Follow Slashdot stories on Twitter

 


Forgot your password?
Close
typodupeerror
×
Technology

TSMC Starts Volume Production of Most Advanced Chips in Taiwan (reuters.com) 58

Posted by msmash from the moving-forward dept.
Chipmaker Taiwan Semiconductor Manufacturing Company began mass production of its most advanced chips in southern Taiwan on Thursday and the company's chairman said it would continue to expand capacity on the island. From a report: The long-awaited mass production of chips with 3-nanometre technology comes as attention focuses on the world's largest contract chipmaker's investment plans at home and abroad. TSMC has a dominant position as a maker of advanced chips used in technology from cellphones to fighter jets. "TSMC is maintaining its technology leadership while investing significantly in Taiwan, continuing to invest and prosper with the environment," TSMC Chairman Mark Liu told a ceremony marking the production and capacity expansion in the southern city of Tainan. Liu said demand for the firm's 3-nanometre chip was "very strong", driven by new technologies including 5G and high-performance computing products. He did not elaborate.
This discussion has been archived. No new comments can be posted.

TSMC Starts Volume Production of Most Advanced Chips in Taiwan More

TSMC Starts Volume Production of Most Advanced Chips in Taiwan

Comments Filter:

  • I suppose they'll wait a bit more, while TSMC shakes it down a little bit more, before going after it.

    • Re:I wondered why China hadn't invaded yet. (Score:4, Insightful)

      by ShanghaiBill ( 739463 ) on Thursday December 29, 2022 @11:06AM (#63165764)

      Did you see the Omaha Beach scene in "Saving Private Ryan?"

      That was a landing with total air supremacy, across a narrow strait, opposed by a tiny fraction of the German Army, with a deception plan that had convinced the enemy the main blow would fall at Calais, and with many times the amphibious shipping capacity that China has today.

      China may be able to invade Taiwan ten years from now, but they do not have the capability today.

      • China can invade now. Holding Taiwan with the infrastructure intact is the problem. If infrastructure is damaged, China is unlikely to get much help rebuilding it.

    • Re:I wondered why China hadn't invaded yet. (Score:5, Insightful)

      by caseih ( 160668 ) on Thursday December 29, 2022 @11:54AM (#63165862)

      I think even the CCP has to know that once they invade Taiwan, TSMC facilities will end up in ruins and they won't get as much out of it as they hope for. Even if they can capture the facilities intact, the people who actually have the knowledge of how it works will be killed. Even they are persuaded to work for new masters, it will still be a huge step backwards and much harder than they think. And with Taiwan preparing to put up resistance, there's no way they can capture Taiwan and have the infrastructure intact, which TSMC depends on.

      That said, I have no doubt China is going to invade relatively soon, probably in the next year. If Russia had not faltered so badly in Putin's war on Ukraine, they would have invaded months ago. I think Xi's ego will be the primary motivation behind it, because Xi will certainly get Taiwan, but it will be largely in ruins. This will devastate our economies. If you think chip shortages were bad in the last couple of years, that's nothing compared to what is coming.

      • Re:I wondered why China hadn't invaded yet. (Score:4, Insightful)

        by jacks smirking reven ( 909048 ) on Thursday December 29, 2022 @12:30PM (#63165916)

        I think if anything the Russia situation has reason to give China even more pause about Taiwan. The conflict has functionally moved Russia from being known as the "second most powerful military" down a couple pegs at least and while China could be regarded as #2 now all it took for Russia was to actually engage in a conflict with western/NATO style opposition (even if indirectly) to show their deficiencies.

        China invading Taiwan will no doubt bring an actual military coalition to it's defense with not just Taiwan but the US, Japan, Korea and really any non-China aligned Asian state (almost all of them really). They would have no choice but ti intervene on Taiwans defense and looking at the trouble Russia has had in Ukraine, a country with a huge physical border compared to invading a heavily defended island, they have to know it will be a slog and the chance of losing is high. Taiwan does not need to "beat" China, only draw them to a stalemate and it's a loss.

        Also for as much as Xi has secured control of his position I don't believe CHina quite operates from a singular position as Russia does. Putin has (from what I have read) much more autocratic control of things whereas Xi, even as he stands now, is still more vulnerable to internal party politics.

        Stack on top of that the seeming reality that China will be still working through repurcussions of their huge covid outbreak for the next year and you have the makings of a disaster if a planned invasion does not go absolutely plan perfect. A huge, unecessary risk that puts China's last 30 years of economic progress on the line. Monumentally stupid decision if they make it.

      • Re:I wondered why China hadn't invaded yet. (Score:4, Insightful)

        by LostInTaiwan ( 837924 ) on Thursday December 29, 2022 @03:38PM (#63166368)

        A successful invasion of Taiwan is nothing more than a Pyrrhic victory. TSMC is more than just the physical assets. The secret sauce is in the people, the institutional knowledge, the software, the recipes and parameters. China has poured hundreds of billions and poached thousands of talents into its own semiconductor industry over the last couple decades and is still no where close to TSMC, Samsung, or Intel. I doubt if an invasion will be of any help to China, other than to knock off a competitor.

        The scale and impact of subsequent sanctions after an invasion of Taiwan will be far greater than that imposed on Russia. The free capitalistic world is not going to forget the country that crashed the world economy, destroyed a vital component of the world's integrated high tech ecosystem. Oh' and those Uyghurs, imprisoned Hong Kong activists, dead Taiwanese in the rubles of their own homes as a result of the invasion are sanction enhancements.

        Today's Chinese people are not the Mao's era Chinese. Most Chinese people of today know what the outside world is like and enjoy a certain living standard that is not possible without global trade. A crippling sanction will take all of that away and very soon the Beijing government will be facing a vote of no confidence in the streets of China.

      • It would be a pyrrhic victory.

        China can easily destroy Taiwan. But it cannot take Taiwan intact.

        Taiwan is heavily armed. Not nuclear that I know of, but heavily armed with conventional missiles and artillery capable of 200 mile range -across a hundred mile straight. This puts the coast of China, including several major cities, within range. The retaliatory strikes from Taiwan would be very costly to China.

        It is the non-nuclear version of MAD.

      • Attack when your advantage is slipping (Score:4, Informative)

        by FeelGood314 ( 2516288 ) on Thursday December 29, 2022 @05:30PM (#63166582)
        China probably can't successfully invade Taiwan right now but they might still try. The distance from China to Taiwan is 160km (for comparison Dover to Calais is only 33km). Amphibious assault ships in WWII travelled at 9 km/h. If China's ships travel at the same speed that would mean being in in range of Taiwanese defenses for 18 hours. Even if the landing craft traveled at twice that speed they would take 9 hours. Next the Taiwanese west coast is such that there are only a few suitable beaches to land on. So best case China loses 90% of the assault force crossing to Taiwan, effectively their entire navy and 75% of their aircraft trying to protect the landing ships. They then lose half the remaining force on the beaches. At this point the two sides would be about equal. China would have destroyed every runway in Taiwan so they would have air superiority but Taiwanese surface to air missiles would reduce the effectiveness of any ground support by Chinese aircraft. There is also no conceivable way China could resupply their beachhead. Taiwan could just sit back and wait for the Americans come to their aid.

        However, here is why now is the most dangerous time, if China doesn't invade in the next 5 years their chances of success will decrease. Taiwan was spending 2% of GDP on defense and most of that not very effectively. If they increase it to 5% and increase the conscription length then China's chances of success fall drastically. Next, drones favour those who can hide and drones are only getting better. An amphibious attacker is about as unhidden as you can get. Finally, China's GDP growth will likely fall behind that of the USA soon if it hasn't already. China's population is growing slower than the USA and China's median age is now older than America (38.4 vs 38.1). With every year that passes China's chance of a successful invasion decreases.

  • The long-awaited mass production of chips with 3-nanometre technology

    Look, can we just use the process names? Because while the process names are designed to be confusing, referring to them in nanometer sizes as if the marketing were relevant is even moreso. I'd go so far as to say that if they're not using the process names, then the article should not be posted here at all as it was clearly written by and for chucklefucks.

