OT: Itanium, Jerry Sanders and x86-64

OT: Itanium, Jerry Sanders and x86-64

Post by Matthew Gardine » Mon, 27 May 2002 03:49:06



After Jerry Sanders took the stand and appeared to support Microsoft,
there was a large backlash in this forum regarding what he did, however,
many people failed to realise this:

1. Itanium has been a commercial flop. Only 443 systems have shipped.
The aim of the Itanium was to compete against current 64bit RISC
companies such as SUN, however, when you charge a wopping $US1300 for
your Itanium and the OS that runs on it runs terribly, it is hard to
justify to both the workstation and server market why they should buy
it. Having tried it, it is like a low end Mac running Windows XP under
emulation.  This has nothing to do with the operating system what so
ever. This crappy performance was seen in both Linux and HP-UX. Add
ontop of that, for $US7,000 you can get a high end SUN Blade station
with 19" screen, a very good operating system and access to a large
number of software titles for 1/2 the price of a comparable Itanium machine.

2. With this crappy performance comes a justifiable unwillingness by
software companies to make their software VLIW 64bit compliant. When the
installation base is 443, the compilers are *due to the complex
nature of VLIW thus producing non-clean code, and the fact that large
portions of software will need to be rewritten no company can justify
pushing a large amount of resources into that direction.

3. The majority of software developed donot need to be 64bit, yet, the
Itanium demands that you must do so, also, on top of that, the majority
of software used is integer, not FPU, which the Itanium performs lower
than Xeon 32bit processor. From the stats so far, even the PA-RISC
outperforms the Itanium.

4. VLIW based computers are also *for basic and complex workstation
work. SUN, who developed MAJC, which was later going to be used in
embedded devices found out the problems regarding compiling etc and
pushed into a different area, and as a result, replaced it with the
generic RISC processor, Ultra Sparc IIe. As a result, the MAJC is now
used in an area where VLIW can be efficiently used in, that is, graphics
processing, aka, a GPU on a graphics card. Graphical computation is
where VLIW and RISC processors excel due to the nature of their FPU
engine, and the ability to "number crunch" very fast.

So, what does this mean for the x86-64. In the future, we need to
realise that he majority of tasks on a computer are integer based, hence
the reason why Xeon processors tend to appear higher on the TPC bench
marking site, however, when it comes to raw numbercrunching, as used for
video compression, CISC chokes, where RISC and VLIW excel. What does
this mean for the future of computers? The future computers will be a
combination of CISC, RISC and/or VLIW. CISC being used as the main
engine driving the computer, either 32 or 64bit. VLIW and/or RISC pushed
into areas requiring strong FPU performance such as graphics cards and
video capturing card. By the OS pushing all the graphical computation
onto the graphics card etc, the processor, which will be CISC can then
concerntrate on what it good at, and the VLIW/RISC can concertrate on it
is good at.

What does this mean for Windows? Microsoft during this time was
reviewing its continuing support for the Itanium platform and when cam
down to it, there are two issues:

1. How can Microsoft justify pouring billions into a OS for a CPU that
has no applications, very little developer support, and a downright
underperforming dog that has not earnt any money back for Microsoft.
2. The poor performance is making Microsoft products look bad, why would
Microsoft want to associate itself with it?
3. Writing software is overly complex. The x86-64 would allow Microsoft
to port their software to 32bit and 64bit long mode, and pick and choose
what should be 64bit-a-lised. Also, since x86-64 is mearly an extension
of the x86 ISA, tuning compilers and assemblers for the processor will
be fairly simple, yet, still provide the performance increases demanded
by consumers.
4. Future applications, in regards to making them 64bit will be made
easier by the fact that they will mearly require retuning, rather than a
re-write, as required by the Itanium.

Any comments about the above issue? I would be pleased to hear them.

Matthew Gardiner

--
Wally: "Hold it right there, buddy ... that scruffy beard...
         those suspenders... that smug expression...YOU'RE ONE
         OF THOSE CONDESCENDING UNIX USERS!"
UNIX User: "Here's a nickel, kid. Get yourself a better computer."

 
 
 

OT: Itanium, Jerry Sanders and x86-64

Post by Erik Funkenbusc » Mon, 27 May 2002 04:50:28



> 1. Itanium has been a commercial flop. Only 443 systems have shipped.

Where precisely did you get this statistic from?  I know for a fact that
there were far more pre-production Itaniums shipped.

Quote:> The aim of the Itanium was to compete against current 64bit RISC
> companies such as SUN, however, when you charge a wopping $US1300 for
> your Itanium and the OS that runs on it runs terribly, it is hard to

Chip prices are always high initiially because they need volume to bring
them down.  This has always been the case since as far back as I can
remember.  I remember 486 DX2/66's costing over $1500 initially.  As for the
OS, which OS are you talking about, several currently run on it.

Quote:> justify to both the workstation and server market why they should buy
> it. Having tried it, it is like a low end Mac running Windows XP under
> emulation.

Yet you had access to one of those 443 Itanium machines.  Uh huh.

Quote:> This has nothing to do with the operating system what so
> ever. This crappy performance was seen in both Linux and HP-UX. Add
> ontop of that, for $US7,000 you can get a high end SUN Blade station
> with 19" screen, a very good operating system and access to a large
> number of software titles for 1/2 the price of a comparable Itanium
> machine.

The current version of Itanium was never expected to be a big seller.  The
reason is that it took too long for it to be developed and therefore the
technology it was designed to use was obsolete.  McKinley, the next version
is inteneded to use more modern process and upgraded to be much more
competitive.

Quote:> 3. The majority of software developed donot need to be 64bit, yet, the
> Itanium demands that you must do so, also, on top of that, the
> majority of software used is integer, not FPU, which the Itanium
> performs lower than Xeon 32bit processor. From the stats so far, even
> the PA-RISC outperforms the Itanium.

All expected events.

Quote:> What does this mean for Windows? Microsoft during this time was
> reviewing its continuing support for the Itanium platform and when cam
> down to it, there are two issues:

And where exactly did you get this information?

Quote:> 1. How can Microsoft justify pouring billions into a OS for a CPU that
> has no applications, very little developer support, and a downright
> underperforming dog that has not earnt any money back for Microsoft.

MS hasn't put billions into development of their 64 bit OS.

Quote:> 2. The poor performance is making Microsoft products look bad, why
> would Microsoft want to associate itself with it?

Hardly.  Nobody expects applications to perform well on current itanium.

Quote:> 3. Writing software is overly complex. The x86-64 would allow
> Microsoft to port their software to 32bit and 64bit long mode, and
> pick and choose what should be 64bit-a-lised. Also, since x86-64 is
> mearly an extension of the x86 ISA, tuning compilers and assemblers
> for the processor will be fairly simple, yet, still provide the
> performance increases demanded by consumers.

I think you grossly underestimate the difficulty of moving to the Hammer
architecture.

Quote:> 4. Future applications, in regards to making them 64bit will be made
> easier by the fact that they will mearly require retuning, rather
> than a re-write, as required by the Itanium.

One of the factors for the development of .NET was that it would allow easy
porting of apps between architectures.  MS already gains from the Mac alone,
and possibly FreeBSD, but Itanium will also be a benefit.

 
 
 

OT: Itanium, Jerry Sanders and x86-64

Post by Matthew Gardine » Mon, 27 May 2002 05:19:57


Quote:>>1. Itanium has been a commercial flop. Only 443 systems have shipped.

> Where precisely did you get this statistic from?  I know for a fact that
> there were far more pre-production Itaniums shipped.

443 was the number of machines commecially available. Where? a news site
reviewing ia64 vs Hammer. Now, it may be more, however, I can't see it
hitting any large numbers soon.

Quote:>>The aim of the Itanium was to compete against current 64bit RISC
>>companies such as SUN, however, when you charge a wopping $US1300 for
>>your Itanium and the OS that runs on it runs terribly, it is hard to

> Chip prices are always high initiially because they need volume to bring
> them down.  This has always been the case since as far back as I can
> remember.  I remember 486 DX2/66's costing over $1500 initially.  As for the
> OS, which OS are you talking about, several currently run on it.

This *y processor has been out for more than a *y year. Why
should a developer develop software for this chip when it costs $US15000
for a machine? how can he justify the investment?  Intel on one hand
want support from the software community yet unwilling to reduce the
price to bring about economies of scale. If they were charging say,
slightly more than a Xeon processor, no problems, however, righ now they
are charging more than a woonded bull.

Quote:>>justify to both the workstation and server market why they should buy
>>it. Having tried it, it is like a low end Mac running Windows XP under
>>emulation.

> Yet you had access to one of those 443 Itanium machines.  Uh huh.

There are ways and means of accessing things. Its who you know and what
you know.

Quote:>>This has nothing to do with the operating system what so
>>ever. This crappy performance was seen in both Linux and HP-UX. Add
>>ontop of that, for $US7,000 you can get a high end SUN Blade station
>>with 19" screen, a very good operating system and access to a large
>>number of software titles for 1/2 the price of a comparable Itanium
>>machine.

> The current version of Itanium was never expected to be a big seller.  The
> reason is that it took too long for it to be developed and therefore the
> technology it was designed to use was obsolete.  McKinley, the next version
> is inteneded to use more modern process and upgraded to be much more
> competitive.

With what? from the pricing it will still be significantly more than a
Xeon with less pep. If they priced the CPU's so that Dell and other
vendors can sell workstations at $AUS5,000, which is slightly more than
a Xeon workstation, then you would find there would be more interest in
developing the applications to run on it. The only thing it really has
going for it is the ability to scale, but considering that there have
been very few 32way servers shipped, instead, the main sellers are 8 and
4 way systems (compaqs reason for discontinuing the 32 way CPU line), I
think AMD has a winner on its hands.

Quote:>>3. The majority of software developed donot need to be 64bit, yet, the
>>Itanium demands that you must do so, also, on top of that, the
>>majority of software used is integer, not FPU, which the Itanium
>>performs lower than Xeon 32bit processor. From the stats so far, even
>>the PA-RISC outperforms the Itanium.

> All expected events.

Yet, Intel is not addressing the problems. they are still charging too
much for too little performance gain.

Quote:>>What does this mean for Windows? Microsoft during this time was
>>reviewing its continuing support for the Itanium platform and when came
>>down to it, there are two issues:

> And where exactly did you get this information?

Insider(s) attitude to Windows XP on Itanium vs x86-64 + Cutlers
response to x86-64 porting rumours.

Quote:>>1. How can Microsoft justify pouring billions into a OS for a CPU that
>>has no applications, very little developer support, and a downright
>>underperforming dog that has not earnt any money back for Microsoft.

> MS hasn't put billions into development of their 64 bit OS.

Ok, maybe millions, but still, it wasn't a very wise investment,
considering they probably haven't made their money back yet.

Quote:>>2. The poor performance is making Microsoft products look bad, why
>>would Microsoft want to associate itself with it?

> Hardly.  Nobody expects applications to perform well on current itanium.

Or the future Itanium, which is still an overprice show pony.

Quote:>>3. Writing software is overly complex. The x86-64 would allow
>>Microsoft to port their software to 32bit and 64bit long mode, and
>>pick and choose what should be 64bit-a-lised. Also, since x86-64 is
>>mearly an extension of the x86 ISA, tuning compilers and assemblers
>>for the processor will be fairly simple, yet, still provide the
>>performance increases demanded by consumers.

> I think you grossly underestimate the difficulty of moving to the Hammer
> architecture.

Compared to Itanium, it is a walk in the park from what I have heard.

Quote:>>4. Future applications, in regards to making them 64bit will be made
>>easier by the fact that they will mearly require retuning, rather
>>than a re-write, as required by the Itanium.

