All of lore.kernel.org
 help / color / mirror / Atom feed
From: Andi Kleen <ak@muc.de>
To: "Justin M. Forbes" <jmforbes@linuxtx.org>
Cc: sander@humilis.net,
	Kernel Mailing List <linux-kernel@vger.kernel.org>,
	"Sergey S. Kostyliov" <rathamahata@ehouse.ru>
Subject: Re: NUMA or not on dual Opteron
Date: Thu, 13 Jan 2005 20:43:54 +0100	[thread overview]
Message-ID: <m1k6qh2jyd.fsf@muc.de> (raw)
In-Reply-To: <20050113170733.GA14524@linuxtx.org> (Justin M. Forbes's message of "Thu, 13 Jan 2005 11:07:33 -0600")

"Justin M. Forbes" <jmforbes@linuxtx.org> writes:


> This is somewhat true.  There are 2 types of dual opteron boards. Those in
> the $200 US range only have one memory bank, which is attached to CPU0.
> They operate as a single node, and may perform better with numa turned off.
> Those in the $400+ range tend to have one bank per CPU and will certainly
> perform better with numa on.  They do usually have a bios option to
> interleave the nodes which would show up as a single node, and probably
> perform better with numa turned off, but a better solution is to turn off
> the node interleave in bios and run the kernel with numa support.
> Basically if you have 2 CPUs and only one memory bank, maybe turning numa
> off will give better performance, but if you have one memory bank per CPU
> numa should be on.

[agreed, except:]

There is no significant performance difference between NUMA on and off
on the crippled boards. So best is to always enable CONFIG_NUMA on x86-64. 
The distribution kernels for x86-64 should ship in this configuration
too.

I would recommend to always disable node interleaved on the non crippled
board with multiple memory banks. Most boards ship with this disabled by 
default, only a few enable it.

-Andi

  reply	other threads:[~2005-01-13 19:50 UTC|newest]

Thread overview: 40+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2005-01-12  5:09 Linux 2.6.11-rc1 Linus Torvalds
2005-01-12  7:07 ` Keith Owens
2005-01-12  7:07   ` Keith Owens
2005-01-12  7:32 ` Markus Trippelsdorf
2005-01-13 22:37   ` Linux 2.6.11-rc1 (ACPI related problems) Hanspeter Kunz
2005-01-12  8:24 ` Linux 2.6.11-rc1 Brice Goglin
2005-01-12  9:20   ` Tino Keitel
2005-01-12  9:38     ` Brice Goglin
2005-01-12 14:13       ` Dmitry Torokhov
2005-01-12 14:35         ` Brice Goglin
2005-01-12 15:50   ` Linus Torvalds
2005-01-12 19:13     ` Vojtech Pavlik
2005-01-12 15:24 ` Sergey S. Kostyliov
2005-01-12 15:32   ` Linus Torvalds
2005-01-12 15:44     ` Gene Heskett
2005-01-12 16:03     ` Sergey S. Kostyliov
2005-01-13  9:45     ` NUMA or not on dual Opteron (was: Re: Linux 2.6.11-rc1) Sander
2005-01-13 10:02       ` NUMA or not on dual Opteron Raphael Jacquot
2005-01-13 17:07         ` Justin M. Forbes
2005-01-13 19:43           ` Andi Kleen [this message]
2005-01-13 19:40         ` Andi Kleen
2005-01-13 15:36       ` NUMA or not on dual Opteron (was: Re: Linux 2.6.11-rc1) Alan Cox
2005-01-13 19:38       ` NUMA or not on dual Opteron Andi Kleen
2005-01-15 23:42         ` Sander
2005-01-13 21:24       ` Bernd Eckenfels
2005-01-14  8:04         ` David Schwartz
2005-01-14  8:10           ` Arjan van de Ven
2005-01-14  8:48           ` Andi Kleen
2005-01-12 19:06 ` Linux 2.6.11-rc1 -- usb_storage and Genesys Jan De Luyck
2005-01-13  5:42 ` [PATCH] contort getdents64 to pacify gcc-2.96 Adam Kropelin
2005-01-13  3:51   ` Linus Torvalds
2005-01-13 23:15 ` Linux 2.6.11-rc1 (compile stats) John Cherry
2005-01-16  9:22 ` Cross-compilation broken (was: Re: Linux 2.6.11-rc1) Geert Uytterhoeven
2005-01-16 16:09   ` Sam Ravnborg
2005-01-17  9:00     ` Geert Uytterhoeven
2005-01-19 21:07 ` Linux 2.6.11-rc1 Daniel Gryniewicz
2005-01-20  4:16   ` Dmitry Torokhov
2005-01-20  4:49     ` Daniel Gryniewicz
2005-01-20  7:17       ` Peter Osterlund
2005-01-20 15:01         ` Romano Giannetti

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=m1k6qh2jyd.fsf@muc.de \
    --to=ak@muc.de \
    --cc=jmforbes@linuxtx.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=rathamahata@ehouse.ru \
    --cc=sander@humilis.net \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.