jdiamond

Yes -and just as Fujitsu designed the SVE SIMD extensions for ARM, it was Apple that created the 64-bit ARM ISA (v8+), totally redesigning it in the process, as they were the first company on Earth to require 64-bit ARM CPUs.  We used to joke about how the new ISA was "ARM" in name only.  So in theory, they could license the ISA to ARM.  And of course, Apple hasn't used the ARM microarchitecture since the Newton - it's a total Apple design.  So what benefit does Apple get from the ARM name?  Just participation in a wide software ecosystem, where people have already created "ARM" versions of popular software, vs the smaller number of companies who would make things for "Apple Silicon".

tskwara said:

timmillea said:

5nM/3nM = 1.6 recurring, suggesting a move from the 5nM process to the 3nM process would yield a 67% improvement in speed/power ratio. We are not seeing that.

Not accounting for other changes and just looking at the size reduction ratio alone, I think the math would suggest that 3nM is a 40% reduction of 5nM, so maybe a 40% increase in speed?

You guys are referring to a concept called "Denard Scaling", but that hasn't actually held up since the late 1990s, and it took a nosedive in the 2000s after 28nm.  Today's denser processor nodes lead to only single digit percent increases in performance.  Some of this is because voltage scaling slowed down.  Some parts, like caches, haven't shrunk at the same rate.  The capacitance of wires and devices is going up.  I remember NVIDIA was freaking out after the 28 nm node, because following that, even the cost per transistor wasn't going down.

The biggest apparent technical issue seems to be the failure of the quad-die Apple Silicon chip.  Perhaps they could make up for this with a dual socket system?  In principle, GPU support is a driver issue, not a HW issue.  It'd also be great if they found a way to do a two tiered memory system, so you could add terabytes of DRAM to the embedded system.  It sounds difficult, but Apple already pulled something similar off with the old two tiered SSD/HDD systems.  You'd let the 192GB of on-die RAM page out to the 8 TB of DIMMs.

I think the main reason to design and support the Mac Pro, even if it loses money overall, is the halo effect.  People who love to use the MacOS, but need to do a larger scale project, particularly with GPUs, suddenly find they are forced to leave the Mac eco system. Just knowing that it is possible to still use MacOS at the higher end provides a layer of comfort in Mac based businesses thinking of expanding in the future.  It also might help with other products on Apple's road map, if it gains them experience in pushing Apple Silicon and GPUs to higher performance levels.

Guys, no one is implying Apple will make a physical car.  It's like AppleTV - initially, everyone hoped for an amazing Apple TV that was transparent and thin.  But it ended up just being a little box you hook up to a normal TV.  There's no reason to think this wouldn't be similar in concept to car play.  Apple would just be helping car companies that need to play catch up on things like self driving features, etc.  And they could go to town integrating the infotainment/entertainment systems with their own stuff.

Intel is partly fabless.  Their high end chips are fabricated at TSMC, and their lower end chips are fabbed using their classic technology.  Intel's hope was that by 2025 to be able to make their own chips again.  We'll see how that timeline plays out.