> One of the factors for the development of .NET was that it would allow easy
> porting of apps between architectures.  MS already gains from the Mac alone,
> and possibly FreeBSD, but Itanium will also be a benefit.

Yet, I don't see any Microsoft products being 100% ported to the .net
framework. Could you please inform me why Microsoft pushes .net yet
doesn't embrace it itself?

Matthew Gardiner

--
Wally: "Hold it right there, buddy ... that scruffy beard...
        those suspenders... that smug expression...YOU'RE ONE
        OF THOSE CONDESCENDING UNIX USERS!"
UNIX User: "Here's a nickel, kid. Get yourself a better computer."

 
 
 

OT: Itanium, Jerry Sanders and x86-64

Post by GreyClou » Mon, 27 May 2002 07:38:03



> After Jerry Sanders took the stand and appeared to support Microsoft,
> there was a large backlash in this forum regarding what he did, however,
> many people failed to realise this:

> 1. Itanium has been a commercial flop. Only 443 systems have shipped.
> The aim of the Itanium was to compete against current 64bit RISC
> companies such as SUN, however, when you charge a wopping $US1300 for
> your Itanium and the OS that runs on it runs terribly, it is hard to
> justify to both the workstation and server market why they should buy
> it. Having tried it, it is like a low end Mac running Windows XP under
> emulation.  This has nothing to do with the operating system what so
> ever. This crappy performance was seen in both Linux and HP-UX. Add
> ontop of that, for $US7,000 you can get a high end SUN Blade station
> with 19" screen, a very good operating system and access to a large
> number of software titles for 1/2 the price of a comparable Itanium machine.

> 2. With this crappy performance comes a justifiable unwillingness by
> software companies to make their software VLIW 64bit compliant. When the
> installation base is 443, the compilers are *due to the complex
> nature of VLIW thus producing non-clean code, and the fact that large
> portions of software will need to be rewritten no company can justify
> pushing a large amount of resources into that direction.

> 3. The majority of software developed donot need to be 64bit, yet, the
> Itanium demands that you must do so, also, on top of that, the majority
> of software used is integer, not FPU, which the Itanium performs lower
> than Xeon 32bit processor. From the stats so far, even the PA-RISC
> outperforms the Itanium.

> 4. VLIW based computers are also *for basic and complex workstation
> work. SUN, who developed MAJC, which was later going to be used in
> embedded devices found out the problems regarding compiling etc and
> pushed into a different area, and as a result, replaced it with the
> generic RISC processor, Ultra Sparc IIe. As a result, the MAJC is now
> used in an area where VLIW can be efficiently used in, that is, graphics
> processing, aka, a GPU on a graphics card. Graphical computation is
> where VLIW and RISC processors excel due to the nature of their FPU
> engine, and the ability to "number crunch" very fast.

> So, what does this mean for the x86-64. In the future, we need to
> realise that he majority of tasks on a computer are integer based, hence
> the reason why Xeon processors tend to appear higher on the TPC bench
> marking site, however, when it comes to raw numbercrunching, as used for
> video compression, CISC chokes, where RISC and VLIW excel. What does
> this mean for the future of computers? The future computers will be a
> combination of CISC, RISC and/or VLIW. CISC being used as the main
> engine driving the computer, either 32 or 64bit. VLIW and/or RISC pushed
> into areas requiring strong FPU performance such as graphics cards and
> video capturing card. By the OS pushing all the graphical computation
> onto the graphics card etc, the processor, which will be CISC can then
> concerntrate on what it good at, and the VLIW/RISC can concertrate on it
> is good at.

> What does this mean for Windows? Microsoft during this time was
> reviewing its continuing support for the Itanium platform and when cam
> down to it, there are two issues:

> 1. How can Microsoft justify pouring billions into a OS for a CPU that
> has no applications, very little developer support, and a downright
> underperforming dog that has not earnt any money back for Microsoft.
> 2. The poor performance is making Microsoft products look bad, why would
> Microsoft want to associate itself with it?
> 3. Writing software is overly complex. The x86-64 would allow Microsoft
> to port their software to 32bit and 64bit long mode, and pick and choose
> what should be 64bit-a-lised. Also, since x86-64 is mearly an extension
> of the x86 ISA, tuning compilers and assemblers for the processor will
> be fairly simple, yet, still provide the performance increases demanded
> by consumers.
> 4. Future applications, in regards to making them 64bit will be made
> easier by the fact that they will mearly require retuning, rather than a
> re-write, as required by the Itanium.

> Any comments about the above issue? I would be pleased to hear them.

The OpenVMS porting team calls the Itanium the Itanic.
They're discovering that the Itanium is a piece of crap.
I've heard that Intel is scrambling to fix it up by using
the next generation of the McKinley processor, but the old
DEC engineers say the Alpha is by far superior in
performance.  The regulars  of the Alpha workstations with
OpenVMS said it would be a downgrade to move to the Intel
processor.
 
 
 

OT: Itanium, Jerry Sanders and x86-64

Post by GreyClou » Mon, 27 May 2002 07:40:12




> > 1. Itanium has been a commercial flop. Only 443 systems have shipped.

> Where precisely did you get this statistic from?  I know for a fact that
> there were far more pre-production Itaniums shipped.

> > The aim of the Itanium was to compete against current 64bit RISC
> > companies such as SUN, however, when you charge a wopping $US1300 for
> > your Itanium and the OS that runs on it runs terribly, it is hard to

> Chip prices are always high initiially because they need volume to bring
> them down.  This has always been the case since as far back as I can
> remember.  I remember 486 DX2/66's costing over $1500 initially.  As for the
> OS, which OS are you talking about, several currently run on it.

> > justify to both the workstation and server market why they should buy
> > it. Having tried it, it is like a low end Mac running Windows XP under
> > emulation.

> Yet you had access to one of those 443 Itanium machines.  Uh huh.

> > This has nothing to do with the operating system what so
> > ever. This crappy performance was seen in both Linux and HP-UX. Add
> > ontop of that, for $US7,000 you can get a high end SUN Blade station
> > with 19" screen, a very good operating system and access to a large
> > number of software titles for 1/2 the price of a comparable Itanium
> > machine.

> The current version of Itanium was never expected to be a big seller.  The
> reason is that it took too long for it to be developed and therefore the
> technology it was designed to use was obsolete.  McKinley, the next version
> is inteneded to use more modern process and upgraded to be much more
> competitive.

> > 3. The majority of software developed donot need to be 64bit, yet, the
> > Itanium demands that you must do so, also, on top of that, the
> > majority of software used is integer, not FPU, which the Itanium
> > performs lower than Xeon 32bit processor. From the stats so far, even
> > the PA-RISC outperforms the Itanium.

> All expected events.

> > What does this mean for Windows? Microsoft during this time was
> > reviewing its continuing support for the Itanium platform and when cam
> > down to it, there are two issues:

> And where exactly did you get this information?

> > 1. How can Microsoft justify pouring billions into a OS for a CPU that
> > has no applications, very little developer support, and a downright
> > underperforming dog that has not earnt any money back for Microsoft.

> MS hasn't put billions into development of their 64 bit OS.

> > 2. The poor performance is making Microsoft products look bad, why
> > would Microsoft want to associate itself with it?

> Hardly.  Nobody expects applications to perform well on current itanium.

> > 3. Writing software is overly complex. The x86-64 would allow
> > Microsoft to port their software to 32bit and 64bit long mode, and
> > pick and choose what should be 64bit-a-lised. Also, since x86-64 is
> > mearly an extension of the x86 ISA, tuning compilers and assemblers
> > for the processor will be fairly simple, yet, still provide the
> > performance increases demanded by consumers.

> I think you grossly underestimate the difficulty of moving to the Hammer
> architecture.

> > 4. Future applications, in regards to making them 64bit will be made
> > easier by the fact that they will mearly require retuning, rather
> > than a re-write, as required by the Itanium.

> One of the factors for the development of .NET was that it would allow easy
> porting of apps between architectures.  MS already gains from the Mac alone,
> and possibly FreeBSD, but Itanium will also be a benefit.

Eric, the Itanium is a dog.  It way under performs what the
current Alpha processors are already doing at a lower price
tag.  The OpenVMS regulars now call the Itanium the Itanic.
 
 
 

OT: Itanium, Jerry Sanders and x86-64

Post by Paul Cook » Mon, 27 May 2002 08:13:24



>>>1. Itanium has been a commercial flop. Only 443 systems have shipped.

>> Where precisely did you get this statistic from?  I know for a fact
>> that there were far more pre-production Itaniums shipped.

> 443 was the number of machines commecially available. Where? a news site
> reviewing ia64 vs Hammer. Now, it may be more, however, I can't see it
> hitting any large numbers soon.

>>>The aim of the Itanium was to compete against current 64bit RISC
>>>companies such as SUN, however, when you charge a wopping $US1300 for
>>>your Itanium and the OS that runs on it runs terribly, it is hard to

>> Chip prices are always high initiially because they need volume to
>> bring them down.  This has always been the case since as far back as I
>> can remember.  I remember 486 DX2/66's costing over $1500 initially. As
>> for the OS, which OS are you talking about, several currently run on
>> it.

> This *y processor has been out for more than a *y year. Why
> should a developer develop software for this chip when it costs $US15000
> for a machine? how can he justify the investment?  Intel on one hand
> want support from the software community yet unwilling to reduce the
> price to bring about economies of scale. If they were charging say,
> slightly more than a Xeon processor, no problems, however, righ now they
> are charging more than a woonded bull.

>>>justify to both the workstation and server market why they should buy
>>>it. Having tried it, it is like a low end Mac running Windows XP under
>>>emulation.

>> Yet you had access to one of those 443 Itanium machines.  Uh huh.

> There are ways and means of accessing things. Its who you know and what
> you know.

>>>This has nothing to do with the operating system what so ever. This
>>>crappy performance was seen in both Linux and HP-UX. Add ontop of that,
>>>for $US7,000 you can get a high end SUN Blade station with 19" screen,
>>>a very good operating system and access to a large number of software
>>>titles for 1/2 the price of a comparable Itanium machine.

>> The current version of Itanium was never expected to be a big seller.
>> The reason is that it took too long for it to be developed and
>> therefore the technology it was designed to use was obsolete. McKinley,
>> the next version is inteneded to use more modern process and upgraded
>> to be much more competitive.

> With what? from the pricing it will still be significantly more than a
> Xeon with less pep. If they priced the CPU's so that Dell and other
> vendors can sell workstations at $AUS5,000, which is slightly more than
> a Xeon workstation, then you would find there would be more interest in
> developing the applications to run on it. The only thing it really has
> going for it is the ability to scale, but considering that there have
> been very few 32way servers shipped, instead, the main sellers are 8 and
> 4 way systems (compaqs reason for discontinuing the 32 way CPU line), I
> think AMD has a winner on its hands.

>>>3. The majority of software developed donot need to be 64bit, yet, the
>>>Itanium demands that you must do so, also, on top of that, the majority
>>>of software used is integer, not FPU, which the Itanium performs lower
>>>than Xeon 32bit processor. From the stats so far, even the PA-RISC
>>>outperforms the Itanium.

>> All expected events.

> Yet, Intel is not addressing the problems. they are still charging too
> much for too little performance gain.

>>>What does this mean for Windows? Microsoft during this time was
>>>reviewing its continuing support for the Itanium platform and when came
>>>down to it, there are two issues:

>> And where exactly did you get this information?

> Insider(s) attitude to Windows XP on Itanium vs x86-64 + Cutlers
> response to x86-64 porting rumours.

>>>1. How can Microsoft justify pouring billions into a OS for a CPU that
>>>has no applications, very little developer support, and a downright
>>>underperforming dog that has not earnt any money back for Microsoft.

>> MS hasn't put billions into development of their 64 bit OS.

> Ok, maybe millions, but still, it wasn't a very wise investment,
> considering they probably haven't made their money back yet.

