[rescue] Re: [geeks] THIS. MAKES. ME. SICK.

[Joshua D. Boyd]

On Wed, 13 Jun 2001 ward at zilla.nu wrote:
> > Itanium = "too little, too late".
> 
> Coulda been cool, but still too late.  And nowhere near as innovative as
> they like to say, seeing as how the PowerPC was basically the same thing
> (take an existing and successful RISC workstation chip, meld in a
> mainstream desktop CPU).

A) How was the PowerPC take a RISC workstation chip (I assume you mean
RS/6000, which I don't think ever got condensed to a single chip until PPC
came along) and meld in a desktop CPU?   I see that it is related to the
RS/6000 (is it compatible with previously existing RS/6000 binaries?
Never found out), but what does it have to do with a mainstream desktop
CPU?

B) How is the Itanium like that process?  Nobody has pointed out any major
similarities between the PA-RISC and the Itanium.  I can't make my own
comparison because I don't know PA-RISC assembly.  Anyone want to point me
to some documents on it (seeing as I just got a PA-RISC machine)?


> > Even HP has moved their focus back to PA-RISC.
> 
> My understanding is that most of the Itanium is just PA-8500 + Intel's
> x86 stuff + the usual improvements/EPIC.  The core instruction set is
> still mostly PA-RISC.  I remember attending a few HP meets and all 
> the HP guys were quite frustrated that Intel was being so slow.  The 
> rumor (and I have no confirmation) was that the 8500 was never supposed 
> to be a fully 'release' CPU, and that the Merced was supposed to be there 
> instead.  Intel must enjoy playing the delay game as much as Motorola
> (remember the 68060?  The PowerPC?).

+the usual improvements?  When has instruction ever been a usual
improvement?  Plus, doesn't the Itanium do away with out of order
execution, a pretty major feature of pretty much every RISC chip that's
come out these past 6-7 years.  Do they even have speculative branching?
I think I heard somewhere that they ditched that in favor of the compiler
marking the most likely branch to execute so that the chip doesn't have to
speculate.
 
> On another note, does anyone else find it terribly funny that Intel's
> old powerhouse CPU, the i860, is now used as a motherboard chip?  I just
> assumed that it was a case of name reuse, but I've been told otherwise.
> Wow!  I remember loving that chip in both HP's graphics box and on a
> NeXTDimension.  Then the i960, its microcontroller cousin, was popping
> up in printers and copiers everywhere I looked.  Next thing you know,
> I'll find an 88110 in my refrigerator.

The i860 also showed up in old SGIs and in the Sun GT, and in the Dec PMAG
highend 3D accelerator.

Has anyone here ever programmed the old i860 hypercube machines that used
to be sold?  I here the l0pht has one working.




--
Joshua Boyd