Does somebody use both, current Intel and AMD tower PCs?

[Liam Proven]

On Sun, 19 Mar 2023 at 18:32, Robert Heller <heller at deepsoft.com> wrote:
>
> What is happening is the *ARM* Chips have caught up to x86 processors and will
> be blowing past, in terms of bang-for-the-buck, CPU processing power, and
> lowest power cost.

Yes and no.

Apple's Arm-ISA chips have. Nobody else's yet.

Apple has also good _really_ good CPU emulation tech, because the
current-gen Apple OS was built on 68030, then moved to Intel and HP
PA-RIS as well, with fat binaries containing multiple CPU architecture
executables in one bundle.

The GUI part also ran on Sun SPARC.

Then it moved to PowerPC, but Intel was kept maintained in the labs.

Then back to x86-32 again, but able to run PowerPC binaries via CPU
emulation, called Rosetta.

Then PowerPC emulation was dropped and it moved to x86-64.

Later, x86-32 was dropped.

Now it's moved to Arm64, but with Rosetta2 x86-64 emulation.

The point being, Apple has more experience with migrating users and
apps from one CPU arch to another than anyone.

Apple also threw a *LOT* of money and talent at very fast Arm-ISA
chips. More than Arm itself has, as it focussed on
performance-per-Watt.

Most of those people have left now and it's not clear if Apple can do it again.

M2 is a lot less impressive relative to M1 than M1 was to Intel x86-64.

>   CISC chips (eg x86 flavors) have pretty much reached their
> "peak" -- at this point transistor size is pushing atomic limits (eg the size
> of a silicon *atom* is the limiting factor in transistor size).

This is equally true of RISC though.

>  CISC chips
> need more complex gate structures / more complex micro code than RISC chips.

Somewhat. Modern RISC is very, very complicated too. The difference is
no longer clear-cut.

> That is RISC processors need fewer transistors / simplier micro code to
> perform the same op, and thus can do more with less (transistors, micro code,
> power consumption, etc.).

Yes, but it's not night and day.

> Apple's M1 already out performs x86s.

Yes it does. But there are no M1 or M2 server chips. It's aimed at
desktop and mobile, and only that.

There isn't even an M1 Mac Pro yet. The M1 has achieved amazing
performance by being _very_ highly integrated. CPU and GPU *and RAM*
on-die, SSDs soldered in.

That's expensive, and it's hot and needs a lot of cooling. There isn't
a lot inside the Mac Studio except huge heatsinks.

It does not adapt well to offboard discrete GPUs, or even to PCIe
slots. That's why the Mac Pro is still x86.

Rip the GPU and video RAM out of the M1 and put it on the other side
of a PCIe bus, and RAM in slots, and SSDs on an NVMe bus or something,
and performance will suffer badly.

-- 
Liam Proven ~ Profile: https://about.me/liamproven
Email: lproven at cix.co.uk ~ gMail/gTalk/FB: lproven at gmail.com
Twitter/LinkedIn: lproven ~ Skype: liamproven
IoM: (+44) 7624 277612: UK: (+44) 7939-087884
Czech [+ WhatsApp/Telegram/Signal]: (+420) 702-829-053