Re: Quad AMD opteron system and Linux

From: SIOL (brankob_at_avtomatika.com)
Date: 09/18/03 

Date: Thu, 18 Sep 2003 19:28:03 +0200

"Pankajkumar Chauhan" <no_spam@dev.null> wrote in message
news:ty4smmu12zk.fsf@bertie.modck.cs.cmu.edu...

<SNIP>

> These are valid concerns.
>
> Here's the reasoning of my group in brief
> for getting a single box.

> - We need large amount of memory, even on a single dual opteron
> machine, we'll need about the same amount of memory, to accomodate our
> processes. Most of the system cost comes from memory.

But you have said that you would use 4-way machine with 8 Gb RAM. Given that
Opteron uses local RAM, that means 2 Gb per CPU.
It's true that Opterons can do memory synchronisation through HT, but heavy
use of that feature would probably kill much of the CPU speed...

>
> - There's a fixed support cost per box (power, b/w, maintenance, software
> install) that the department facilities charges. With 10
> boxes, the support cost would be 10 times higher. On average, it is
> about $80 a month per box.

O.K. I see that it wouldn't be optimal for your case. But for a power user
in small company it could be the best deal around.
Four dual CPU Opterons or even eight heavily overclocked A64 could be an
interesting option.
Not quite real SMP, but lightyears ahead of networked computing...

What I'm asking is, is it possible to do something like that and how ?

Branko

Relevant Pages

  • Re: pcib allocation failure
    ... pcib1: attempting to grow prefetch window for ... attempting to grow memory window for ... cpu0: on acpi0 ... <ACPI PCI bus> on pcib0 ...
    (freebsd-current)
  • Next July 27: boot failure(hang) on x86_64 box.
    ... Freeing unused kernel memory: 1360k freed ... ACPI: PM-Timer IO Port: 0x488 ... CPU: L2 Cache: 1024K ... # AX.25 network device drivers ...
    (Linux-Kernel)
  • [PATCH] Document Linuxs memory barriers [try #3]
    ... The attached patch documents the Linux kernel's memory barriers. ... I've tried to get rid of the concept of memory accesses appearing on the bus; ... barring implicit enforcement by the CPU. ...
    (Linux-Kernel)
  • Oops in 2.6.28-rc9 and -rc8 -- mtrr issues / e1000e
    ... Bios 1.04beta did show correct memory sizing in dmidecode, ... I hope this is as simple as me doing something glaringly wrong in the kernel ... DMI present. ... CPU: L2 cache: 6144K ...
    (Linux-Kernel)
  • Re: read vs. mmap (or io vs. page faults)
    ... not fit in main memory, and there are overheads related to the heuristics ... But because the CPU is underutilized, ... reasonably sized user buffer). ... You have to measure the actual overhead to see what the actual cost is. ...
    (freebsd-questions)