* Xeon 5500 series support in kernel
@ 2009-07-02 7:45 lchampig
0 siblings, 0 replies; 7+ messages in thread
From: lchampig @ 2009-07-02 7:45 UTC (permalink / raw)
To: linux-kernel
^ permalink raw reply [flat|nested] 7+ messages in thread
* Xeon 5500 series support in kernel
@ 2009-07-02 8:59 xb
2009-07-02 9:46 ` Andi Kleen
0 siblings, 1 reply; 7+ messages in thread
From: xb @ 2009-07-02 8:59 UTC (permalink / raw)
To: Linux Kernel Mailing List
Hi all,
I wonder if there are specific kernel configuration parameters that
allow to optimize performances on the recent Xeon 5500 processors (aka
Nehalem).
In particular, should we set CONFIG_GENERIC_CPU or CONFIG_MCORE2 on such
processors ?
Thanks in advance.
Xavier
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: Xeon 5500 series support in kernel
2009-07-02 8:59 Xeon 5500 series support in kernel xb
@ 2009-07-02 9:46 ` Andi Kleen
2009-07-04 11:41 ` Nix
0 siblings, 1 reply; 7+ messages in thread
From: Andi Kleen @ 2009-07-02 9:46 UTC (permalink / raw)
To: xb; +Cc: Linux Kernel Mailing List
xb <xavier.bru@bull.net> writes:
> I wonder if there are specific kernel configuration parameters that
> allow to optimize performances on the recent Xeon 5500 processors (aka
> Nehalem).
The most important optimization is to use a 64bit kernel with NUMA
support enabled (even if you use 32bit userland)
> In particular, should we set CONFIG_GENERIC_CPU or CONFIG_MCORE2 on
> such processors ?
The kernel doesn't really care about compiler optimizations,
so it shouldn't make any difference. CONFIG_GENERIC_CPU is fine.
-Andi
--
ak@linux.intel.com -- Speaking for myself only.
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: Xeon 5500 series support in kernel
2009-07-02 9:46 ` Andi Kleen
@ 2009-07-04 11:41 ` Nix
2009-07-04 16:07 ` Andi Kleen
0 siblings, 1 reply; 7+ messages in thread
From: Nix @ 2009-07-04 11:41 UTC (permalink / raw)
To: Andi Kleen; +Cc: xb, Linux Kernel Mailing List
On 2 Jul 2009, Andi Kleen outgrape:
> xb <xavier.bru@bull.net> writes:
>
>> I wonder if there are specific kernel configuration parameters that
>> allow to optimize performances on the recent Xeon 5500 processors (aka
>> Nehalem).
>
> The most important optimization is to use a 64bit kernel with NUMA
> support enabled (even if you use 32bit userland)
Surely NUMA only benefits you if you have a >1 socket system? Where's
the nonuniform access on a 1-socket machine? Every core is as far from
all the RAM as every other core, isn't it?
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: Xeon 5500 series support in kernel
2009-07-04 11:41 ` Nix
@ 2009-07-04 16:07 ` Andi Kleen
2009-07-04 18:01 ` Morten P.D. Stevens
0 siblings, 1 reply; 7+ messages in thread
From: Andi Kleen @ 2009-07-04 16:07 UTC (permalink / raw)
To: Nix; +Cc: Andi Kleen, xb, Linux Kernel Mailing List
On Sat, Jul 04, 2009 at 12:41:15PM +0100, Nix wrote:
> On 2 Jul 2009, Andi Kleen outgrape:
>
> > xb <xavier.bru@bull.net> writes:
> >
> >> I wonder if there are specific kernel configuration parameters that
> >> allow to optimize performances on the recent Xeon 5500 processors (aka
> >> Nehalem).
> >
> > The most important optimization is to use a 64bit kernel with NUMA
> > support enabled (even if you use 32bit userland)
>
> Surely NUMA only benefits you if you have a >1 socket system? Where's
Xeons are typically multi socket. On single socket the overhead
of a NUMA kernel is negligible, so it's best to enable it always.
There are actually some use cases of NUMA kernel code for single
socket too, e.g. it is sometimes used to isolate different
applications on the VM level.
-Andi
--
ak@linux.intel.com -- Speaking for myself only.
^ permalink raw reply [flat|nested] 7+ messages in thread
* RE: Xeon 5500 series support in kernel
2009-07-04 16:07 ` Andi Kleen
@ 2009-07-04 18:01 ` Morten P.D. Stevens
2009-07-05 9:55 ` Andi Kleen
0 siblings, 1 reply; 7+ messages in thread
From: Morten P.D. Stevens @ 2009-07-04 18:01 UTC (permalink / raw)
To: Andi Kleen; +Cc: linux-kernel@vger.kernel.org
Hi Andi,
a general question with regard to the new intel nehalem platform:
Which linux version is needed for an intel nehalem based xeon in view of the new intel 5500/5520 chipset drivers?
Does it work with 2.6.18? (Debian Etch)
thanks
--
Morten
^ permalink raw reply [flat|nested] 7+ messages in thread
* Re: Xeon 5500 series support in kernel
2009-07-04 18:01 ` Morten P.D. Stevens
@ 2009-07-05 9:55 ` Andi Kleen
0 siblings, 0 replies; 7+ messages in thread
From: Andi Kleen @ 2009-07-05 9:55 UTC (permalink / raw)
To: Morten P.D. Stevens; +Cc: Andi Kleen, linux-kernel@vger.kernel.org
On Sat, Jul 04, 2009 at 08:01:44PM +0200, Morten P.D. Stevens wrote:
> Hi Andi,
>
> a general question with regard to the new intel nehalem platform:
>
> Which linux version is needed for an intel nehalem based xeon in view of the new intel 5500/5520 chipset drivers?
>
> Does it work with 2.6.18? (Debian Etch)
You would need backported drivers for Ethernet etc; the plain
integrated drivers in 2.6.18 will not work. The e1000 project
at sourceforge has driver backports for Ethernet.
I don't know if Debian does regular backports of drivers,
some other distributions with support for old kernels do.
Also some power saving features and performance features
which are related to power saving (turbo mode) will not work in that old a
kernel. For full performance it's better to use newer kernels or
kernels with appropiate backports.
-Andi
--
ak@linux.intel.com -- Speaking for myself only.
^ permalink raw reply [flat|nested] 7+ messages in thread
end of thread, other threads:[~2009-07-05 9:55 UTC | newest]
Thread overview: 7+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2009-07-02 8:59 Xeon 5500 series support in kernel xb
2009-07-02 9:46 ` Andi Kleen
2009-07-04 11:41 ` Nix
2009-07-04 16:07 ` Andi Kleen
2009-07-04 18:01 ` Morten P.D. Stevens
2009-07-05 9:55 ` Andi Kleen
-- strict thread matches above, loose matches on Subject: below --
2009-07-02 7:45 lchampig
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox