Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 04 Nov 2000 01:32:58 -0800
From:      Kent Stewart <kstewart@urx.com>
To:        Heredity Choice <stork@QNET.COM>
Cc:        Jeremiah Gowdy <jgowdy@home.com>, Allen Landsidel <all@biosys.net>, Nick Slager <nicks@albury.net.au>, jadream <jadream@chat.ru>, freebsd-questions@FreeBSD.ORG
Subject:   Re: FBSD & Itanium?
Message-ID:  <3A03D7CA.1EA714C7@urx.com>
References:  <000901c04631$5a8e4930$4fc7ddd1@STORK>

next in thread | previous in thread | raw e-mail | index | archive | help


Heredity Choice wrote:
> 
> I attribute many of the shortcomings of the x86 family to its backward
> compatibility with the Intel 4044. It has eight registers, compared with 64
> on the Alpha. The architecture is in many ways much simpler on the Alpha.
> 
> Apples and oranges is a poor analogy for 32 bit versus 64 bit.  Nobody
> prefers 32 bit unless it is cheaper. The Alpha is a relatively cheap
> processor to manufacture. The Itanium never will be. Porting software to 64
> bit can be painful, but 32 bit software can run on 64 bit hardware. The
> Ultrasparc and the HP are 64 bit and much new software development is on a
> 64 bit platform even if it will later be ported to a 32 bit platform.

One of the reasons HP and Sun never made it in some markets is their
architecture. Some of the data arrays these old codes used required
logical operations on 64-bit integers. Dec Fortran would compile a
Cray Fortran program with almost no modification. Conversion to a DEC
Alpha required 4 hours of programming and serveral days to get the
documentation done. Going from a CDC (the first company Seymour worked
for) to a Cray also had similar characteristics. We had one project
that was the conversion of a German Saftey analysis program from a
CDC-990 to the XM/P. We made it run in two weeks. They were astounded
when we showed them the output. What they hadn't told me before we
started was that a 5 man team in Japan had required 1.5 years to
convert it to a slightly different system. I had used both the CDC-990
and the Cray. Vi did a wonderful job of sweeping through the Fortran
modules. It seemed like all I did was "n" and "." for several days.
Our conversion was KISS simple. The 5-man team didn't have that
experience and we never purchased their version to see what they had
done. 

There are a lot of people out there still using DEC Alpha's that don't
believe any of the PR-hype until the see a real 64-bit computer
sitting by their desk. They don't care if FreeBSD or Linux runs in
64-bit mode. They have been doing that for around 20 years. What they
don't want to have to do is change 100's of lines of code over weeks
or months of time to make their program work on the new "64-bit"
machines. They are afraid conversion ratios like 4 man weeks to 7.5
man years will pop up again :).

> 
> Despite its high performance, modest cost, and elegant architecture, the
> Alpha has never sold well. Microway is alone marketing the low-end Alphas,
> while Compaq now offers only high-end workstations and servers with the
> Alpha processor.

It also depends on the market you are looking at. For example, I came
from the supercomputer environment where Cray still means something. A
line from their web pages shows
Cray T3E scalable parallel systems use the DECchip 21164 (DEC Alpha
EV5)

They have combinations with up to 2048 processors. They liquid cool
anything over 128 processors. What distinguishes then from things we
use are parameters like I/O channels with 1 GB/sec per channel.


> 
> FreeBSD Alpha does not have the maturity of FreeBSD Intel, but it is making
> great progress because of a few enthusiasts who recognize its great
> potential.
> 
> Meanwhile the Itanium, if it ever becomes reality, is an answer in search of
> a question.

I agree. I think 64-bit will mean the most to sites with huge
databases that want to use GB's of memory to handle the queries. To
really optimize many calculations you need 16-64 64-bit floating point
register's so you can optimize your calculations better.

