Re: It is almost certain now, INTEL will have 64bit x86 !!
From: Andrew Harrison SUNUK Consultancy (Andrew_No.Harrison_No_at_nospamn.sun.com)
Date: 02/24/04
- Next message: Hunter Goatley: "ANN: FLIST V2.5-3 for VMS"
- Previous message: _at_SendSpamHere.ORG: "Re: Bruce Ellis book OpenVMS Troubleshooting"
- In reply to: Robert Klute: "Re: It is almost certain now, INTEL will have 64bit x86 !!"
- Next in thread: Robert Klute: "Re: It is almost certain now, INTEL will have 64bit x86 !!"
- Reply: Robert Klute: "Re: It is almost certain now, INTEL will have 64bit x86 !!"
- Reply: JF Mezei: "Re: It is almost certain now, INTEL will have 64bit x86 !!"
- Reply: Dirk Munk: "Re: It is almost certain now, INTEL will have 64bit x86 !!"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Date: Tue, 24 Feb 2004 18:24:06 +0000
Robert Klute wrote:
> On Tue, 24 Feb 2004 13:19:36 +0000, Andrew Harrison SUNUK Consultancy
> <Andrew_No.Harrison_No@nospamn.sun.com> wrote:
>
> I didn't say that 32bit systems haven't been popular. In the 1 - 4 way
> space they dominate the market, showing that a most business
> applications can be solved cost effectively with such systems. I merely
> said that the limited address space puts a cap on how many processors
> can be effectively used in a single system.
>
But since this point is demonstrably untrue it seems pointless
to have made it.
Solaris 2.6 is a 32bit OS that does not stop it making use of the
64 CPU's in a Sun E10000, Dynix was a 32bit OS that did not
stop it using the 64 CPU's in the Sequent NUMA-Q box.
3.75 GB for a single process may be a problem for some applications
it is not for most. As an example the majority of Oracle apps apart
from the DBMS are 32bit because they don't benefit from having 64bit
support.
The apparent need for all applications and all systems to be 64bit was
a marketing campaign of Digitals, it was mostly rubbish. As the basis
of a Digital marketing campaign it had to be dubious though obviously
from this conversation some of the rubbish seems to have stuck.
>
>>>No, I didn't miss their announcement. I seem to remember that they are
>>>focusing on 1 - 2 way in the short term with their offering. Still
>>>waiting for proof of an Opteron box bigger than 4 way.
>>>
>>
>>No that is not what Intel announced, though the 1-2 way only may
>>well be HP spin on the announcments.
>>
>>Intel announced that it would be available for all the x86 server
>>CPU's starting with 2 way, with 4way etc arriving later this year.
>
>
> They announced 64-bit support on Prescott (1way), Nocona (2way), and
> Potomac (8way). Potomac isn't forecast to show up for at least a year.
> Both Potomac and the Twin Castle chipset have slipped at least once,
> already.
>
>
>
>>You seem to have lost the plot, the 1-8 way market is where most server
>>revenues are and except in HPC people do not cluster commodity servers
>>in any large quantities to get additional throughput. They buy bigger
>>SMP servers instead because they cost less when you add in the
>>cluster interconnect, SAN and cluster SW/DBMS costs.
>
>
> That's what I have been saying all along. Thank you for agreeing with
> me.
>
You seem to have missed the point, no one is interested in very large
Linux boxes or very large Windows boxes and they are also not going
to create clusters these smaller systems running either OS to try
to acheive the same effect.
>
>>Because of that the number of large Linux servers and large Windows
>>servers is tiny making an Integrity SuperDome 64 running Linux
>>or Windows a system designed to address a market that simply
>>does not exist.
>
>
> The large SQL Server market does exist. HP is happily selling Integrity
> servers running Windows to customers.
>
Hardly, most of the Integrity SuperDomes are running HP-UX and while
there may be very small numbers of customers running instances of
SQL-Server on them using Windows 2003 they hardly create a blip on
anything other than a marketeers radar.
>
>>>I guess, this just means that we can be expecting an announcement from
>>>Sun, any day now, that they are dropping the 15K in favor of V20z
>>>clusters.
>>
>>No for the above reasons.
>
>
> Thought not. So, don't expect HP to drop Itanium any time soon.
>
Again you seem to have entirely missed the point, HP are highly
unlikely to drop Itanium, it would be rather like Turkeys voting
for Christmas/Thanksgiving. You have all your eggs in the Itanium
basket. The issue that is excercising people is how long Intel
will continue to fund Itanium given that it has become increasingly
obvious that Itanium is not going to capture the server market as
Intel expected and nor is it going to make money for Intels
shareholders.
Many people armed with P&L numbers have asked why HP bothers
making anything other than printers including the families
of your co-founders, the same question is being asked about
Intel and Itanium.
Regards
Andrew Harrison
- Next message: Hunter Goatley: "ANN: FLIST V2.5-3 for VMS"
- Previous message: _at_SendSpamHere.ORG: "Re: Bruce Ellis book OpenVMS Troubleshooting"
- In reply to: Robert Klute: "Re: It is almost certain now, INTEL will have 64bit x86 !!"
- Next in thread: Robert Klute: "Re: It is almost certain now, INTEL will have 64bit x86 !!"
- Reply: Robert Klute: "Re: It is almost certain now, INTEL will have 64bit x86 !!"
- Reply: JF Mezei: "Re: It is almost certain now, INTEL will have 64bit x86 !!"
- Reply: Dirk Munk: "Re: It is almost certain now, INTEL will have 64bit x86 !!"
- Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Relevant Pages
|