Pentium vs. Itanium

* Rick Jones:
Benjamin Gawert wrote:
It is dead, even HP (the biggest Itanium vendor) knows this....

We do?

You should. At least our HP contacts already admit that even inside HP
it seems to be clear that Itanium figures aren't going to rise. The
delay of Montecito is just a small part of this misere...

Which is sad because technically the Itaniums are great, and we're
extremely satisfied with it. But for some time now it lacks a severe
performance increase while everything else around it gets faster and
faster. Montecito is just too little too late...

Benjamin

"Yousuf Khan" wrote in message
...
***** charles wrote:
It is too bad about the Itanium. It has "some" good engineering
in it and it was supposed to last for the "next 20 years". If only
Intel could fiture out how to make them CHEAP. The biggest
problem is price that is why AMD64 and EM64T are both
blowing it out of the water.

Not really, Intel had the ability to sell it for very cheap, if it
wanted to. Initially, it could've subsidized the price of the Itaniums,
and then eventually it could've gotten the price down by simply
migrating it to its latest miniaturization process node. However,
neither option would've proved to be fruitful, as a cheap processor that
nobody wants will only sell slightly better than an expensive processor
that nobody wants.

It's major problem was, and is, and will always be it's lack of software
compatibility. If Itanium had the ability to efficiently run 32-bit x86
software like the Opteron did, then it would've been only a matter of
biding time until 64-bit applications start coming in. Opteron simply
ran existing 32-bit apps happily, while people slowly (and sometimes
speedily) ported those apps to 64-bit.

Yousuf Khan

I wish I could remember where I read an article about the transitioning
process from one arch to another since it was the writers' premise that
the ability to run old software was NOT a hurdle for the adoption of
the newer arch. The writer proposed 3 or 4 cases for his conclusion.
I think Apple was one were it has gone through 3 major transitions:
Apple II, Power and now Intel x86.

IA-64 had the ability in the beginning to run x86 software but that
ability was dropped because the demand was so low in its' customer
base, people who actually bought one.

Everyone (people, organizations, governments, etc...) is price sensative.
If it was cheaper and faster, it would sell.

It would be interesting to see what the percentages are for OS's:
Linux
Windows
and HP-UX in the existing owner space.

later,
charles.....

"Benjamin Gawert" wrote in message
...
* Rick Jones:
Benjamin Gawert wrote:
It is dead, even HP (the biggest Itanium vendor) knows this....

We do?

You should. At least our HP contacts already admit that even inside HP
it seems to be clear that Itanium figures aren't going to rise. The
delay of Montecito is just a small part of this misere...

Which is sad because technically the Itaniums are great, and we're
extremely satisfied with it. But for some time now it lacks a severe
performance increase while everything else around it gets faster and
faster. Montecito is just too little too late...

Benjamin

I have tried to buy one of them at several times during the last 2 or 3
years. First you have to "qualify" and HP sends you to an internal
rep in HP. Finally you can not just go to a website and order one
and have it shipped to you. So, the purchase process is a HUGE
hurdle. Then the prices are way out of wack compared to comperable
x86-64 clones either from amd or intel. I hear that it runs Linux just
great. My proposal to HP/Intel/whoever is to drop the hurcles to
purchase, drop the prices according to Yousuf and sell them through
the Internet to anyone who has a credit card or will send a check.
Another idea it to design and make a motherboard that is IA-64
based that uses an ATX and or BTX motherboard that can handle
other clone parts to make an Itanium clone computer. Sell it on
places like Pricewatch just like all the other motherboards.
That along with a better marketing campaign would turn it around.

later.....

* ***** charles:

I have tried to buy one of them at several times during the last 2 or
3 years. First you have to "qualify" and HP sends you to an
internal rep in HP. Finally you can not just go to a website and
order one and have it shipped to you. So, the purchase process is a
HUGE hurdle. Then the prices are way out of wack compared to
comperable x86-64 clones either from amd or intel.