    • Re: (Score:1)

      by CityZen ( 464761 )

      Everywhere you see the term "nanometer", just replace it with "nonsense-meter", which describes the process more accurately.
      Other possible replacements are:
      - bogo-meter
      - nanu-manu
      - nahna-mahna

  • But performance of computers is limited by the considerably greater distances on motherboards. 3nm won't help as much as you might think. Instead of shaving off a nm here and a nm there, a larger die size could trim entire decimetres off distances.

    • Nanometer doesn't mean jack shit anymore, it's a marketing term not an actual measurement.

      • Re: Very impressive (Score:4, Insightful)

        by timeOday ( 582209 ) on Thursday December 29, 2022 @12:59PM (#63165990)
        Even so I would bet TSMC's "3-nm" process does have the highest transistor density and lowest power consumption for e.g. the 5g chips they are producing. TSMC does know what they're doing, and didn't invest in this new fab without actual advancement.
        • Not necessarily, there've been several instances where the older Intel node was denser than the newer TSMC node, but it wasn't practically usable for things like general purpose CPUs due to thermal concerns, fine for certain types of memory though. Intel ain't really behind TSMC like a lot of the online and investor crowd made them out to be. Technology node comparison is a lot trickier than most folks make it out to be.

          • Not necessarily

            Yes necessarily. This isn't some theoretical discussion we're having. This is an actual fab process with actual numbers you can look up right now:

            TSMC N3: 220 MTr/mm^2
            TSMC N4: 146.5 MTr/mm^2

            Intel ain't really behind TSMC like a lot of the online and investor crowd made them out to be.

            TSMC has the 3nm node in production. Intel hope to start production at the end 2023. TSMC plans to have its second iteration N3E in production by then. At this stage it looks like TSMC may even get N3E out the door before Intel finishes its Intel 4 node and scales that to full production. It's a node that is beaten in *a

            • Yes necessarily. This isn't some theoretical discussion we're having. This is an actual fab process with actual numbers you can look up right now:

              And yet here you are, quoting theoretical numbers. Funny how that works, isn't it? Those densities can only be achieved on highly repetitive structures (aka memory). In reality you cannot achieve that density on actual designs due to yield, thermal and timing concerns. They have little to no meaning for actual general purpose logic, so unless if you're planning to make memory chips you're kind of out of luck.

              TSMC has the 3nm node in production. Intel hope to start production at the end 2023. TSMC plans to have its second iteration N3E in production by then. At this stage it looks like TSMC may even get N3E out the door before Intel finishes its Intel 4 node and scales that to full production. It's a node that is beaten in *all* metrics by TSMC's N3 node.

              Yes Intel is behind. Intel's CEO even acknowledge they are behind and announced a strategy change to catch up which they hope to do before 2025.

              And again, you are blindly staring at process nodes and PR numbers without understanding how to actu

    • TSMC and any chip fab can make larger die sizes anytime they want. For example the latest Threadripper. Yields and cost is the main reason their customers do not have huge chips. It is not so much a technical problem as a cost issue.
      • there's a physical limit to die size: the masks are 6'' square, with a usable of around 5''x4'', and all that is shrunk 4x onto the wafer. That's your maxiumum die you can print, around an inch in size, no more. Of course, fitting more smaller dies onto said mask is advantageous in terms of yield.

        • there's a physical limit to die size: the masks are 6'' square, with a usable of around 5''x4'', and all that is shrunk 4x onto the wafer. That's your maxiumum die you can print, around an inch in size, no more. Of course, fitting more smaller dies onto said mask is advantageous in terms of yield.

          My point was not chip fabs can make any sized die. My point was that chip fabs can make larger dies any time they want including right now as evidenced by Threadripper. The limitation is yield and cost. The OP seemed to imply that smaller feature size would not be as much help in performance as larger dies. In the grand scheme of things, many of TSMC's customers will not be pursuing larger dies as a strategy to increase performance as larger dies are normally for server/workstation chips which start at thou

    • Re: (Score:1)

      by CityZen ( 464761 )

      This is why Apple has incorporated all the DRAM onto the same substrate as the CPU in the M1 and M2 SOC's.
      It's also why AMD has done the same with some of its recent GPUs.
      Too bad for upgrade-ability, though.