>>>2. The poor performance is making Microsoft products look bad, why
>>>would Microsoft want to associate itself with it?

>> Hardly.  Nobody expects applications to perform well on current
>> itanium.

> Or the future Itanium, which is still an overprice show pony.

>>>3. Writing software is overly complex. The x86-64 would allow Microsoft
>>>to port their software to 32bit and 64bit long mode, and pick and
>>>choose what should be 64bit-a-lised. Also, since x86-64 is mearly an
>>>extension of the x86 ISA, tuning compilers and assemblers for the
>>>processor will be fairly simple, yet, still provide the performance
>>>increases demanded by consumers.

>> I think you grossly underestimate the difficulty of moving to the
>> Hammer architecture.

> Compared to Itanium, it is a walk in the park from what I have heard.

>>>4. Future applications, in regards to making them 64bit will be made
>>>easier by the fact that they will mearly require retuning, rather than
>>>a re-write, as required by the Itanium.

>> One of the factors for the development of .NET was that it would allow
>> easy porting of apps between architectures.  MS already gains from the
>> Mac alone, and possibly FreeBSD, but Itanium will also be a benefit.

> Yet, I don't see any Microsoft products being 100% ported to the .net
> framework. Could you please inform me why Microsoft pushes .net yet
> doesn't embrace it itself?

> Matthew Gardiner

who were you quoting??? I know it was one of the trolls in my Killfile,
but please attribute... :)

--
Paul Cooke
  Registered Linux user 273897 Machine registration number 156819
  Linux Counter: Home Page = http://www.veryComputer.com/
  7:10am  up 1 day, 16:35,  1 user,  load average: 0.22, 0.32, 0.22

 
 
 

OT: Itanium, Jerry Sanders and x86-64

Post by Erik Funkenbusc » Mon, 27 May 2002 08:53:52





>>> 1. Itanium has been a commercial flop. Only 443 systems have
>>> shipped.

>> Where precisely did you get this statistic from?  I know for a fact
>> that there were far more pre-production Itaniums shipped.

>>> The aim of the Itanium was to compete against current 64bit RISC
>>> companies such as SUN, however, when you charge a wopping $US1300
>>> for your Itanium and the OS that runs on it runs terribly, it is
>>> hard to

>> Chip prices are always high initiially because they need volume to
>> bring them down.  This has always been the case since as far back as
>> I can remember.  I remember 486 DX2/66's costing over $1500
>> initially.  As for the OS, which OS are you talking about, several
>> currently run on it.

>>> justify to both the workstation and server market why they should
>>> buy it. Having tried it, it is like a low end Mac running Windows
>>> XP under emulation.

>> Yet you had access to one of those 443 Itanium machines.  Uh huh.

>>> This has nothing to do with the operating system what so
>>> ever. This crappy performance was seen in both Linux and HP-UX. Add
>>> ontop of that, for $US7,000 you can get a high end SUN Blade station
>>> with 19" screen, a very good operating system and access to a large
>>> number of software titles for 1/2 the price of a comparable Itanium
>>> machine.

>> The current version of Itanium was never expected to be a big
>> seller.  The reason is that it took too long for it to be developed
>> and therefore the technology it was designed to use was obsolete.
>> McKinley, the next version is inteneded to use more modern process
>> and upgraded to be much more competitive.

>>> 3. The majority of software developed donot need to be 64bit, yet,
>>> the Itanium demands that you must do so, also, on top of that, the
>>> majority of software used is integer, not FPU, which the Itanium
>>> performs lower than Xeon 32bit processor. From the stats so far,
>>> even the PA-RISC outperforms the Itanium.

>> All expected events.

>>> What does this mean for Windows? Microsoft during this time was
>>> reviewing its continuing support for the Itanium platform and when
>>> cam down to it, there are two issues:

>> And where exactly did you get this information?

>>> 1. How can Microsoft justify pouring billions into a OS for a CPU
>>> that has no applications, very little developer support, and a
>>> downright underperforming dog that has not earnt any money back for
>>> Microsoft.

>> MS hasn't put billions into development of their 64 bit OS.

>>> 2. The poor performance is making Microsoft products look bad, why
>>> would Microsoft want to associate itself with it?

>> Hardly.  Nobody expects applications to perform well on current
>> itanium.

>>> 3. Writing software is overly complex. The x86-64 would allow
>>> Microsoft to port their software to 32bit and 64bit long mode, and
>>> pick and choose what should be 64bit-a-lised. Also, since x86-64 is
>>> mearly an extension of the x86 ISA, tuning compilers and assemblers
>>> for the processor will be fairly simple, yet, still provide the
>>> performance increases demanded by consumers.

>> I think you grossly underestimate the difficulty of moving to the
>> Hammer architecture.

>>> 4. Future applications, in regards to making them 64bit will be made
>>> easier by the fact that they will mearly require retuning, rather
>>> than a re-write, as required by the Itanium.

>> One of the factors for the development of .NET was that it would
>> allow easy porting of apps between architectures.  MS already gains
>> from the Mac alone, and possibly FreeBSD, but Itanium will also be a
>> benefit.

> Eric, the Itanium is a dog.  It way under performs what the
> current Alpha processors are already doing at a lower price
> tag.  The OpenVMS regulars now call the Itanium the Itanic.

Do you think you're contradicting me?  I said so myself.  What is your
point?
 
 
 

OT: Itanium, Jerry Sanders and x86-64

Post by Erik Funkenbusc » Mon, 27 May 2002 09:56:54


Matthew Gardiner <matga...@nospam.com> wrote:
>>> 1. Itanium has been a commercial flop. Only 443 systems have
>>> shipped.

>> Where precisely did you get this statistic from?  I know for a fact
>> that there were far more pre-production Itaniums shipped.

> 443 was the number of machines commecially available. Where? a news
> site reviewing ia64 vs Hammer. Now, it may be more, however, I can't
> see it hitting any large numbers soon.

