Re: Intel to chip away at Itanium prices <- or ... I want my cheap

From: Andrew Harrison SUNUK Consultancy (Andrew_No.Harrison_No_at_nospamn.sun.com)
Date: 01/28/04 

Date: Wed, 28 Jan 2004 18:53:45 +0000

Robert Klute wrote:
> On Wed, 28 Jan 2004 13:24:20 +0000, Andrew Harrison SUNUK Consultancy
> <Andrew_No.Harrison_No@nospamn.sun.com> wrote:
>
>
>
>>The reasons for abandoning IA64 on the desktop are pretty obvious.
>>
>>1. Its too hot, some of the professional Computer gaming
>> people may be happy to lug an external cooler arround to
>> cope with the overclocked CPu they are using but most
>> desktop users arn't.
>
>
> My zx2000 didn't come with an external cooler. Do I need to complain
> that I didn't get one?
>

You zx2000 is hardly a desktop system its a deskside
tower ~19x11x19 (inches) and quite a big one
at that.

It has a maximum power requirement of 650 watts with
fans to match.

Compare that with an HP D530 4x13x15 in
maximum power requirement of 185 watts.

>
>>2. Its too expensive, you need to be able to produce a ~1K
>> or less desktop and you cannot do that with Itanium.
>
>
> Why? Every IA-32 workstation I have priced out costs a lot more than
> $1k, once you configure more than the barebones teaser configuration.
>

The z2000 with a CPU that can compete ish with the IA32
costs ~5K with 1 GB of RAM and a 1.4 GHZ 1.5 MB cache
Itanium II. I say Ish because it isn't a fast processor
compared with IA32 or Opteron except on FP.

The D530 with a 3.2 GHZ processor, 1 GB of RAM etc is
$1300 and thats just shopping from HP, Dell are cheaper
and its faster than your z2000 on Int by a wide margin.

Integer SPECint
1261 for the x86 vs 824

FP is better SPECfp
1267 for the x86 vs 1444

Since what most people use on corporate desktops is
integer based you are looking at a huge price performance
dissadvantage for the Itanium based system for the apps
that matter to customers.

>
>>3. Its not integrated enough. It has onchip cache but cannot
>> hope to compete price wise with Opteron/SPARC which
>> have that plus memory controlers and interconnects.
>>
>>4. It has a vanishingly tiny software portfolio and a terrible
>> x86 emulation story (even when HP introduce the faster
>> x86 support). While it may be possible to control the
>> software stack on a server it becomes much more difficult
>> on the desktop and any Itanium based system is always going
>> to be suffering from the it doesn't run on Itanium problem.
>
>
> I guess that is Intel's curse with Itanium - it is continually compared
> to IA-32 when it was designed to compete with non IA-32 RISC
> architectures. SPARC doesn't run x86 code and no seems to care.
>

We don't have to we have the 2and largest SW portfolio after
WIN32/x86 on SPARC it would be nice to get access to x86
SW portfolio as well but in the server space SPARC is not
short of apps, the opposite in fact of Itanium.

And if you are looking at a desktop platform then having access
to the x86 SW catalogue is very important.

>
>>5. The mid term solution to the heat/cost issue (dearfield)
>> sacrifices performance to reduce heat and cost making
>> Itanium even less competitive compared with x86 or x86-64
>> based systems.
>>
>>6. Corporate desktops are a huge support hole, having to maintain
>> an entirely separate application tree for Itanium is likely
>> to disuade many corporates from moving, x86-64 for example
>> does not have this issue.
>>
>>
>>>Had Intel not said that IA64 wasn't going to be for desktop, it might have had
>>>a chance at making IA64 industry standard, especially if AMD and others were
>>>allowed to copy it. (multiple source is an important factor when defining
>>>"industry standard").
>
>
> Intel has a history of saying that their new chip is for servers and not
> the desktop. I remember they did that when the 80286 came out and then
> again when the 80386 came out.

Except this time Intel said that Itanium is for everything and
then when the reality of their situation slapped them hard in
the face they backtracked to a slightly more possible claim
for Itanium being a server CPU.

regards
Andrew Harrison

Relevant Pages

  • Re: HP Dumps the Itanic
    ... > event for the Itanic platform. ... base still bought more PA-RISC workstations than Itanium desktops. ... Itanium sells quite well in the server market. ... The last time I heard that Itanium should replace x86 has been before the ...
    (comp.sys.sgi.misc)
  • Re: VC++6 LNK1112 Error
    ... Itanium into a piece of code destined for an x86. ... correctly select your target to be an Itanium. ... Note also that you can't really compile IA64 code unless you have a compiler that targets ...
    (microsoft.public.vc.mfc)
  • Re: Question to Kerry Main
    ... "The Itanium is great, it's got a future, and it's selling pretty well ... to be more accurate some in the trade press are reporting that Intel certainly *wants* people to think that Itanic has a future: if you read the above carefully all of it refers to Intel's message rather than expressing the opinion of the author. ... So Itanic has finally managed to acquire a sufficient percentage of the former PA-RISC market, the remnants of the Alpha market, and peripheral contributions from Windows, Linux, and the seven dwarfs who adopted Itanic along with HP to match the nearly 1/3 share of the market that PA-RISC used to enjoy all by itself. ... There's no hope that Itanic will even begin reducing the devastating performance lead that POWER currently enjoys for about the next two years until Tukwila arrives with CSI - and the tepid projection that Tukwila will only double Montecito's performance suggests that on a per-core basis POWER's lead on a per-core basis may instead continue to grow - so, just as has been the case since its ill-fated debut, Itanic seems destined to continue to be sandwiched between ever-increasing x86 capabilities on the low end and unassailable competition on the high end: exactly the opposite of the situation that Intel and HP originally envisioned for it, and hardly a recipe for the kind of overall market domination they invested their $billions to achieve. ...
    (comp.os.vms)
  • Re: Intel to sell itanium?
    ... I expect them both to lose to SPARC at the high end and SPARC and ... both Sun and Fujitsu outperforms Itanium on a per core basis for almost ... In terms of scale the largest SPARC server ... that market dual core AMD Opteron based servers require a minimum of 2 ...
    (comp.os.vms)
  • Re: Fazing out x86
    ... His main point has little to do with Itanium*. ... Due to the competitive market, x86 will never go away as long as customers keep paying for it; if any player tries to leave, someone else will step in to take the money and new players will emerge. ... HP has learned how to handle an ISA transition, and I'm sure they'd also study up on how Apple handled things after Motorola and then IBM decided not to build the PPC products they were depending on. ...
    (comp.arch)