Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/root-scientific-causes-of-tsmcs-leadership-intel-and-samsungs-failure.14630/page-2
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Root scientific causes of TSMC's leadership / Intel and Samsung's failure

They are 12 hours. 996 was used in China has it worked? Smic? Stem can be an art and a creative process is involved. Companies do need creatives.

Taiwan is part of coalitions. They're better team players because they have to be. That a site from Arizona will be using a Taiwan process from sales at Taiwan says something.
 
Last edited by a moderator:
Working hours is not the key for TSMC Arizona fab.
Communication and leadership from different language and culture is the key to success.

That's why Morris Chang said a good manager in Taiwan will not necessary become a good manager in US.
 
More about 996
It’s unofficially official, for exempt employees. It doesn’t apply to hourly. It doesn’t violate any laws because exempts, are, you know, exempt from those laws.
Cafeterias stay open late, so it’s official enough that a meal is served in the evening specifically for 996 workers.
I think it’s important to discuss because it affects Americans too. It’s part of the reason fabs moved to Asia; a better deal from the perspective of the owners. There was recently an agreement among some nations for a global minimum corporate tax to limit the incentive for large enterprises to go tax venue shopping. But enterprises can still venue shop for the longest exempt work hours.
 
I have worked as a supplier to all three. There are very significant cultural differences between the 3 organizations. TSMC: thousands upon thousands of engineers and PhDs marching to the SAME drum. Their ability to achieve large scale coordination of purpose is unmatched. TSMC: wide open to new ideas, no matter how small, if they will advance the company's goals of performance and cost. I never had a problem getting a last minute appointment at TSMC fabs to discuss new ideas. True on cutting edge and true on 150 mm wafers! Intel: no air gets in. Most PhDs are picked 6 months after graduation with no industry experience as a rule. Intel: no interest in cost cutting or performance improvements if technology is "released". Intel: no older fabs generating cash to pay for everything else. Intel: focus on share buy-backs not technology. Samsung: very much like TSMC in terms of coordination but without a single focus. Making DRAM + foundry logic + own ASICS + mobile devices and have CEO in jail at same time, that's probably your explanation right there. The secret sauce: people. @benb Surely you mean Taiwanese contractors, not Chinese.
You nailed it, Taiwanese contractors!
 
Kintestsu has a hub in Portland and their big client is tsmc. It's 996 and 997 there. They're having supply crunch and labor problems as well.
 
I'd like to hear more about the 996 in a future podcast or articles. This sounds like something with significant implications on the industry.

My experience in non semi (but other high tech and defense) industries in the US is that something approaching 996 tends to be the expectation for executives but also often has health impacts for the workers involved at ages typical for those execs.
 
996 definitely doesn't apply to Korea. Korea's Labor Standards Act amended in 2018 allows for a maximum of 52 hours of work per week, down from 68 hours before. They're quite strict on it to the chagrin of many in the tech/IT/media industry, making it much harder for them to meet deadlines on projects and such, there's plenty of news covering the complaints of game developers, TV production sets and so on and so forth about the difficulties posed by the labor law.

As for contractors, Korean contractors are among the fastest in the world so that part is irrelevant vs Taiwan or China. When Taipei 101 was under construction, the TFCC, which owns the building, transferred the entire construction authority over to Samsung C&T Corporation to expedite the build out. Earlier this year, Samsung C&T was selected to build a new terminal at the Taoyuan International Airport, so no, the Koreans do not have a disadvantage in that regards.

On the other hand, labor cost in Korea is definitely higher than Taiwan. Minimum wage is substantially higher in Korea (roughly $9.25/hr) vs Taiwan (roughly $5.25/hr). In fact, minimum wage in Korea is higher than Japan (roughly $7.50/hr) and the United States ($7.25/hr). Just three days ago, there was a report released showing Korean college graduates starting salary at companies with 500 employees or more is 59% higher than the equivalent salaries at conglomerates in Japan.

Taxes are generally higher in Korea vs Taiwan and R&D deductions are lower in Korea vs Taiwan, military spending as percentage of GDP much higher in Korea vs Taiwan, and another factor to consider is, Korea's always had cheap electricity due to favorable policies towards nuclear energy, that all went to the crapper with Fukushima and President Moon's administration, causing huge losses for KEPCO and raising electricity bills across the board. Taiwan will face energy issues too with their decision to follow suit with phasing out nuclear energy, both countries have very limited land for solar and to meet renewable energy targets, have to spend heavily on offshore wind which is $$$.

But as someone else pointed out above, Korea has a wide variety of industries that are competitive on the world stage. Car manufacturing, ship building and locomotives, household appliances, media and pop culture, batteries, petrochemicals, food and pharmaceuticals, Korea has a strong presence in all these industries, whereas Taiwan, outside of TSMC and semiconductors, you have a few EMS supplying IT equipment with relatively low margins, Formosa Plastics in chemicals (also affiliated with Nanya), some banks/financial institutions, and Evergreen? In that sense, I fear Taiwan has too many eggs in one basket with TSMC. As much as I dislike TSMC having a near monopoly on leading edge processes and not having a viable competitive second source, maybe the fact they're dominant in semis will be a good enough reason for the rest of the world to rally behind them to stand up to China, which to me is more important in the big picture of things.
 
