Date   : Sun, 24 Jul 2011 08:41:10 +0100 (BST)
From   : tommowalker@... (Tom Walker)
Subject: Risc PC (Was 'Minitel in France')

> Premature cease? The A300/A400 range? Well, it is logical given the 
> A5000 and how much of a step forward that was over the older machines 
> (ARM3, faster, better spec, IDE...). Sadly it was still based around the 
> original chipset (so your 800x600 SVGA was somewhat limited).

Don't forget the stupid memory restrictions. Plus video was still from main
memory on the A5000, which hurt performance quite a bit. It was a bit better
than an ARM3'ed A300, but not that much better.

It didn't age well either. In 1991 it was somewhat competitive on most fronts,
by 1993 it was being outperformed (sometimes quite considerably) in many
areas by cheaper PCs. Acorn's reluctance to reduce their prices didn't help.

> > and the early ones at least had no backwards compatibility.
> 
> !!??
> 
> AFAIK the early ones only missed the 16 bit sound. What, in particular, 
> failed as compatibility?

Most games failed. Some other stuff that relied on edge cases in the OS.
Acorn never really gave that much attention to backwards compatibility.

Then of course you've got the endless compatibility problems when the StrongARM
came in. But that's less the fault of Acorn's coders/engineers and more the
reality of them not being ARM's biggest customer anymore (and of course ARM
screwing up the ARM8).

> Don't forget *all* (no exceptions) RISC OS 2 programs that used the
> ???SYS "OS_UpdateMEMC", 64, 64
> trick to speed things up *failed* (no exceptions) on RISC OS 3 which 
> behaved somewhat differently.

Exactly the sort of thing the OS should have trapped and worked around, given
this was hardly an unknown technique.

> Wasn't there a processor that actually follows branches, starts 
> executing *both* paths of code, then tosses away the one not taken?

Cyrix 6x86 does this from memory.

> Not to mention some versions of the Atom (IIRC) that don't execute code 
> as it is written, but rearrange for better performance.

Atom doesn't do this, but it's about the only modern chip that doesn't. 
All of Pentium Pro/II/III/4, Core 1/2/i3/i5/i7, AMD K5/K6/Athlon/Phenom/Fusion,
Cyrix 6x86/MX/M2, VIA Nano, PowerPC 604/G3/G4/G5, ARM Cortex M9 and various
others do this. It's pretty standard.

> I don't think Phoebe would have made the grade. :-(

It didn't even make the grade in 1998. The intended price was around ?1800 
(+ VAT) from memory, for a 233MHz StrongARM machine which would be vastly
outperformed by PCs half the price. Factor in the effect the iMac had at
the same time and it's no wonder Stan Boland decided to throw the towel in.

Tom