FWIW I was a hardcore, dyed-in-the-wool CarPlay user before buying a Tesla. But after using the Tesla touch screen and voice controls for five minutes I've never missed CarPlay. If you haven't tried the Tesla touch screen I'd encourage you to give it a shot. It is, by a million miles, the fastest, most usable, well laid out infotainment system by a long shot. It's comparable to the performance you get from your smartphone. You really cannot compare it to the garbage infotainment systems in every (literally) other brand.

Have you seen Shin Godzilla yet? That was fantastic. It was like they beat Godzilla with collectivism and Japanese government bureaucracy. It was great!

I say this as someone who just bought an iPhone 15 Pro to replace his iPhone 12 (which replaced an iphone which replaced an iphone etc back to the iPhone 4) but let's not forget Apple's engineering teams also built a phone where literally holding it affected the antenna signal quality. They are not immune from poor design decisions.

But I agree with your conclusion, I assume it's just a software thing they will fix. Add some additional throttling while charging, whatever, I don't know. Surprised they didn't find this in their testing and address it already. Tim Cook seems a LOT more apt to push something out for financial reasons where Steve repeatedly over the years delayed releases until it was "perfect."

I use m700 specifically with i5 7400T which are 35w parts and idle at well under 10w. I believe it was around 3-4w last time I checked one with the killawatt.

There are certainly scenarios where rpi make a lot of sense if you truly need something VERY small and next to zero power consumption. But I think for most of what the slashdot crowd does with them, like NAS, media consumption, etc. they would be better served with something like an m700 that is anywhere from four to 10 times (or more!) more powerful in any benchmark and the difference in power costs would amount to maybe $5 a year and still extremely compact in size. Not to mention you don't have to deal with ARM and any compatibility issues there.

Even five nanometer would require octuple-patterning from what I understand, which is absolutely not realistic at scale. Their failure rates at 7nm are already 50% and the costs are estimated to be 30-35% higher to produce 7nm DUV vs 7nm EUV. But to your point, this is absolutely a dead end for China. The only real next step is acquiring, or building (good luck!) EUV machines. (Source and Source)

Agree completely. And it wouldn't affect ad spending online one iota because there's nothing else even close to as effective as online advertising even when its not hyper targeted.

You seem to be only thinking about this from a military perspective and today. Sure, military weapons today can use node sizes that are literally 10x larger (or more) than 22nm. But I think you discount the short term use of AI in the very near term and how important massive amounts of compute will be. It's absolutely the future of warfare I don't think anyone doubts that. China can't compete against the US in a kinetic conflict today and that will just get progressively worse as the west invests in massive compute, dwarfing China's capabilities and using it to develop advanced AI models.

But with all that said, I wasn't talking specifically about the semiconductors that would be used in a traditional conflict between China and the US. The fact of the matter is that China's stated goal is to develop a domestic semiconductor industry that makes them independent from western sources. Producing 22nm wafers in 2023 means you are lightyears behind the west and will be operating at a huge disadvantage. For example, just imagine if Alibaba had to compete against AWS with CPU's that are 10 years behind in performance. Or government agencies doing climate modeling or nuclear fusion simulations or all of the other incredibly important HPC workloads.

I'll add and clarify a couple things. One, AMD was producing 7nm CPUs for retail sale at volume in 2017. Second, those 7nm CPU China is producing are ASIC and not general x86 CPU. Analysts don't even believe that's possible for them today at all. And they aren't even producing simple 7nm ASIC at significant scale yet.

Compound that with the fact that this is the end of the road node size they can produce using DUV equipment and you really see the terrible position China is in. They have to perform a leap from DUV to EUV which took the rest of the world several decades to make any improvement in node size at this point. This isn't an incremental improvement we're talking about it. This requires revolutionary technological developments.

I hear this argument a lot. And for the average desktop computer, I totally agree with you.

But that's not the concern. The concern is cloud, AI, etc. Massive compute requirements. TSMC began mass production of a 22nm node nine years ago. It's going to require many, many times more power and cooling to operate at the same levels of performance as modern 7, 5, 4 and even 3nm node sizes. This is a truly massive disadvantage. And while China has proven they can produce chips using a 7nm node size, it is a dead end best-it-will-ever-get DUV process, not EUV. And the 7nm chips they've produced so far are very simple things like ASIC and a far cry from general purpose CPUs. And as far as I know, they can't even produce that at scale. And for reference, AMD was building 7nm CPU seven years ago.

The most important thing to remember is that China has reached a dead end. There's no smaller node sizes with DUV while the west will continue to drive down node sizes and power consumption while improving efficiency and performance. The next step for China took the collective world about 30 years to create (see: ASML) and China alone won't crack that for decade(s) by which time we will be producing node sizes unimaginably small to China. So not only are they far, far behind today it's actually going to get much, much worse for them in the next few years.

You should be extremely thankful that you're willing to sacrifice that much of your life to work and he's not. Otherwise you might not have that job or the salary would be vastly lower because of all of the additional labor supply in the market.

There's no question that China will be able to produce mass volume chips that are not leading edge nodes. They already do that today, as you pointed out. The problem is that if all they can produce is several generations behind, they will be operating at a huge disadvantage in things like HPC, cloud computing, AI model training, etc. Is your argument that they can produce "good enough" node sizes for general desktop computing? Sure, no argument from me. But for all those really critical use cases they will be operating at a huge disadvantage against other countries that are using the latest node sizes.

7nm would be great, but they cannot even build advanced microprocessors on that node size. It's all simple things like ASICs. The question is, by the time they can manage that, what node size will ASML EUV machines be producing? For basic desktop use, 10nm is more than adequate. That's not the problem. They could buy used desktop PCs if they wanted from all over the world. The problem is advanced chips that are being used in cloud datacenters, HPC, AI training, etc. They will be competing against the rest of the world at a huge disadvantage.

It was hyperbole, it was meant to imply "so long as to be impractical or irrelevant" I didn't actually calculate this and come to exactly 50 years :)

They will never, ever be at generation n-1 given the current trajectory. All they have are DUV machines which are dead ends stuck at 7nm and zero EUV machines. That's like saying they are really close to building a spaceship because they built a steam locomotive except they didn't even build the steam locomotive, they bought it.

It's really, really hard to overstate how difficult of an engineering challenge EUV was, and is. TSMC is planning 2nm for next year. China is like n-10 and has ZERO path to getting to the next step beyond DUV 7nm other than "completely reinvent EUV lithography domestically without the international support that it took to create." For reference, it took ASML seventeen years with zero trade restrictions and international support to build EUV machines.

So maybe in only 10 years, instead of 17, China can develop an EUV lithography machine to compete with the node sizes we have in mass production today? I feel like that's a generous estimate.

Sure, eventually they will. Unfortunately for China it'll be in about 50 years and the rest of the world will be so far ahead it won't matter.

China had already made their intentions of creating a domestic semiconductor industry ("Made in China 2025") very clear. By doing this and forming a coalition this will put China forever generations behind the rest of the world in semiconductor manufacturing. There's no scenario where China alone (or with Russia, Iran, etc lol) will get anywhere close to the leading edge node production. There's no scenario where we could slowly spoon feed them technology to stop them from creating a domestic semiconductor industry. That ship sailed years ago, by China's own public admission. So sure, they would have continued to buy foreign produced chips while they developed their own. That was going to happen either way, so why wouldn't we do everything to hurt China in the process?

Great summary and I agree all around. Chinese's domestic chip manufacturing industry is at a dead end. They cannot even produce advanced 7nm chips. If we look at just the node size, TSMC was producing 7nm SRAM memory in 2016.

So not only are they still far behind today, they also have absolutely zero chance of producing EUV lithography machines so they will continue to fall farther and farther behind.