  • mass production of chips with 3-nanometre technology

    All such advanced facilities will be targeted by the enemy in the second or third wave of attacks. It'd be best to:

    1. Avoid publicizing the exact locations — and geo-tagged photographs — of the buildings.
    2. Have reliable bomb-shelters for the workers — within minutes of walking (running) distance from each workplace.
    3. Reinforce the most crucial parts of the structures — to survive a near-by (if not direct) hit.
    4. Have contingency plans to cont

    • Bombing semiconductor fabs makes no sense.

      Anything being fabbed will be weeks or months away from going into any weapon.

      Capturing the fabs intact would be a huge win for China. Their greatest fear is that they will be sabotaged beyond repair before they can be seized. To bomb the fabs themselves would be nuts.

      • Re: (Score:2)

        by caseih ( 160668 )

        Very little about war makes sense. Invasion for made-up reasons even less so. I've never understood this irrational need to control other people and force them to another's will.

      • So eliminating the sole purpose of an invasion makes no sense?

        Just threatening to put a few high explosive packages through the roof of the place would probably be enough to stop China in it's tracks.

        There is literally no purpose to taking Taiwan if they can't get the TSMC facilities and tech. Any political "victory" would be crushed under the weight of the rest of the world's response.

        • Re: (Score:2)

          by dgatwood ( 11270 )

          There is literally no purpose to taking Taiwan if they can't get the TSMC facilities and tech. Any political "victory" would be crushed under the weight of the rest of the world's response.

          And presumably China realizes this, and recognizes that their best bet is to infiltrate, rather than invade. They can gain far more by letting Taiwan and TSMC continue operating normally, while exfiltrating information about their technology through surreptitious means.

          • Exactly.

            Why invade and risk burning the place down and having a humanitarian crisis on your hands at the same time as global rejection, when you can just bypass all that by stealing the tech?

      • Re: (Score:2)

        by mi ( 197448 )

        If they don't prevail in the first week or two, they'll try to hurt the victim economically. Destroying the most profitable firms will be part of the strategy...

    • Unless you build your chip fab 100 feet underground in a steel-reinforced concrete bunker, you're not going to be able to prevent critical damage to the facility if someone decides to drop a few 1200 pound JDAMs on it. And, it's very likely that Taiwan has JDAMs, since we're sending them to Ukraine right now, so it's not like they are NATO-only.

      And if you think that Taiwan wouldn't hit their own fabs in order to screw China over if they couldn't repel the invasion, you're crazy.

      • Why would Taiwan need to drop bombs on their own facility even if they wanted to destroy it? Send in the explosive experts and wire the place up so that one button push does as much (or as little) damage as you like. And the ability to do this quickly, in the unlikely situation that they even thought this a useful tactic, means that the mere threat of doing it is as good as actually doing it if you want to hold you own facility hostage.

        • Because you probably don't want to have live explosives just laying around in a multi-billion dollar facility "just in case" when you can deliver it within a few minutes from an aircraft and confirm the job got done?

          Would you want to go work in a facility that could accidentally have multiple explosions triggered without warning in order to self-destruct the facility in case of dictatorial invasion? Would you trust a system built by the lowest bidder to not have a control failure and crater the place accid

          • Notice that I said "the ability to do this quickly" - that is to wire the place up (since pushing button takes no time at all) - and thus they wouldn't need to means that the facility never has any explosives in it. And of course the enemy has no way of verifying if it was actually done or not until demolition happens.

            The threat of dropping bombs on your own plant also the same property of not needing to every actually be done, but you cited the fact that it was not a hardened facility as if that was instru

  • The TSMC fabs mean squat to China, zero, zip.
    It is Taiwan's proximity to China that matters.
    Taiwan is about as close to China as Cuba is to the US. If nuclear missiles were placed in Taiwan, there is dangerous first strike vulnerability; the missiles could hit China before a response could be launched. Of course the US would not do this, but the mere possibility is what geopolitics is all about.

Slashdot Top Deals

Never test for an error condition you don't know how to handle. -- Steinbach

Close