So in other words, that's old data.

>>> The aim of the Itanium was to compete against current 64bit RISC
>>> companies such as SUN, however, when you charge a wopping $US1300
>>> for your Itanium and the OS that runs on it runs terribly, it is
>>> hard to

>> Chip prices are always high initiially because they need volume to
>> bring them down.  This has always been the case since as far back as
>> I can remember.  I remember 486 DX2/66's costing over $1500
>> initially.  As for the OS, which OS are you talking about, several
>> currently run on it.

> This bloody processor has been out for more than a bloody year.

Well, it's been almost a year exactly (three more days mark the 1 year
mark).

However, Intel had great difficulty in actually producing any quantity of
chips in the beginning, which also contributed to its price.  Another
contributor is that Itanium machines tend to have top of the line
components.

For instance, the Compaq Proliant DL590/64:

http://www.smb.compaq.com/dstore/default.asp?page=ctoBases&ProductLin...
&Family_Id=1087

It's dual processor, with 2MB L2 cache, 1GB Memory (ECC), SCSI RAID Array,
SCA Drive Bay, 11 64-Bit PCI slots, Hot Plug power supplies, etc..

Those extras are what cost so much, not so much the processor itself.  Now,
try an HP Workstation instead:

http://www.hp.com/workstations/products/itanium/i2000/summary.html

That's quite a bit cheaper.  Still expensive, as is to be expected with
small volume products, but that's also small volume in terms of HP's
chipsets as well which add to the price.

>>> justify to both the workstation and server market why they should
>>> buy it. Having tried it, it is like a low end Mac running Windows
>>> XP under emulation.

>> Yet you had access to one of those 443 Itanium machines.  Uh huh.

> There are ways and means of accessing things. Its who you know and
> what you know.

You're so mysterious.

>>> This has nothing to do with the operating system what so
>>> ever. This crappy performance was seen in both Linux and HP-UX. Add
>>> ontop of that, for $US7,000 you can get a high end SUN Blade station
>>> with 19" screen, a very good operating system and access to a large
>>> number of software titles for 1/2 the price of a comparable Itanium
>>> machine.

>> The current version of Itanium was never expected to be a big
>> seller.  The reason is that it took too long for it to be developed
>> and therefore the technology it was designed to use was obsolete.
>> McKinley, the next version is inteneded to use more modern process
>> and upgraded to be much more competitive.

> With what? from the pricing it will still be significantly more than a
> Xeon with less pep. If they priced the CPU's so that Dell and other
> vendors can sell workstations at $AUS5,000, which is slightly more
> than a Xeon workstation, then you would find there would be more
> interest in developing the applications to run on it. The only thing
> it really has going for it is the ability to scale, but considering
> that there have been very few 32way servers shipped, instead, the
> main sellers are 8 and 4 way systems (compaqs reason for
> discontinuing the 32 way CPU line), I think AMD has a winner on its
> hands.

Of course AMD has a winner on their hands.  There is no doubt about that.
However, Intel is not out for the count on this.  As I said, this was
expected, and Intel and HP were warning customers almost a year before
Itanium shipped that the first version would be disappointing.

>>> 3. The majority of software developed donot need to be 64bit, yet,
>>> the Itanium demands that you must do so, also, on top of that, the
>>> majority of software used is integer, not FPU, which the Itanium
>>> performs lower than Xeon 32bit processor. From the stats so far,
>>> even the PA-RISC outperforms the Itanium.

>> All expected events.

> Yet, Intel is not addressing the problems. they are still charging too
> much for too little performance gain.

They're charging what they have to charge, based on its volume.  It's a
chicken and egg scenario, prices can't come down till volume goes up, and
volume won't go up until prices come down.  In this case though, Even
dropping the price isn't likely to gain more sales because of the Itanium's
poor performance.  The people that are buying them will buy them anyways.

>>> What does this mean for Windows? Microsoft during this time was
>>> reviewing its continuing support for the Itanium platform and when
>>> came down to it, there are two issues:

>> And where exactly did you get this information?

> Insider(s) attitude to Windows XP on Itanium vs x86-64 + Cutlers
> response to x86-64 porting rumours.

I haven't read those.  Do you have a link?

>>> 1. How can Microsoft justify pouring billions into a OS for a CPU
>>> that has no applications, very little developer support, and a
>>> downright underperforming dog that has not earnt any money back for
>>> Microsoft.

>> MS hasn't put billions into development of their 64 bit OS.

> Ok, maybe millions, but still, it wasn't a very wise investment,
> considering they probably haven't made their money back yet.

It was a fine investment.  MS needs 64 bit support, and that investment
certainly will go a long way to making it cheaper to port to x86-64.

>>> 2. The poor performance is making Microsoft products look bad, why
>>> would Microsoft want to associate itself with it?

>> Hardly.  Nobody expects applications to perform well on current
>> itanium.

> Or the future Itanium, which is still an overprice show pony.

I'm not saying Itanium is going to take over the current x86 market.  It's
not, at least not anytime soon.  Desktop apps just don't need 64 bit
performance, and 64 bits can add to increased bloat (Being both RISC and
twice the instruction size, that means more instructions per operation and
twice the size of both the instruction and it's data).  But there is
certainly a cross section of the marketplace that needs 64 bit computing.
Why do you think Sun has such a hard time selling SPARC systems for
desktops?

>>> 3. Writing software is overly complex. The x86-64 would allow
>>> Microsoft to port their software to 32bit and 64bit long mode, and
>>> pick and choose what should be 64bit-a-lised. Also, since x86-64 is
>>> mearly an extension of the x86 ISA, tuning compilers and assemblers
>>> for the processor will be fairly simple, yet, still provide the
>>> performance increases demanded by consumers.

>> I think you grossly underestimate the difficulty of moving to the
>> Hammer architecture.

> Compared to Itanium, it is a walk in the park from what I have heard.

Since MS has already created a 64 bit version of XP, it's unlikely that
they'll do a partial port to Hammer.

