Pentium-D support

Pentium-D support

[Bill Davidsen]

Since this is really a question in several areas I'll put it here. Now 
that the Pentium-D processors are available at reasonable prices and for 
quick delivery, can anyone speak to the ACPI issues? The available 
boards use the 945 and 955 chipset. Is there any reason to think that 
the scheduler would get confused by the CPU, such as thinking it was HT 
or some such?

The specs indicate that 64 bit is supported, is there any actualy Linux 
support for the Intel 64 bit stuff in gcc and the kernel? One of the 
people I work with reports that the distro he runs on his Athlon64 lock 
solid after reading the boot sector, so obviously this isn't Athlon 
compatable.

The price is lower than a dual Xeon setup if you have an application 
which needs SMP, and initial power values make it look like a lower 
power solution overall.
-- 
    -bill davidsen (davidsen@tmr.com)
"The secret to procrastination is to put things off until the
  last possible moment - but no longer"  -me

Re: Pentium-D support

[Michael Thonke]

Bill Davidsen schrieb:

> Since this is really a question in several areas I'll put it here. Now
> that the Pentium-D processors are available at reasonable prices and
> for quick delivery, can anyone speak to the ACPI issues? The available
> boards use the 945 and 955 chipset. Is there any reason to think that
> the scheduler would get confused by the CPU, such as thinking it was
> HT or some such?

There are only issues with CPUFREQ/speedstep and some ACPI related
things on Intel 955X chipset I have. So I use a ASUS P5WD2-Premium.
The Pentium D will reach me in few days so I only guess and  I think the
kernel know how to handle Physical Cores and SMT/HT Processors.

>
> The specs indicate that 64 bit is supported, is there any actualy
> Linux support for the Intel 64 bit stuff in gcc and the kernel? One of
> the people I work with reports that the distro he runs on his Athlon64
> lock solid after reading the boot sector, so obviously this isn't
> Athlon compatable.
>
Yes the 64bit support of the Intel CPUs are well supported, I use a
Intel Pentium 4 640 and 64bit compiled system (Gentoo 2005)
I have no problems everything is working nearly perfect better as my
AMD64 system.

For GCC 3.4.x or 4.0.x you only need the compile switch -march=nocona as
the name of the XEON..but its the same piece of technologie

> The price is lower than a dual Xeon setup if you have an application
> which needs SMP, and initial power values make it look like a lower
> power solution overall.

Cheep and good performance.

Greets
Best regards
    Michael

Re: Pentium-D support

[Bill Davidsen]

Michael Thonke wrote:
> Bill Davidsen schrieb:
> 
> 
>>Since this is really a question in several areas I'll put it here. Now
>>that the Pentium-D processors are available at reasonable prices and
>>for quick delivery, can anyone speak to the ACPI issues? The available
>>boards use the 945 and 955 chipset. Is there any reason to think that
>>the scheduler would get confused by the CPU, such as thinking it was
>>HT or some such?
> 
> 
> There are only issues with CPUFREQ/speedstep and some ACPI related
> things on Intel 955X chipset I have. So I use a ASUS P5WD2-Premium.
> The Pentium D will reach me in few days so I only guess and  I think the
> kernel know how to handle Physical Cores and SMT/HT Processors.
> 
> 
>>The specs indicate that 64 bit is supported, is there any actualy
>>Linux support for the Intel 64 bit stuff in gcc and the kernel? One of
>>the people I work with reports that the distro he runs on his Athlon64
>>lock solid after reading the boot sector, so obviously this isn't
>>Athlon compatable.
>>
> 
> Yes the 64bit support of the Intel CPUs are well supported, I use a
> Intel Pentium 4 640 and 64bit compiled system (Gentoo 2005)
> I have no problems everything is working nearly perfect better as my
> AMD64 system.
> 
> For GCC 3.4.x or 4.0.x you only need the compile switch -march=nocona as
> the name of the XEON..but its the same piece of technologie

Okay, unless I can find an Intel-64 build of Fedora I'll have to do it 
myself, but now that I know what option to use it's possible. Trust 
Intel to have a 64 bit standard which isn't Itanium and isn't ..quite.. 
AMD compatible.
> 
> 
>>The price is lower than a dual Xeon setup if you have an application
>>which needs SMP, and initial power values make it look like a lower
>>power solution overall.
> 
> 
> Cheep and good performance.

Thanks for the pointer!
-- 
    -bill davidsen (davidsen@tmr.com)
"The secret to procrastination is to put things off until the
  last possible moment - but no longer"  -me

Re: Pentium-D support

[Dave Jones]

On Tue, Jun 07, 2005 at 01:45:43PM -0400, Bill Davidsen wrote:

 > >For GCC 3.4.x or 4.0.x you only need the compile switch -march=nocona as
 > >the name of the XEON..but its the same piece of technologie
 > 
 > Okay, unless I can find an Intel-64 build of Fedora I'll have to do it 
 > myself, but now that I know what option to use it's possible.

The standard x86-64 build of Fedora will work just fine.
(I use it daily on em64t)

		Dave

Re: Pentium-D support

[Lennart Sorensen]

On Tue, Jun 07, 2005 at 01:45:43PM -0400, Bill Davidsen wrote:
> Okay, unless I can find an Intel-64 build of Fedora I'll have to do it 
> myself, but now that I know what option to use it's possible. Trust 
> Intel to have a 64 bit standard which isn't Itanium and isn't ..quite.. 
> AMD compatible.

It should be instructionset compatible (at least almost) but
optimization wise it is different since it is a very different design.

Certainly Debian's amd64 runs on both just fine, but recomends different
kernel builds for each (but also has one generic that runs on both).

Len Sorensen

Re: Pentium-D support

[Siddha, Suresh B]

On Tue, Jun 07, 2005 at 11:06:50AM -0400, Bill Davidsen wrote:
> Is there any reason to think that 
> the scheduler would get confused by the CPU, such as thinking it was HT 
> or some such?

2.6.12-rc5 has the Intel dual core detection patches. With those, kernel
can distinguish between cores and HT logical siblings.

thanks,
suresh