🚀 go-pugleaf

In article <43va7q$4qq@news.eecs.umich.edu>,
Chung-Hsiung Yang <chyang@quip.eecs.umich.edu> wrote:

>	P7 maybe the end of the line, but from all speculation it
>would incorporate the so called "new" technology from HP that is supposed 
>to trascend "RISC" technology ( whatever that is.  People think it is
>VLIW which is sort of an extension to the so called "RISC" technolgy )
>I am not going into what VLIW is, other people will be happy to tell you.
>As far as it being the end of the line, I don't know, they may chose to do 
>that.  But in my view, the only way to catagorically say that X86 
>architiecture has been squeezed of all its "computation power" is if you
>assume the manufacturing technolgy remains constant.  I mean to make a 
>chip faster all one has to do is speed up the clock!  This of course is 

You can't just increase the clockspeed indefinitely on an existing chip
design.  You have to redesign the chip to allow for higher clock speeds.
Even if you could, you'd quickly fall behind because newer chips aren't
just running at faster clock speeds, they're doing more work per clock
cycle as well.  If Intel was making 133 and 150MHz 486's right now
and working on 200MHz 486's -- with no P5 or P6 chips -- people would
probably start flocking to PPC in droves pretty soon.  A 200MHz 486
would get what?  100 SPECint92?  Compare that to at least 250 SPECint 
for a 200MHz PPC 604.


>a simplistic view, but I just wanted to point out "RISC", "X86", "VLIW"
>sometimes are just marketing terms that people throw  around.  And 
>Instruction Set Architecture is not the only thing that determines 
>the success of a chip. 

Tell that to Intel :-)  I think they'll say that x86 compatibility
is a VERY BIG factor in the success of their chips.  And it may
be a very big factor when it comes to keeping up with the performance
of other chips which aren't constrained to such an old architecture
(so far Intel's been keeping up very well, but lately I think it's
because they can throw so much more money into developing each new
generation of 'x86 chip, and that may not work indefinitely.)



>	That is probably the best attitude to take as a consumer.
>But I just wanted to point out one thing, the only think IBM and
>APPLE have in common at this piont is that they both use the 
>PowerPC chips.  Anybody who have any illusion as to IBM and APPLE
>working together to produce a all world machine would probably have
>to wait awile.  IBM never had the intention to work with APPLE 
>to develop a machine which both would benefit.  IBM is much more

Well, you're probably right in the sense that IBM isn't going to ditch  
OS/2 in favor of MacOS, but they are working together to the extent that
they've defined a new architecture such that IBM machines will be
compatible with Apple machines in the same way that Intel-based machines
are compatible with each other now.  The goal is to set a new "IBM
compatible" standard that eventually replaces the old standard (which
Intel has taken over from IBM).


>>Anyway, the point is that suppose a major PC manufacturer, say 
>>Compaq or Toshiba or Dell, or whatever, decided that they'd better get 
>>into the PowerPC game. The PowerPC market could grow enormously 
>>and do serious damage to Intel. Existing PowerPC processors are 
>
>	Lets put it this way, these PC manufactures wouldn't get into
>the PPC market unless there is a market.   So someone else would have to 
>figure out where this PPC market is.  The fact remains, the successs
>of the PPC does not depend on its architecture, but software.  There
>are no demonstrable OS and software on the PPC as of today that 
>people can trust and buy into.  That is your market, right there,
>and it isn't big.

You say there's no OS and software on the PPC that "people can trust and
buy into"?   Ever heard of the MacOS?  The OS that Microsoft has been  
trying to copy for almost 10 years because it's generally considered to
be the best mainstream OS for personal computers?

And you don't think 10% of the personal computer market is a market?  If
you were a clone maker, you'd rather compete with literally thousands of
other clone makers for the miniscule profit margins on PC clones, rather
than competing with about 4 Mac manufacturers, all of which have pretty
high profit margins and one of which has $1 billion in backorders they 
can't even fill?  The Mac market is several million units per year, and
almost any PC clone maker would *love* to have a chunk of that.  They
have no loyalty to Intel, and in fact they'd probably love to play
IBM/Motorola against Intel in order to get the chips they need when
they need them and at the best price.  Right now they're pretty much
at the mercy of Intel.



>	Frankly, I really don't know what Apple management is doing 
>right now.  I am not so sure if they are sincere in opening up the 
>Mac architecture and sacrifice their control to gain more market share.
>desperately need pre-emptive multitasking and better memory protection. 
>In doing so, the OS is going to be bloated like W95 and better and faster
>machines will be necessary to do all the tasks that the users have been
>doing before.  

Actually, I believe Apple will do a much better job with Copeland than
Microsoft did with Win 95 or even NT.  Remember that the MacOS now is
kind of bloated and largely emulated.  Copeland is almost totally native
and practically a complete rewrite from what I hear, and making it lean
and mean is one of the top priorities.  It's much easier for Apple to
do this than Microsoft, because Mac software is much better behaved
and most of it is already being modified as it's being recompiled for
PPC to be even better behaved.  The worst-behaved software (extensions)
probably won't even be compatible with Copeland so it won't have to
have all that bloat to handle them.  It also won't support 680x0 Macs
so generally it should be very clean compared to Windows (which has to
support an unbelievable amount of poorly-designed hardware and software).
But even if Copeland had substantial bloat over the current MacOS, just
the fact that it'll be almost completely native should at least make
up for that, so that it'll be at least as fast as the current MacOS and
won't require faster hardware.  The memory requirements are actually going 
*down* from what I hear, due to a cool new virtual memory architecture.



>	As a proud owner of an ancient  Quadra 700, I am a fan 
>Apple.  But the fact remains, Intel is still doing all the right things
>to keep X86's momentum going.  In fact, Intel just landed a multi-billion 
>dollar government contract to build a parallel machine based on the P6.

I doubt it will be running Windows :-)  So it really doesn't matter that
it's using P6's.  The number of P6's will be miniscule compared to those
in consumer machines.  IBM has some of the fastest parallel machines
right now using the POWER architecture, but it doesn't really affect
anything.

Much more interesting, IMHO, is the use of an architecture in the embedded
market, where the number of chips is even greater than the personal
computer market.  We see all kinds of deals to use PPC chips in everything
from Ford automobiles to TV set-top boxes.  I haven't seen anything
like that for the P6.  With the PPC it seems that IBM and Motorola can
effortlessly crank out variations of the PPC architecture which are
geared for embedded systems, whereas the 'x86 architecture seems a lot
less flexible.

--
Steve Kanefsky