[drewprops]

I hope they're red.


...

[dglow]

Originally Posted by Kickaha RISC, on the other hand, the principle behind ARM designs such as Apple Silicon, says keep the design simple, but streamlined and capable of inter core connectivity from day one…

Originally Posted by Kickaha This has literally been brewing for 20 years, but held back by the need to rely on Intel. Now that is dropping off rapidly, it'll be highly interesting to see what comes to fruition. Nobody right now has the groundwork laid like Apple. ... CISC winning out over RISC was a tragic misstep.

Originally Posted by Kickaha The philosophies remain though… clean design that is intended to be simple but scalable, or complexity that has diminishing returns. It’s a testament to the Intel and AMD engineers that they’ve been able to keep it going this long, but runway is running out.

My understanding is that RISC design principles won just about everywhere, and nearly every modern chip architecture leverages simple instructions and strategies like branch prediction and speculative execution in similar ways.

For both Intel and AMD today, x86/x64 instruction set compatibility means decoding those instructions into a simpler, private, RISC-like set which then executes in a very RISC-like manner. The decoders consume non-trivial amounts of power, so there's an efficiency hit that Apple avoids with ARM, but it's not the primary reason Apple Silicon is so vastly superior in performance-per-watt.

Originally Posted by chucker I don’t think there’s anything clean about ARM having an instruction specific to JavaScript.

It may not be clean but it's really, really smart. Laying down dedicated silicon to optimize something that's very frequently used? It's the same reasoning that leads Apple to put high-end media engines on even its A-series chips: people record a lot of video with their phones.

Originally Posted by Frank777 It will be interesting to watch if the Compute Module expansion rumours turn out to be true.

Agree completely. Compute modules might be the next step in the story that began with Apple putting an A13 into the Studio Display.

[chucker]

Originally Posted by dglow It may not be clean but it's really, really smart. Laying down dedicated silicon to optimize something that's very frequently used? It's the same reasoning that leads Apple to put high-end media engines on even its A-series chips: people record a lot of video with their phones.

Sure. But you can’t have it both ways. Either ARM is a clean and simple design that avoids legacy, or it’s one that embraces weird specializations to eke out more performance.

[dglow]

Originally Posted by chucker Sure. But you can’t have it both ways. Either ARM is a clean and simple design that avoids legacy, or it’s one that embraces weird specializations to eke out more performance.

Then fine – I don’t need my CPU to pass a purity test. Apple also added instructions to simulate x86’s memory model, which makes Rosetta emulation blazing fast compared to any other ARM chip. Legacy well-spent, I say.

[turtle]

So Apple is gonna make blades cool again?

[PB PM]

Originally Posted by turtle So Apple is gonna make blades cool again?

Not likely, I don’t see how Apple could go anywhere with blades. They dropped doing that for a reason, iCloud exists.

[chucker]

Originally Posted by dglow Then fine – I don’t need my CPU to pass a purity test. Apple also added instructions to simulate x86’s memory model, which makes Rosetta emulation blazing fast compared to any other ARM chip. Legacy well-spent, I say.

Agreed.