Kent

> 
> Paul Smith
> 
> > -----Original Message-----
> > From: Jeremiah Gowdy [mailto:jgowdy@home.com]
> > Sent: Friday, 03 November, 2000 12:55 PM
> > To: Heredity Choice; Allen Landsidel; Nick Slager; jadream
> > Cc: freebsd-questions@FreeBSD.ORG
> > Subject: Re: FBSD & Itanium?
> >
> >
> > > Yes, an Athlon 600 is cheaper than an Alpha 21164 533 MHz, and
> > about half
> > as
> > > fast.
> > >
> > > I am afraid there is a little more to porting an OS than having gcc for
> > the
> > > processor, especially as FreeBSD is not optimized for
> > portability. FreeBSD
> > > will be ported to the Itanium if people want to make the considerable
> > > effort. So far the Itanium has been all promises and no substance.
> >
> > Why are we comparing 64bit CPUs to 32bit CPUs ?  Apples to Oranges if you
> > ask me.  I say, wait for the tainted smell of vaporware to disappear, and
> > see who comes out ahead on the Itainium vs Sledgehammer front.  I believe
> > for FreeBSD users, Sledgehammer would be the CPU of choice, as AMD claims
> > that in 32bit mode, it would still be the fastest 32bit x86 processor ever
> > released, even if you ignore the 64bit mode altogether.  So since
> > FreeBSD is
> > situated in the x86 architecture quite well, seems like AMD would be the
> > only option in this next generation of the Intel/AMD offerings.  The Intel
> > Itainium's core is truly emulating x86 (I know all of today's cpus emulate
> > CISC with RISC, however, Itainium's focus isn't x86), the Itainium will
> > certainly perform slower per clock than a Coppermine.  Willamette performs
> > slower per clock than Coppermine.  So if we say, ok FreeBSD is x86
> > optimized, what is the x86 cpu market looking like in the next few years,
> > one has to conclude that the newer AMD offerings, especially the
> > Hammer, are
> > the cpus of choice for FreeBSD x86 systems.  As for Alpha, I've never used
> > FreeBSD/Alpha, however, I know Alphas are absolutely awesome.  My question
> > is, is our Alpha support awesome ?  And are people willing to go Alpha ?
> > I'm not saying anything against Alpha, I'm just putting forth my opinion
> > that once it comes out, FreeBSD x86/Hammer combo will be the best
> > option for
> > FreeBSD x86 users.
> >
> > Linux will probably come out with both an Itainium and Hammer version.  I
> > believe both such projects are underway.  But will they be hack
> > jobs or the
> > real thing ?  FreeBSD x86-64 makes sense, it's just another
> > extention of the
> > architecture.  Ax8664_CPU  :)
> >
> >     I have great doubts about willamette and it's design which
> > performs less
> > per clock than the Coppermine, but allows higher clock speeds.
> > Remember the
> > 486 DX4 ?  Then comes the Pentium and all those empty mhz the DX4s were
> > hitting were stomped by a greater design.  Seems to me the idea shouldn't
> > just be how many gigahertz you can kick, but the design of the
> > core.  If you
> > could make a 386 that clocked to 3 ghz, I would rather have a Thunderbird
> > that ran at 1 ghz.
> >
> > My theory is this, if Hammer running in 32bit mode is far far faster than
> > Thunderbird, and Thunderbird is just a little bit faster per clock than
> > Coppermine, and Coppermine is faster per clock than Willamette, the winner
> > is clear.  Itainium isn't even entering into the picture, because
> > even Intel
> > says the x86 emulation isn't going to be very fast.  It's more of
> > a utility
> > thing, it's there, it's available if you need it.  I personally don't see
> > the need for Willamette.  They can keep their empty clock speed
> > and all the
> > heat that comes with it.
> >
> > Oh, and one more thing, think about the price difference.  How
> > much cheaper
> > do you think the Hammer will be than the Itainium ?  And how much more
> > production will AMD have ?
> >
> > AMD Athlon Thunderbird 1 ghz $254
> > Intel Pentium III Coppermine 1ghz $469
> >
> > AMD Athlon Thunderbird 1.1 ghz $350
> > Intel Pentium III Coppermine 1.1 ghz RECALLED
> >
> > Intel Pentium IV Willamette 1.5ghz $1139  (If you can get ahold of one)
> >
> > What makes Intel CPUs worth twice as much money, the inferior performace ?
> > :)
> >
> > Sorry this got off on an AMD vs Intel Jihad, but I truly can't take any of
> > Intel's upcoming offerings, even the Itainium, seriously.  It's
> > time for IT
> > Managers to stop buying Intel CPUs rather than AMDs because of the brand
> > name and the "stability".  People with old memories of K6/K6-2 cpus, can't
> > seem to wake up to the fact that Intel cpus are the ones having
> > the heat and
> > stability issues.  I won't even go into the rambus fiasco.  If
> > you happen to
> > work under an Intel Only IT Manager, you know what I'm talking about.
> >
> >
> >
> 
> To Unsubscribe: send mail to majordomo@FreeBSD.org
> with "unsubscribe freebsd-questions" in the body of the message

-- 
Kent Stewart
Richland, WA

mailto:kbstew99@hotmail.com
http://kstewart.urx.com/kstewart/index.html
FreeBSD News http://daily.daemonnews.org/


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-questions" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3A03D7CA.1EA714C7>