996 definitely doesn't apply to Korea. Korea's Labor Standards Act amended in 2018 allows for a maximum of 52 hours of work per week, down from 68 hours before. They're quite strict on it to the chagrin of many in the tech/IT/media industry, making it much harder for them to meet deadlines on projects and such, there's plenty of news covering the complaints of game developers, TV production sets and so on and so forth about the difficulties posed by the labor law.

As for contractors, Korean contractors are among the fastest in the world so that part is irrelevant vs Taiwan or China. When Taipei 101 was under construction, the TFCC, which owns the building, transferred the entire construction authority over to Samsung C&T Corporation to expedite the build out. Earlier this year, Samsung C&T was selected to build a new terminal at the Taoyuan International Airport, so no, the Koreans do not have a disadvantage in that regards.

On the other hand, labor cost in Korea is definitely higher than Taiwan. Minimum wage is substantially higher in Korea (roughly $9.25/hr) vs Taiwan (roughly $5.25/hr). In fact, minimum wage in Korea is higher than Japan (roughly $7.50/hr) and the United States ($7.25/hr). Just three days ago, there was a report released showing Korean college graduates starting salary at companies with 500 employees or more is 59% higher than the equivalent salaries at conglomerates in Japan.

Taxes are generally higher in Korea vs Taiwan and R&D deductions are lower in Korea vs Taiwan, military spending as percentage of GDP much higher in Korea vs Taiwan, and another factor to consider is, Korea's always had cheap electricity due to favorable policies towards nuclear energy, that all went to the crapper with Fukushima and President Moon's administration, causing huge losses for KEPCO and raising electricity bills across the board. Taiwan will face energy issues too with their decision to follow suit with phasing out nuclear energy, both countries have very limited land for solar and to meet renewable energy targets, have to spend heavily on offshore wind which is $$$.

But as someone else pointed out above, Korea has a wide variety of industries that are competitive on the world stage. Car manufacturing, ship building and locomotives, household appliances, media and pop culture, batteries, petrochemicals, food and pharmaceuticals, Korea has a strong presence in all these industries, whereas Taiwan, outside of TSMC and semiconductors, you have a few EMS supplying IT equipment with relatively low margins, Formosa Plastics in chemicals (also affiliated with Nanya), some banks/financial institutions, and Evergreen? In that sense, I fear Taiwan has too many eggs in one basket with TSMC. As much as I dislike TSMC having a near monopoly on leading edge processes and not having a viable competitive second source, maybe the fact they're dominant in semis will be a good enough reason for the rest of the world to rally behind them to stand up to China, which to me is more important in the big picture of things.
"I fear Taiwan has too many eggs in one basket with TSMC."
As a Taiwanese, I don't think we have enough resource to be successful in too many different industries.
 
If you for something more scientific I suggest the link between volumes and process control. TSMC has been consistently capable of aggregating enough volume of processed wafers in R&D and early ramp-up to make process control and yield feasible. Do not forget that variance goes as the root square of the sample size so the higher the volumes the better your process control becomes possible. Especially when you want to achieve things like less than one defect per trillion vias.
I would go to the point of postulating that Intel needs to become a foundry as it doesn't have anymore the volumes required to achieve and maintain a below 7nm process.
 
The glut will hit and tsmc as well as Samsung are in a better position. There is a lot of trash where I live but some good people will be hurt.
 
Chiming in after a couple of months, the 996 is fascinating. It reminds me of academia (in America). In academia, at least in scientific fields, including social sciences for this purpose only, 60 hours a week is normal. Anyone passionate about their work, their research, is working at least that much for most of their career. 50 hours is probably the minimum, and 70 hours might be common for some labs. The 996 is more, since it's 72 hours. More than 60 seems to be the cut-off for some academic researchers I know – beyond that they can't have a normal life that involves any other activity (I know several who teach at ballet studios or group exercise classes at gyms). There's also the issue of family.

I'm sure that 996 will not work for Americans in a corporate context. The mentality and context of academic scientists is very different. They're autonomous and self-guided. They own themselves and their work. They're not doing it for anyone else, certainly not a corporation. Americans won't be willing to work 72 hours a week as the default for their careers. Maybe for a year at an early startup that they have a big financial stake in with options or whatever, but not permanently, at a big fab that brings no such financial windfall. Americans aren't that loyal to or trusting of large employers anymore. They might never have been – 72 hours a week is pretty much insane. It means no real life outside of work, except maybe church on Sundays. That kind of work setup also requires a spouse who is not doing 996, else no one would be able to take care of children, or run the myriad errands of a modern family life. It seems like it's designed for Asian men with stay-at-home wives, and that's obviously not portable to the US in 2022. It's a non-starter.