Not really. At least here in Germany buying Itanium is totally easy.
Besides HP itself there are dozens of resellers like MCL that do
Itanium. And the price really isn't an issue as a zx2000 is sub-1kEUR
and a rx26x0 is sub-2kEUR - both with decent CPU, RAM, and HP RENEW with
3yrs onsite.

I hear that it runs Linux just great.

As it does with Windowsxp/Windows Server 2003 64bit Edition, and it runs
HP-UX even better. Honestly, if you want Linux then x64 would the the
better choice, simply because there are much more distributions for x64
than for IA64...

My proposal to HP/Intel/whoever is to drop the hurcles to purchase,
drop the prices according to Yousuf and sell them through the
Internet to anyone who has a credit card or will send a check.

Yeah, right. As if people would buy Itanium servers just by a click on a
website ;-)

Another idea it to design and make a motherboard that is IA-64 based
that uses an ATX and or BTX motherboard that can handle other clone
parts to make an Itanium clone computer.

You mean something like that
http://www.supermicro.com/products/motherboard/Itanium/

Sell it on places like
Pricewatch just like all the other motherboards. That along with a
better marketing campaign would turn it around.

You mus be kidding, right? Besides the fact that Itaniums performance is
stagnating for some time now while x64 gets faster and faster the fact
that it doesn't run acceptable with the majority of 32bit software is
just one of the little annoying things that would prevent average user
from buying it. The fact that it's also difficult to program for Itanium
is just a second one. With no standard software and insufficient
compatibility, who in his right mind would buy a more expensive platform
which is even slower than the much cheaper x64 which also supports all
generic applications out there?

Your suggestion has the same quality like the ones that from time to
time suggest that SGI or Sun should bring out a low-end MIPS/SPARC
workstation for the consumer market to gain marketshare. These ideas
look good, at least as long as you ignore the real world and its
economics out there...

Benjamin

* ***** charles:

I wish I could remember where I read an article about the transitioning
process from one arch to another since it was the writers' premise that
the ability to run old software was NOT a hurdle for the adoption of
the newer arch.

Well, this writer was wrong. The very bad 32bit performance when running
x86 code definitely was a very bad thing for Itanium...

The writer proposed 3 or 4 cases for his conclusion.
I think Apple was one were it has gone through 3 major transitions:
Apple II, Power and now Intel x86.

Right, but Apple also always provided a very good application
compatibility...

IA-64 had the ability in the beginning to run x86 software but that
ability was dropped because the demand was so low in its' customer
base, people who actually bought one.

It was reduced (not dropped since that isn't possible). But you miss the
reason for that. Itanium has a very poor performance when running 32bit
x86 code, and no-one wanted to spend thousands of dollars for a 64bit
machine that runs his 32bit applications at P2-266 speed. Since the
majority of desktop applications is 32bit Itanium had already lost from
the beginning there. So with low 32bit performance Itanium simply was
uninteresting for the desktop market, and because Itanium was
uninteresting for the desktop market no ISV wanted to invest money to
develop 64bit applications for IA64 and Windowsxp 64bit. So Windowsxp on
Itanium was terminated since no-one bought it. With Windowsxp being
terminated the primary remaining OSes for IA64 were HP-UX and Linux (IBM
terminated AIX for IA64, Solaris for IA64 was killed even during the
alpha stage). But Linux was much better on x86 and later on x64, so
there remained HP-UX. But the traditional workstation market was already
shrinking, and no ISV who offered his desktop software for PA-RISC
wanted to take the risk and port it to IA64, so HP finally killed the
PA-RISC workstation...

In short, the 32bit execution layer lost importance since the audience
that needed that feature already abandoned Itanium because of the slow
32bit performance. And in the market where Itanium is today (midrange to
highend servers and supercomputers) the relevance of running x86 code is
nil...

Everyone (people, organizations, governments, etc...) is price sensative.
If it was cheaper and faster, it would sell.

Yeah, right. I know the government market quite well but there are other
criterias for them than pricing. Besides that, Itanium has its place in
the highend server market. But as x64 gets better the days Itanium can
keep this niche are countable...

It would be interesting to see what the percentages are for OS's:
Linux
Windows
and HP-UX in the existing owner space.

