD? I thought Huawei were banned from doing service with TSMC?
Thank you for another excellent video!
Thank you.
AMD , TSMC. Global Foundries .... major share holder..... the same.... Mubadala. All companies are run from Abu Dhabi.
Thanks bro for great info
Funny how Intel is having the same problems that AMD had
the irony... the increasing demands for RandD and to keep pace has resulted in essentially 2 alliances right now
Samsung,Intel,IBM are one group --- TSMC,AMD is basically the other at the moment
I don't know much about this kind off business but just can't stop wondering how big a deal was dr Lisa Su for AMD. Looks like under she's control things start finally working.
Dr. Lisa su was no Steve Jobs but was one of the best.
@@seapirate5323 🤣
@@seapirate5323 Steve Jobs was no Steve Wozniak, so good thing Dr. Su is not a Jobs.
@@kiseitai2 Steve Jobs is one of the greatest engineer the world has seen,
He gave us 3 important things and the world didn't ever stop to not using those.
He was a visionary. You may disagree all you want but Jobs was a legend.
As for Dr Lisa su she didn't invent anything new but saved a company from going bankrupt for 4 years. I wouldn't say that an easy task.
Lisa Su is more like Tim Cook. All they needed was someone who is competent and didn’t worry about stock manipulation which was what intel was doing for a whole decade.
So the most advanced foundries are now in East Asia. We see if Intel's Alder Lake will finally start catching up to AMD/TMSC. How the turntables.
Awesome content about semiconductors!!
Alot of speculation in this video, for example what is your source that 12nm came from Samsung?
In the video, showed the presentation of the 14nm node joint venture, which included the evolution to 12nm.
The issue for GR, as per their February 2024 announcement, customers are defecting due to their needs for sub 10nm (7nm, 5nm, 4nm) far faster than expected.
10:17 This is where the training montage and swelling music begin.
Pretty sure global foundries was older amd fabs
What is AMD fab 36 LLC?
Thanks for the great stuff!
This is excellent
the CEO of AMD and Nvidia are taiwanese descent. maybe it's also a factor why they often chose Taiwan smc
Intel opened up their foundry business to other customers now. Maybe we will see AMD chips being fabbed at Intel foundries?
I could see it may open the can of legal troubles,
a sponsored customer may strike legal trouble by claiming that intel's foundries 'illegally copying' said customer's trade-secret designs/IP,
while intel may win in legal court by properly showing the receipts it could very likely costly for them in term of other customer's trust.
@@rashidisw Intel would need to reverse engineer the design back up to a high-level to understand what the different circuits are doing. Sounds like an immense amount of very specialised work. It's probably simpler to innovate.
@@SerBallister simpler to innovate? well, the already defunct Cyrix have completely different opinion.
@@rashidisw Reverse engineering a billion+ transistor circuit into discrete features that you can copy and implement into your existing chips is extremely difficult. Cyrix had spec sheets to make their x86 clone (from scratch), they didn't copy circuits.
It doesn't make sense for amd to use intel foundries ,amd need leading edge node that intel cant provide
Also remember amd also have incentive to use tsmc node
Not every chip needs the newest, smallest process. Lots of business for chips from process nodes 1-3 nodes behind the state of the art.
zen was mutidie from the begining they were never monothithic as that was part of why its design was so brilliant the diffrence was before they had the io with the rest of the cpu and with chiplets they seperated the io as being less importand left it at 14nm and fabricated the pure cpu at 7
Never lock yourself into a single supplier.
Strategy for noobz.
Apple tried to use both TSMC and Samsung and they failed. Consumers complained when they bought Samsung processor iPhones. It runs slower, hotter and consumes more power.
@@mlai2546 that's funny, because if anything happened to TSMC, Apple would not be able to make phones.
@@tensevo They can go to Samsung or Intel.
@@mlai2546 That's my point! You always want to keep suppliers on your side, in case you need them. Do you think Samsung would just drop everything they were working on, to make chips for Apple? Supplier relations are as important as employee relations. If your product depends on one supplier, they can and will hold you to ransom. Thus TSMC make more money than God.
@@tensevo I agree with you that it is better to have multiple supplier, however TSMC is not known to hold companies for ransom. If they did, they wouldn't be largest in the world. Just look at their profit vs Intel. Intel is the worst at overcharging customers.
The other issue with Apple outsourcing to Samsung or Intel is that both make same product. Do you remember when Apple was using Samsung fab? Well at the time, Samsung Exynos was very competitive and was leading both Qualcom and Apple. I am sure there is some sort of copying or technology stealing there. Since Apple has moved to TSMC, Exynos processors has dropped to 4th place. Even slower than Huawei's Kirin.
Also I don't think Apple will go to Intel as Intel makes computer processors too. They are basically rivals now. Apple going to Intel is like giving away their design to a competitor.
Another thing. GlobalFoundries bought IBM fabs in 2014 or 2015 I think. They got 22nm from IBM but AMD never used IBMs 22nm on the CPUs and GPUs. Wonder why because before Zen, AMD was stuck at 32nm from 2010 to 2016
I'm there and you are not quite correct. Just so you know 14 NM was being successfully produced. Can't go further into detail.
What a great and informative video
I subscribed
Couldn't AMD just have not signed the wafer supply agreement? Why did they do it?
i always thought amd procuded chips for intel in the beginning ;)
@@domainmojo2162 even in the 90s they did
When do you think that Gloflo will restart EUV/7nm development?
Almost never. As mentioned in the video, GF gave up 7nm development due to (1) high investment (2) lack of 7nm customers. No.2 reason is more than No.1. Even if GF suddenly got dozens of billion dollars and genius engineers to make their 7nm working, there's no way the cost is competitive to TSMC: TSMC's 7nm is on mass production since 2018 and all the cost spending on 7nm has been paid by Apple/Nvidia/AMD/Qualcomm/MediaTek's orders. Even if US government want some US' 7nm capability to secure military's use, they would invest Intel (TSMC's 7nm~Intel's 10nm) but not GF.
never, they would only do so if the market would be so gigantic that investments woul make sense. Well thats not the case.
The only other solution would be if another company would share their knowledge with them, but even then Gloflo would have to pay a high price for a market which is not that big for Gloflo to even begin with.
So: never.
If they make 7nm products, it will probably be with a licensed process from TSMC. At least we know it's not necessary to use EUV for that node. TSMC doesn't used it at all with N7.
Whether or not that ever becomes a sensible thing to do will depend on the market. There are still plenty of use cases where anything lower than 12nm is pointless.
as soon as 2023 maybe. Saw a forecast of TSMC mass producing 3nm in 2023 and moving all the top customers A A G N X etc there, it will pressure second tier foundries to evolve towards more advanced nodes.
@@GoogleUser-ee8ro wrong, there is a lot of money to be made with 12/14nm or higher nodes, no need for Gloflo to go lower if they dont want to
Lis Su
We Love You!
Awesome!
Hello! Recent subscriber here. I remember one of your videos you recommended a book that on China but I can’t remember which one. Could you please recommend some of your favorite books about China, particularly modern China and it’s evolution culturally and economically
Qualcomm will follow soon.
You said "next year" in the script, it would be better to say the actual year. Now, 11 months later, when I watch it, I don't know if that means 2022 or 2023.
So when China invades Taiwan and all these expensive fabs get destroyed...good for those who have other options then.