[IP] Why the Intel switch is bad
Begin forwarded message:
From: gep2@xxxxxxxxxxxxx
Date: June 7, 2005 11:49:38 AM EDT
To: dfarber@xxxxxxxxxx
Subject: [IP] Why the Intel switch is bad
The FACT of the matter is that NONE of the RISC-based processors have
ever
really lived up to their promises.
The argument always was that because they were simple, they would use
less power
and you could clock them faster, and supposedly that would make up
for the fact
that you had to do things with subroutines on RISC-based processors
that were
part of the built-in instructions on CISC-based processors.
That never really made sense, even back then. Even if one argues that
CISC-based processors use "subroutines" to implement their more-
complicated
instructions too, the fact is that the related firmware is tightly
bound to the
processor logic itself, meaning both PERFECTLY optimized, and with
minimal
path/propagation distances between the firmware store and where the
firmware
actually gets used. RISC processors virtually never achieve that.
Even in terms of simple raw clock speed, the CISC CPUs don't have
anything at
all to apologize about in comparison to their RISC competitors.
If you go back ten or fifteen years, you'll see all sorts of RISC
processors
which promised to replace (or obsolete!) Intel's and AMD's CISC
processors in
PCs... and in truth NONE of them have done so, and today Intel and
AMD CISC-type
processors are still the winners in the race (and I don't expect that
to change
anytime soon).
When IBM introduced the PowerPC CPU at CeBIT in Germany those many
years ago, I
told the IBM guys at their booth that there was NO way this was going to
'revolutionize' PCs... and for the same reasons I'm saying today.
Eventually,
people have to admit the truth!
<---- Begin Forwarded Message ---->
From: David Farber <dave@xxxxxxxxxx>
Subject: [IP] Why the Intel switch is bad
Date: Mon, 6 Jun 2005 20:07:00 -0400
To: Ip ip <ip@xxxxxxxxxxxxxx>
Reply-To: dave@xxxxxxxxxx
Begin forwarded message:
From: Santi Alonso <santi@xxxxxxxxxxxxxx>
Date: June 6, 2005 6:08:53 PM EDT
To: dave@xxxxxxxxxx
Subject: Why the Intel switch is bad
From: http://angelusfade.spymac.com/blog/
The reasons why this is bad:
- x86 is a technology from the 70s. Yes they made large steps forward
using new technology and incorporating this into new Processors. But
at its inner core the processor is still 16bit, bored once (starting
with the i386SX) to 32 bit and now again with AMD64 or E64MT to 64-
bit. The whole Pnetium Processor range including AMDs are big chunk
of patchwork around a crappy 16bit processor. (actually the patchwork
is 95% of the processor nowadays and the original processor is barely
there but it still is). Sure they prolly can push this forward and
forward with even more patchwork... looks to be fun.
- Altivec is one of the best vector engines which made the Cell
processor possible. It is way beyond anything SSE2 can do on Intel.
- The PowerPC design was 64-bit from the beginning, they saw the time
was not right and downtuned it, but with the G5 they just had to pull
the "64-bit" switch and transition was seemless. The G5 uses 1/4th of
the transistors of a P4, giving it much more room for future growth.
Derived from the Power Processors, which are without any doubt the
best server processors one can buy, this architecture could have
blown away in a year or two, if IBM would be able to produce chips
derived from the Power5 and Power6, which are looking to be the best
CPUs ever made.
- Though Cell technology is nothing for a consumer PC, some of the
technologies used in this processor blow my mind. IBM already
announced parts of this design will go towards the Power family of
chips, if this would have been directed also to the PowerPC lines,
imagine what would have been possible.
- Every single Company that has the oportunity to start with a fresh
design (mainly gaming consoles but also many embedded) are going for
a PowerPC design. Not only is the design more promising for the
future, it is simply better than tuned up 70s technology.
- No matter what "the Steve" says: The switch when we had to transit
from 68k to PPC did hurt. I had one of the last "quadras", it was a
good machine, but there were only a few to distribute fat binaries
and I had to buy a new machine sooner than I wanted to keep stuff.
- No matter what "the Steve" says: The transition from OS 9 to OS X
did hurt. Yes I agree that OS X is by far better and that it was
necessary as OS 9 was outdated technology. But it was a hard time and
there are still things from OS 9 that I miss in OS X. The good side
on this was: They went with the best available technology. Mach
undoubtly being one of the best kernel architectures, FreeBSD being
one of the most secure and best Unix systems available... This is
just not the fact with Intel. Yes they are currently making quite
fast processors (not that much faster than my G5 I think or was all
that Steve told me crap?), but they do it by pressing the last juice
out of outdated technology.
- The BIOS issue. Yes I know Apple can still use theoretically
OpenBoot and stull like that. But some of the smart things Apple did
was use special hardware that was not that PC market stuff but there
are problems that arise... If they use PC stuff they will be up to
the same level in speed than current PCs, having the same problems as
you have in Wintel machines today, not the OS, but some parts have
such a short time to market, that problems have to occur. As seen in
various chipsets and so on, many problems the Wintel crowd blames MS
for are actually hardware issues. Or Apple could go with "we only use
premium", but then they will be a tad behing the crowd. Or the 3rd
choice they become Alienware 2.0 and for using handselected measured
chipsets they have to cut down other things to a point where the
things prolly work due to selected hardware but if not you are in for
a whole hell off support pain, still you would have faster things on
hand-build Wintel PCs.
- The NeXT thing: Fat binaries anyone? Yay now they named them
"universal" binaries. Who the hell wants them? I can see many many
developers that stick with the Mac go "Uh that thing may be universal
but it sure is fat, avtually twice as big as it needs to be... Hey
all geeks should go Intel anyway I skip the PPC compile..." Same
thing happened when we switched from 68k to PPC, most fat binaries
were shareware from devoted fans, first were games to skip the old CPU
(Apple still sold 68k Macs when Command and Conquer came out only for
PPC), then there was Adobe and the rest of the crowd. I bet Adobe CS3
next year will run on Mactel, but fat binary? I doubt that, the CS
products are fat enough as it is.
And there are even more reasons but I am to upset and need to go sulk
in a corner.
Yes maybe in 3-4 years the whole thing will settle and Intel will
again own the whole PC (Wintel and Mactel) market and the PowerPC
will be history. It will be a sad day then, but one thing is for
sure: I will rather go and buy an Alienware Laptop to replace my
dated PowerBook than buy an Apple for the next 3 years. Maybe I will
buy a Mac again someday, but for the time being I am switched, back
to an Intel Notebook prolly running Linux.
-------------------------------------
You are subscribed as gep2@xxxxxxxxxxxxx
To manage your subscription, go to
http://v2.listbox.com/member/?listname=ip
Archives at: http://www.interesting-people.org/archives/interesting-
people/
<---- End Forwarded Message ---->
Gordon Peterson http://personal.terabites.com/
1977-2002 Twenty-fifth anniversary year of Local Area Networking!
Support free and fair US elections! http://stickers.defend-
democracy.org
12/19/98: Partisan Republicans scornfully ignore the voters they
"represent".
12/09/00: the date the Republican Party took down democracy in America.
-------------------------------------
You are subscribed as roessler@xxxxxxxxxxxxxxxxxx
To manage your subscription, go to
http://v2.listbox.com/member/?listname=ip
Archives at: http://www.interesting-people.org/archives/interesting-people/