That depends on the system vendor. Most Itanium systems made by HP
probably run HP-UX (and since HP is the biggest Itanium vendor it's very
likely that most Itanium systems out there run HP-UX, too). Then there
are HPs NonStop servers running NonStop-Unix but they are sold in very
few numbers. Most of the few other vendors selling Itanium use some kind
of Linux...

Benjamin

* Benjamin Gawert:
But the traditional workstation market was already
shrinking, and no ISV who offered his desktop software for PA-RISC
wanted to take the risk and port it to IA64, so HP finally killed the
PA-RISC workstation...

Sorry, of course HP killed their Itanium workstations...

Benjamin

Rick Jones wrote:

wrote:
In article ,
says...
Benjamin Gawert wrote:
It is dead, even HP (the biggest Itanium vendor) knows this....
We do?
Didn't you get the memo? :-)


He's the guy that talks to the customers so
the programmers don't have to.

Benjamin Gawert wrote:
* Benjamin Gawert:
But the traditional workstation market was already shrinking, and no
ISV who offered his desktop software for PA-RISC wanted to take the
risk and port it to IA64, so HP finally killed the PA-RISC workstation...

Sorry, of course HP killed their Itanium workstations...

Microsoft Windows expects graphics, and OpenVMS I64 DECwindows (X
Windows) works just fine with current Integrity server configurations,
for instance, and readily available displays. HP-UX also has graphics
capabilities.

All these operate on Integrity servers with Itanium processors.

FWIW, one of the low-end servers in the Integrity series (the rx2600)
basically started out as the graphics workstation box (the zx6000), and
the second generation of the box (the rx2620) is available in the server
line. And yes, with graphics.

The usual choices are an optional add-on graphics controller, or the
graphics that are part of the Management Processor (MP). The MP is a
component of various of the Integrity servers, and optional on a few
other of the series. And DECwindows runs on the MP graphics just fine,
for instance.

Unfortunately, the term "workstation" is a whole lot like "real-time"
-- it's ambiguous. Without some context around the particular
requirements or the particular application, "workstation" is ambiguous
at best and not meaningful at worst. If you need or want to use your
"workstation" for status displays, typical X Windows operations, or as
part of a software development interface, for instance, these
"workstation" products are certainly available in the Integrity product
line.

* Hoff Hoffman:

Microsoft Windows expects graphics, and OpenVMS I64 DECwindows (X
Windows) works just fine with current Integrity server configurations,
for instance, and readily available displays. HP-UX also has graphics
capabilities.

Of course these machines do have some gfx but that is not really
comparable to what was available for the workstations...

All these operate on Integrity servers with Itanium processors.

Yeah, fine. And what 3D adapters are supported on any of the Itanium
servers? Exactly, zero.

FWIW, one of the low-end servers in the Integrity series (the rx2600)
basically started out as the graphics workstation box (the zx6000),

Nope. zx6000 and rx2600 are indeed _very_ similar but not identical
(zx6000 has AGP, the rx2600 is PCI only)...

and
the second generation of the box (the rx2620) is available in the server
line. And yes, with graphics.

And yes, still 2D-only gfx, and yes, still no fast gfx adapter or AGP.
So what?

The usual choices are an optional add-on graphics controller, or the
graphics that are part of the Management Processor (MP). The MP is a
component of various of the Integrity servers, and optional on a few
other of the series. And DECwindows runs on the MP graphics just fine,
for instance.

Ever tried to do something 3D? Of course not. Moving a few Windows from
one corner to another is all what the server gfx adapters are for...

Unfortunately, the term "workstation" is a whole lot like "real-time"
-- it's ambiguous. Without some context around the particular
requirements or the particular application, "workstation" is ambiguous
at best and not meaningful at worst. If you need or want to use your
"workstation" for status displays, typical X Windows operations, or as
part of a software development interface, for instance, these
"workstation" products are certainly available in the Integrity product
line.

Only that they still are servers without any workstation attribute -
mainly a wide choice of gfx adapters, a fast gfx interface (i.e. AGP),
and (probably most important) the lack of ISVs that offer their
workstation software products for Itanium...

Benjamin