Sadly, AMD is just perpetuating the crappy x86 architecture that Intel has
been trying to dump for 10+ years.  Itanium has some cool features, the
least of which are true general purpose registers, register remapping, and
various pipelining and parallellization.  It's definately a good
architecture, it's just that Intel has had a hell of a time getting it to
work properly.

Itanium is well ahead of it's time.  It will scale well 10 years from now,
which may not be true of existing processors like Alpha and PPC.  It may
just be a waiting game for Intel.  People had similar complaints about the
Pentium Pro, but Intel managed to turn that into the Pentium II/III/4 which
have worked pretty well.

Of coruse Intel is hedging its bets as well, Prescott is rumored to be a 64
bit version of the P4 that's in development.  This may be a way for Intel to
transition people to a future Itanium as well.

>>> 4. Future applications, in regards to making them 64bit will be made
>>> easier by the fact that they will mearly require retuning, rather
>>> than a re-write, as required by the Itanium.

>> One of the factors for the development of .NET was that it would
>> allow easy porting of apps between architectures.  MS already gains
>> from the Mac alone, and possibly FreeBSD, but Itanium will also be a
>> benefit.

> Yet, I don't see any Microsoft products being 100% ported to the .net
> framework. Could you please inform me why Microsoft pushes .net yet
> doesn't embrace it itself?

That's missing the point.  .NET apps don't have to be 100% ported for .NET
to be a huge benefit.  Say you write your app in 80% .NET and 20% native
code.  You now only have to port 20% of your app to a new architecture
instead of the whole thing.  Microsoft is far more pragmatic than Sun is
regarding portability.  They know that true portability has too high of a
cost in terms of performance for it to ever be successful.
 
 
 

OT: Itanium, Jerry Sanders and x86-64

Post by Matthew Gardine » Mon, 27 May 2002 15:56:33






>>>>1. Itanium has been a commercial flop. Only 443 systems have
>>>>shipped.

>>>Where precisely did you get this statistic from?  I know for a fact
>>>that there were far more pre-production Itaniums shipped.

>>>>The aim of the Itanium was to compete against current 64bit RISC
>>>>companies such as SUN, however, when you charge a wopping $US1300
>>>>for your Itanium and the OS that runs on it runs terribly, it is
>>>>hard to

>>>Chip prices are always high initiially because they need volume to
>>>bring them down.  This has always been the case since as far back as
>>>I can remember.  I remember 486 DX2/66's costing over $1500
>>>initially.  As for the OS, which OS are you talking about, several
>>>currently run on it.

>>>>justify to both the workstation and server market why they should
>>>>buy it. Having tried it, it is like a low end Mac running Windows
>>>>XP under emulation.

>>>Yet you had access to one of those 443 Itanium machines.  Uh huh.

>>>>This has nothing to do with the operating system what so
>>>>ever. This crappy performance was seen in both Linux and HP-UX. Add
>>>>ontop of that, for $US7,000 you can get a high end SUN Blade station
>>>>with 19" screen, a very good operating system and access to a large
>>>>number of software titles for 1/2 the price of a comparable Itanium
>>>>machine.

>>>The current version of Itanium was never expected to be a big
>>>seller.  The reason is that it took too long for it to be developed
>>>and therefore the technology it was designed to use was obsolete.
>>>McKinley, the next version is inteneded to use more modern process
>>>and upgraded to be much more competitive.

>>>>3. The majority of software developed donot need to be 64bit, yet,
>>>>the Itanium demands that you must do so, also, on top of that, the
>>>>majority of software used is integer, not FPU, which the Itanium
>>>>performs lower than Xeon 32bit processor. From the stats so far,
>>>>even the PA-RISC outperforms the Itanium.

>>>All expected events.

>>>>What does this mean for Windows? Microsoft during this time was
>>>>reviewing its continuing support for the Itanium platform and when
>>>>cam down to it, there are two issues:

>>>And where exactly did you get this information?

>>>>1. How can Microsoft justify pouring billions into a OS for a CPU
>>>>that has no applications, very little developer support, and a
>>>>downright underperforming dog that has not earnt any money back for
>>>>Microsoft.

>>>MS hasn't put billions into development of their 64 bit OS.

>>>>2. The poor performance is making Microsoft products look bad, why
>>>>would Microsoft want to associate itself with it?

>>>Hardly.  Nobody expects applications to perform well on current
>>>itanium.

>>>>3. Writing software is overly complex. The x86-64 would allow
>>>>Microsoft to port their software to 32bit and 64bit long mode, and
>>>>pick and choose what should be 64bit-a-lised. Also, since x86-64 is
>>>>mearly an extension of the x86 ISA, tuning compilers and assemblers
>>>>for the processor will be fairly simple, yet, still provide the
>>>>performance increases demanded by consumers.

>>>I think you grossly underestimate the difficulty of moving to the
>>>Hammer architecture.

>>>>4. Future applications, in regards to making them 64bit will be made
>>>>easier by the fact that they will mearly require retuning, rather
>>>>than a re-write, as required by the Itanium.

>>>One of the factors for the development of .NET was that it would
>>>allow easy porting of apps between architectures.  MS already gains
>>>from the Mac alone, and possibly FreeBSD, but Itanium will also be a
>>>benefit.

>>Eric, the Itanium is a dog.  It way under performs what the
>>current Alpha processors are already doing at a lower price
>>tag.  The OpenVMS regulars now call the Itanium the Itanic.

> Do you think you're contradicting me?  I said so myself.  What is your
> point?

Maybe he interpreted your post as a pro-itanium angle.

Matthew Gardiner

--
Wally: "Hold it right there, buddy ... that scruffy beard...
        those suspenders... that smug expression...YOU'RE ONE
        OF THOSE CONDESCENDING UNIX USERS!"
UNIX User: "Here's a nickel, kid. Get yourself a better computer."

