I confess that I do not understand. Sure, I know that the supply of available RAM cannot meet today’s demand. Prices rise as a result.
But with Apple silicon and unified memory that memory is on the same chip as the CPUs and whatever else, right?
And Apple itself designs that chip before it’s ever etched.
So why can’t Apple just print all the RAM it wants?
I had the same misconception. I believe ATP addressed this a few weeks ago. They’ll explain it much better than I can. But basically Apple still buys ram from other producers that gets sodered/packaged together with the processor that goes into Apple devices. It’s not upgradable for that reason (it’s not on some separate bus that you can plug different ram chips into). But it does mean that Apple is subject to the same pricing and supply pressures as other purchasers from the RAM makers.
The foundry that makes the chips doesn’t make the RAM. It’s a different process. Same chip doesn’t mean the same silicon wafer. An MX chip is a package, not a single chip.
RAM is a different part of the package.
The underlying problem is that very few people in the world have the knowledge to make cutting edge DRAM. That supply is limited until either new foundries are built, or the AI business implodes. I’m betting on the latter.
Is RAM hard? I mean, they’re putting eco- power- etc cores on a chip with neural and GPU engines but they can’t handle the RAM? RAM would seem to be something simple that they could start with.
If they really put it all together wouldn’t that improve the RAM bandwidth a lot?
It’s worth pointing out that I have only a vague idea of what I’m talking about.
Yes it is hard. Worse it requires different equipment and expertise. TSMC is a fab for chips. The people who make the RAM have not only the fab, but also the research and development for designing the RAM.
The companies that have this expertise have a 10yr head start. So even if Apple or TSMC wanted to start today, they’re still behind.