KennyR wrote:
@Hammer
You raise some interesting points, and you've proven me wrong. I can't argue against your technical knowledge. But, basically the reason I say PPC can be theoretically pushed farther than x86 is because it has no legacy to emulate.
Note that PPC 970 protects PPC based investments i.e. via decode/crush stages. Why not we remove these stages and lets see IF it can run existing 32bit PPC code.
‘Legacy’ itself is not the problem it’s flawed the 8086 era instructions. IF they’re perceived to be flawed, a programmer can just use the modern Pentium Pro era instructions. It's the programmer's choice to use these instructions.
Note that both Intel and AMD have released the ideal programming guidelines for their modern processors.
The modern x86 has many workarounds for the old 8086 architecture,
The lack of general registers is addressed via AMD** and Transmeta** CPU solutions.
**Both vendors supports AMD64/X86-64 ISA.
and yet even the best workarounds have overheads.
That's is why you see Linux kernels complied for a specific x86 class CPUs i.e. 586 kernels only works with Pentium and later x86 CPUs.
Tricks and kludges don't make for a very efficient CPU, no matter how well they're done.
IF they’re perceived to be flawed, a programmer can just use the post-Pentium era instructions.
Granted, that doesn't really matter as things stand, since even with its inefficiencies it's still much faster and will stay that way for the forseeable future.
Note that, there are X86 system programmers who optimize their 'own' code.
(And sliding OT, this doesn't really remove the problem of any OS on x86 being unmarketable.
Note that the Athlon 64/Opteron has increased its general registers to 16, while it keeping the advantages of register renaming regime.
X86 ISA concerns in the higher level languages (3GL and above) is not quite a big deal at this stage.
Software doesn't sell, hardware does, as I'm sure you know.
Not quite, the legacy desktop software investment is the boat anchor for the dominance of X86. Even Intel has to include X86 software (FX32 style) emulator for it’s IA-64****. A response from AMD64/X86-64’s threat. (****IA-64 does have a poor performing X86-32 compatibility mode).
The software is the key. Refer to Beta Max vs VHS wars to illustrate this point. Hardware without software doesn’t offer the total solution.
Any commercial OS would have the threefold problem of coming up against Windows, being pirated like crazy,
I recall X86 Solaris was still available for X86 class CPUs. To bad they are not seriously targeting for home/office desktop use i.e. lack of SUN support for leisure based applications.
and being consigned to The Hell of Multi-Boot as a subordinate to Windows or Linux. Until this changes, my x86 bias will remain.)
Will a different ISA stop Microsoft? IF the PowerPC market size make sense for Microsoft, who can stop them? Remember, they also followed the RISC hype for their Windows NT 4.0 products. I can still remember DEC Alpha version Windows NT 4.0. I think, I still have non-X86 Windows NT 4.0 CDs sitting on the self (somewhere).
At a smaller extent, MS’s Windows CE still covers MIPS and ARM RISC CPU families. It's no surprise that AMD and Intel support both either one of these RISC CPU families.
Note that Linux is available for PPC platform, thus one can not stop the potential dual AmigaOS and Linux setups.
Personally, the type of ISA is not an issue to me (since I program with 3GL and above), it about price and offering the solution.
Playing with AmigaOS remains as a hobby(i.e. "leisure computing") of mine, just like my original A500/A1200. IF Eyetech delivers a (reasonably high performance) PPC solution within the A500/A1200 target price bracket, then I would be open for purchasing the product. I do like Eyetech’s goals for the return of AmigaOS platform (via reasonably cheap PPC solution) into the mainstream shops.