I was thinking of something else recently. I think being in America is going to become a significant liability in the near future, if it hasn't become so already, for political reasons. American companies are letting leftists co-opt their workplaces to such an alarming degree that I think a lot of people will savor the prospect of getting away from the madness. We're clearly seeing a cult at this point, and it's such a weird cult that I'm not sure we could even explain it to executives in Singapore or Japan or pretty much any non-Western country. The indoctrination in crackpot pseudoscience (those slides that Coke was using were amazing: "be less White"), the obsession with race and permanent hyper-neurotic victim narratives, the assertions that their favored invisible forces are the most important factors in our world and decisively shape every outcome (every cult has its thing, its invisible force that explains everything around us), the incredible tension and stress leftists are creating for everyone else, the extremely predictable explosion in violent crime – it's like having Scientologists take over, and a lot of smart people would rather be somewhere else. Not to mention the de facto tax of the awful American legal system, what with the unjust and capricious lawsuits, patent predators, etc. All this BS is completely avoidable by just not being in the US. Even just moving to Mexico is a huge win to get away from the cult, the bogus lawsuits, and the patent predators (née trolls).

Back to the science of node development, it's amazing to me that nature is carved this way. I still don't understand how TSMC has done it. It's the same physics, the same laws. Even now, it's not at all clear that Intel will ever be able to achieve what TSMC already has circa December, 2021. That is, can Intel, or anyone else, actually produce something on par with the improved N5 or the imminent N3 with the same production ability/efficiency? From what I read, Intel's 10nm node has been extremely expensive and inefficient on the production side. Intel can hide costs more than TSMC since it's not selling it to anyone. I think it was also reported that their 10nm was so complex that it was unattractive to potential customers during their first aborted go at a foundry business (LG might have been the prospect). It just seems far short of what TSMC achieved. I don't believe in any of Intel's scheduled nodes, all the renamed stuff. There's something wrong there, and I don't understand what it is. They don't seem to be able to develop nodes in a smooth way that results in something unambiguously better than what came before. It's been so long since Intel offered anything like that. I guess 14nm was the last time, and it was weirdly not that great at first, with Broadwell. Haswell was actually pretty amazing in retrospect. Maybe one of the 14nm x-Lake chips amounts to a decent bump over Haswell, but the 10nm stuff has been treading water.

It's just so strange that TSMC can do it, but not anyone else. I'm grasping for some sort of R&D methodological breakthrough, a breakthrough in the science of delivering breakthroughs, but maybe the answer is more mundane. I'm curious about the nature of the cognitive work involved in node development, compared to other forms of engineering and innovation domains. Is it just chopping wood with thousands of 996 engineers? Can brilliance or genius have a lot of leverage like in software, design, etc.? I even wonder about the tools. TSMC works with Synopsis or Cadence on the end-user tools, but I wonder what tools node development might involve, and whether high performance computing is used to develop new nodes. It still seems like very few fields or teams ever think about HPC and how it might be exploited.
 
@BlueNode Japan was once the king of long hours. Look where that landed them. 996 or whatever can be beat through high productivity any time. Present AND contributing is the key. Fabs are fully automated in the end so hours worked do not have the impact that they would in other industries. The answer to the question you raise is covered in the beginning of this thread. There is no silver bullet, just a singular purpose reinforced throughout the organization. How did ASML develop EUV? Patiently, systematically over many years. It's the same with TSMC. Slowly but surely and consistently. Kind of like eating an elephant or moving a mountain. Having lots of cash helps (but then you can't do share buy-backs).

As to your other point, the USA has not been competitive in manufacturing for years and this is unlikely to change ever. Consider: US$ as a reserve currency, the medical complex, the industrial-military complex and Wall Street. Add to that NIMBO and a few other quirks (legal, political) and you are in the present. We are an excellent place to generate ideas, software products, research, medicines etc. For lower margin products, we can't cut it (see bicycles, textile, etc.). It's neither good or bad, it is what it is. Diversity initiatives may well prove a disaster in the end but I would say they have little or nothing to do with the difference between Intel and TSMC.
 
I'd put it even more succinctly, Intel, overhead cost outpacing the other two and setting unreasonable goals whilst being run by MBAs. Samsung, their business too diversified with R&D and capex more focused on memory. TSMC, singular focus on foundry and foundry only with reasonable lock in step node improvements.

Foundry semi is all TSMC does, they can't screw it up. Not the case for either Intel or Samsung. Intel is nearly a pure semi company but they are heavily vested into design. Samsung...semi is just one of many many things they do. Work culture isn't all too different between TSMC and Samsung, 996 isn't applicable in Taiwan or Korea due to labor laws, cost overhead is similar between the two vs Intel, and Morris Chang would still identify Samsung as the one for them to worry about. Intel has excellent R&D but without a singular focus on HVM and commercialization.
 
Back
Top