This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
And, most critically, time. EUV took decades to get to the state where it could barely be called working. Remember the mythical man-month? China's supply of STEM grads is sufficient to undertake nation-state scientific giga projects, yes, but they will be rediscovering the required physics and engineering for years. And until they figure it out, they're stuck at quad-patterning DUV yields (which aren't great below 14nm), or even worse yields on even more multipatterning. And that means more subsidies the government has to pay to top chip manufacturers to produce otherwise unprofitable chips. I know you linked to a guy who says they got the yields up above 50% and they can probably improve it, and that this could be profitable if their devices sold for comparable prices to next-gen TSMC tech, but this comes at considerable cost to power efficiency, places intense demands on the fab to customize process to fit design issues, and sucks money and time away from longer-term research to pivot away from DUV. They would spend the next ten years on the same trajectory as Intel 14nm, where they keep squeezing minor miracles out of the chip design in exchange for increasing power consumption and overspecialization. They'd keep getting tiny improvements, for years; meanwhile, Western EUV will also be improving, yields will also be slowly catching up, and costs will start falling off as the EUV nodes mature (look at TSMC wafer prices for 7nm at start vs today). The whole reason the major players are transitioning to EUV now is because they all recognize that, long-term, the future of DUV is unsustainable, and EUV is going to be less costly. Chinese fabs also know this. If China manages to stay competitive against EUV, they will be required to pay dearly for it. Lay-journo interpretations notwithstanding, this is, and has always been, the aim of sanctions - to penalize and slow access to cutting-edge technology.
I suspect that an invasion of Taiwan in which the outcome appears dire would result in the EUV machines being rendered into indistinguishable welded slag and scraps, at least in the molten tin UV source. Several hundred billion dollars and decades of research and development is tied up in the design and manufacturing of these machines. The critical trade secrets are going to be unrecognizably destroyed. If TSMC doesn't destroy their machinery, and ASML can't, the US will, covertly if possible, overtly if unavoidable. China will eventually get access to domestic EUV with enough will and investment; but without intact examples of working EUV machines to study, they're stuck on the long, slow grind of figuring it out themselves. How many years of progress will be made in western chip fabrication while the Chinese are busy re-solving EUV?
They do industrial espionage, they're very rich and can pay people who developed these processes for ASML, TSMC and so on more than they would get paid in the West.
Yeah, and imagine if they didn't manage to get these great chips, just built some more nuclear power and bigger datacenters? What then? I mean, assuming starting fake companies in the neutral countries and straw purchasing Yankee Haram compute that way wasn't an option.
In spite of their best efforts, they still have no domestic EUV industry to speak of. Contrast with the domestic advanced semiconductor industry they do have. Clearly one of these is harder to replicate than the other.
They do pay well, provided you don't mind living in a country where the government can and will disappear or execute you for wrongthink, regardless of your station or importance to the technology roadmap. Many such cases!
Again, no one can stop them with sufficient motivation and willpower. But it can certainly be made to take a while and cost a boatload. As HPC systems get bigger, there are topological and latency challenges that compound on each other. And they will have to keep getting bigger, slower, and less efficient to keep up with the pace of innovation in equivalent western systems. It's not an insurmountable obstacle, but it costs resources and time. ETA: More power also means more cooling required. More power density in less efficient devices places huge demands on cooling, or forces more lateral scaling, which also compounds the problems above.
I'm assuming that the big concern with high-volume Chinese HPC is them achieving AI dominance and outpacing western innovation long-term - correct me if I'm wrong. I think, for very large AI, the software, and particularly the training techniques, have orders of magnitude of improvements awaiting discovery and implementation. This could go either way, western or Chinese advantage, with the understanding that Chinese industrial espionage is highly effective at extracting and implementing mathematical innovations in software, substantially moreso than hardware. I think it's more likely that the high costs of hardware force China to be parsimonious with their compute resources and ultra-efficient with their software craftsmanship, than that they just steamroll through the obstacles and disregard the expense.
More options
Context Copy link
More options
Context Copy link
More options
Context Copy link