 
 
 

OT: Itanium, Jerry Sanders and x86-64

Post by Matthew Gardine » Mon, 27 May 2002 15:58:47




>>After Jerry Sanders took the stand and appeared to support Microsoft,
>>there was a large backlash in this forum regarding what he did, however,
>>many people failed to realise this:

>>1. Itanium has been a commercial flop. Only 443 systems have shipped.
>>The aim of the Itanium was to compete against current 64bit RISC
>>companies such as SUN, however, when you charge a wopping $US1300 for
>>your Itanium and the OS that runs on it runs terribly, it is hard to
>>justify to both the workstation and server market why they should buy
>>it. Having tried it, it is like a low end Mac running Windows XP under
>>emulation.  This has nothing to do with the operating system what so
>>ever. This crappy performance was seen in both Linux and HP-UX. Add
>>ontop of that, for $US7,000 you can get a high end SUN Blade station
>>with 19" screen, a very good operating system and access to a large
>>number of software titles for 1/2 the price of a comparable Itanium machine.

>>2. With this crappy performance comes a justifiable unwillingness by
>>software companies to make their software VLIW 64bit compliant. When the
>>installation base is 443, the compilers are *due to the complex
>>nature of VLIW thus producing non-clean code, and the fact that large
>>portions of software will need to be rewritten no company can justify
>>pushing a large amount of resources into that direction.

>>3. The majority of software developed donot need to be 64bit, yet, the
>>Itanium demands that you must do so, also, on top of that, the majority
>>of software used is integer, not FPU, which the Itanium performs lower
>>than Xeon 32bit processor. From the stats so far, even the PA-RISC
>>outperforms the Itanium.

>>4. VLIW based computers are also *for basic and complex workstation
>>work. SUN, who developed MAJC, which was later going to be used in
>>embedded devices found out the problems regarding compiling etc and
>>pushed into a different area, and as a result, replaced it with the
>>generic RISC processor, Ultra Sparc IIe. As a result, the MAJC is now
>>used in an area where VLIW can be efficiently used in, that is, graphics
>>processing, aka, a GPU on a graphics card. Graphical computation is
>>where VLIW and RISC processors excel due to the nature of their FPU
>>engine, and the ability to "number crunch" very fast.

>>So, what does this mean for the x86-64. In the future, we need to
>>realise that he majority of tasks on a computer are integer based, hence
>>the reason why Xeon processors tend to appear higher on the TPC bench
>>marking site, however, when it comes to raw numbercrunching, as used for
>>video compression, CISC chokes, where RISC and VLIW excel. What does
>>this mean for the future of computers? The future computers will be a
>>combination of CISC, RISC and/or VLIW. CISC being used as the main
>>engine driving the computer, either 32 or 64bit. VLIW and/or RISC pushed
>>into areas requiring strong FPU performance such as graphics cards and
>>video capturing card. By the OS pushing all the graphical computation
>>onto the graphics card etc, the processor, which will be CISC can then
>>concerntrate on what it good at, and the VLIW/RISC can concertrate on it
>>is good at.

>>What does this mean for Windows? Microsoft during this time was
>>reviewing its continuing support for the Itanium platform and when cam
>>down to it, there are two issues:

>>1. How can Microsoft justify pouring billions into a OS for a CPU that
>>has no applications, very little developer support, and a downright
>>underperforming dog that has not earnt any money back for Microsoft.
>>2. The poor performance is making Microsoft products look bad, why would
>>Microsoft want to associate itself with it?
>>3. Writing software is overly complex. The x86-64 would allow Microsoft
>>to port their software to 32bit and 64bit long mode, and pick and choose
>>what should be 64bit-a-lised. Also, since x86-64 is mearly an extension
>>of the x86 ISA, tuning compilers and assemblers for the processor will
>>be fairly simple, yet, still provide the performance increases demanded
>>by consumers.
>>4. Future applications, in regards to making them 64bit will be made
>>easier by the fact that they will mearly require retuning, rather than a
>>re-write, as required by the Itanium.

>>Any comments about the above issue? I would be pleased to hear them.

> The OpenVMS porting team calls the Itanium the Itanic.
> They're discovering that the Itanium is a piece of crap.
> I've heard that Intel is scrambling to fix it up by using
> the next generation of the McKinley processor, but the old
> DEC engineers say the Alpha is by far superior in
> performance.  The regulars  of the Alpha workstations with
> OpenVMS said it would be a downgrade to move to the Intel
> processor.

Also, the fact is, the Alpha chip wasn't broken, so why does Compaq want
to replace it? it has all the promised performance of the Itanium, and
it has actually been delivered. Intel would have been much better off
buying the Alpha chip and build upon it a next generation processor, one
that is based on proven technology with proven results.

Matthew gardiner

--
Wally: "Hold it right there, buddy ... that scruffy beard...
        those suspenders... that smug expression...YOU'RE ONE
        OF THOSE CONDESCENDING UNIX USERS!"
UNIX User: "Here's a nickel, kid. Get yourself a better computer."

 
 
 

1. Linux on itanium 64 bit -- Very poor performance

Has anyone else tested Linux on an itanium server?

I tested RH 7.1 on a Dell 7150 using the Unixbench suite.
Overall performance was abysmal. On all the tests (including
double-precision FP arithmetic) a Pentium III out-performed
the itanium by a factor of 2.

We tried downloading GCC 3, which supposedly has better
itanium support but the results did not improve.

What's going on?
Is RH 7.2 any better?
If so, are there published benchmarks somewhere?

--
Michael Maloney
Ilex Engineering, Inc.

2. HELP - need bug-free version of LDP postscript

3. Cannot boot Linux 2.4.19 for IA 64 on an Itanium 2 single processor box

4. Dual boot / Partition size

5. Linux on itanium 64-bit

6. VMWARE

7. FWD: 64 Days to 64-bit x86 Computing!

8. imake support for shared libraries

9. Fixes to make x86-64 arch compile

10. Joystick timing for x86-64

11. Use MTRRs by default for vesafb on x86-64

12. PATCH: x86-64 typo fixes

13. x86-64 fixes for 2.4.21-pre5