[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-smp
Subject:    Re: Fw: Hp LH3000/6000
From:       Jeff Chua <jchua () fedex ! com>
Date:       2001-02-22 1:51:26
[Download RAW message or body]


This is good info. Thanks Mark!

Just want to share this with everyone.

Jeff.




On Wed, 21 Feb 2001, Mark Hahn wrote:

> > Anyone got linux running on HP LH3000 (4 CPUs) or LH6000 (8 CPUs) ?
>
> as far as I know, these are rather unexceptional machines:
> 4-way based on something like the GX/NX/etc, and 8-way based
> on the corrolary chipset.  Linux has always run on them.
>
> > How about Compaq or other makes?
>
> unless the board vendor (usually not the same as the box vendor)
> has screwed up the bios MPS initialization, linux just runs.
>
> > Any limit on number of CPUs???
>
> last time I looked, ia32 had a limit of 32 CPUs, but that's trivial
> (exists only because nothing above 8-way exists).
>
> but in terms of performance, *YES*, there is a limit:
> in general, even 2-way SMP are memory starved, though it's possible
> to find some codes that are so very cache-friendly that this is
> not true.  this is because 4 and 8-way systems have 4 very hungry
> processors all sharing a bus that can deliver at most 800 MB/s.
> obviously, 200 MB/s is pretty pathetic (even a slow celeron sees
> more memory bandwidth.)
>
> or codes that do nothing but IO, and thus could not care less about
> either dram or cache.  in other words, a mainframe app, where the whole
> point of the box is to stuff it full of 50 or so PCI cards, each with
> a whack of disks hanging off.  there, the only real utility of multiple
> processors is that modern Linux can bind a processor to a particular
> IO device, can support many, many irq's, etc.
>
> regards, mark hahn.
>

-
To unsubscribe from this list: send the line "unsubscribe linux-smp" in
the body of a message to majordomo@vger.kernel.org

[prev in list] [next in list] [prev in thread] [next in thread]