Re: Mip-o-suction (horrible performance)
On Jul 10, 2006, at 1:16 PM, Ron Johnson wrote:
[...snip...]
Anyway, the problem goes away with a 2.6 kernel. This is with no
changes to BIOS or any of the package configurations. This
gives _me_ an acceptable solution....for now.
What do you mean "for now"?
I mean, until I either understand what's going on, or make my peace
with it. It also raised a red-flag about performance under load. I
just don't like ignoring red flags.
... there are those nagging doubts. Surely I'm not the only one
who's seen such behavior. From a community perspective it's
troubling to see a basic, no-frills, stable installation on a
name-brand machine produce such horrible results.
Maybe NCR changed at some point, but I do know that they were always
hot for *highly* customized systems. "Added value", it's called.
The mfgr is NEC, not NCR.
The fact that the 2.6 kernel fixes it strongly suggests to me that
the PG350 *is* such an oddball, where a work-around was only put
into 2.6.
...or its could be symptomatic of a race condition, which (under
light loads) fails in 2.4 but succeeds in 2.6. Under heavier loads,
who knows? All-in-all I much prefer affirmative identification of
problems.
And THAT's why I inquried about diagnostics in my first post. I'd
like to verify that the problem doesn't recur.
It's shaken
my confidence that the problem will not reappear at the most
inopportune time.
Oh, puh-leeze. Mobo, chipset and card makers all do some pretty
weird shit, and it's not easy for Linux to handle every combination
of custom chipsets.
I would hope it does at least as good a job as Windoze on the same
machine.
Hmm..is there a utility to report the chipset/driver config, as
resolved by the kernel? Is there a way to get a report of the
scheduler's activity?
Reply to: