Why we don’t see any bleeding edge process node mixed signal audio ICs, but might in the near future

[voodooless]

So, we all know about the AKM fire destroying a significant part of the semiconductor industry for more than a year. This factory obviously made more than just our beloved AKM DACs.

The reason why we didn’t see these chips simply made by other foundries is that these chips are predominantly mixed-signal and analog ICs. The analog part of them is very complex to make and highly tailored to the foundries' production processes. That is also why after the rebuild, they didn’t start outputting the old chips again: new iterations had to be made to work with the new equipment.

Analog IC design is also very reliant on the process mode size. Usually, they pick larger process modes, because the smaller you go, the harder it is to design. Obviously, a smaller process node is not always needed, but with the every high degree of integration we see in SOCs and smartphone chips, this is a hot topic.

The following video gives a good overview of why that is, and what’s going to happen in the future to make small process mode mixed-signal IC a reality. Quite interesting! And obviously, it involved AI

If you like this video, you may like his other videos as well, very informative. It’s often about semiconductors, but not exclusively.

 

[IAtaman]

Interesting video, thanks for sharing.

I think we should stop calling our ML tech AI. Among other things, it might be offensive to our future overlords for whom we seem to be removing the last obstacle for them to be able to design and improve their own hardware without our help.

 

[voodooless]

I think we should stop calling our ML tech AI.

AI is more than just machine learning. It can basically be any algorithm used to perform a task, not just the kind that can actually “learn”. Things like genetic algorithms can be classified as AI, but they are not ML.

 

[IAtaman]

AI is more than just machine learning. It can basically be any algorithm used to perform a task, not just the kind that can actually “learn”. Things like genetic algorithms can be classified as AI, but they are not ML.

Yes, my point is also AI is lot more than ML. People call every software that has ML capability or developled with ML models as "AI" which is incorrect. All 3 examples mentioned in the video as "Analog Design AI SW" are using ML too. [AI] can be any algorithm used to perform a task, not the kind that cab actually "learn" is an incorrect and unfortunate statement.

 

[voodooless]

Yes, my point is also AI is lot more than ML. People call every software that has ML capability or developled with ML models as "AI" which is incorrect.

It is, ML is one type of AI. Like an apple is fruit.

All 3 examples mentioned in the video as "Analog Design AI SW" are using ML too.

Maybe, I don’t remember the specifics.. I do know genetic algorithms were mentioned, and that is not machine learning.

[AI] can be any algorithm used to perform a task, not the kind that cab actually "learn" is an incorrect and unfortunate statement.

That is a misrepresentation. You omit the word “basically”. And obviously it’s a vast oversimplification. It may have been better to substitute the words “perform a task” with “solve a problem”. But from where I stand, that is very similar.

 

[simbloke]

The following video gives a good overview of why that is, and what’s going to happen in the future to make small process mode mixed-signal IC a reality. Quite interesting! And obviously, it involved AI

If you like this video, you may like his other videos as well, very informative. It’s often about semiconductors, but not exclusively.

I really like his videos, even when they're not about semiconductors.

 

[IAtaman]

It is, ML is one type of AI. Like an apple is fruit.

Current ML tech is AI in the same way beyond burger is meat... Maybe next year.

 

[voodooless]

Current ML tech is AI in the same way beyond burger is meat... Maybe next year.

Why would that be so?

 

[IAtaman]

Why would that be so?

I don't know. Something about the texture is a bit off I think. And it does not brown in the same way meat would.

 

[voodooless]

I don't know. Something about the texture is a bit off I think. And it does not brown in the same way meat would.