On Saturday 18 August 2001 10:52, Kirby Urner wrote:
> > YMMV, all the above is nothin but my own conjecture and opinions
> >
> >--
> >Tom Brinkman Galveston Bay
>
> I appreciate all the info. The impression I got from
> Civileme's post is that detection of this KT133? chipset
> is going to put the kernel in hobbled mode on boot,
> sans udma with real impact on performance, e.g. in the
> realm of cdr-burning.
>
> Sounded to me like it didn't matter what other variables
> might be in play, i.e. I can count on this behavior given
> my chipset ("no ifs-ands-or-buts"). If so (civileme
> sounds authoritative and posts from mandrake.org), then
> really *not* too much is being made of the problem, from
> my point of view (a) as any board with this chipset is
> getting the same workaround and (b) the workaround doesn't
> really restore the board to full functionality.
>
> So now I'm wondering what cosmic forces might, in the best
> possible world, conspire to make an 8.1 be worth installing
> over such a board, or is it a lost cause? What gives me
> hope is I'm under the impression WinME is getting the full
> udma features now, maybe because of a patch (seems VIA
> website is most concerned about the SB Live! problem --
> was reading elsewhere about bus cycles and buffers, maybe
> losing data because of the bridge design).
>
> Kirby
Actually WinME is getting all that performance because the op system
does absolutely nothing to work around the bug. You have HDDs on IDE0 master
and slave. If they were on IDE0 master and IDE1 either master or slave (which
is the usual configuration for best performance since both disks can be active
at the same time), then the first time you decided to copy a huge directory
like 100 Mb or more, from dirve C to drive D the computer would chug along
for a while and reset itself in the midst of the copy operation. This is
reproducible on an ABIT board of your model just about 100% for
Win95 win98 Win2k WinME and kernel 2.4 Kernel 2.2 appears unaffected, and
so does winXP beta.
The northbridge KT133A has another problem about not being able to power down
softly (you have to switckh off the power supply.) This is limmited to a specific lot
af the chips and I have observed it on 2 FIC AZ11s and several MSIs and ABITs, but
never on an ASUS (maybe a remainder of their once terrific quality).
There are some configurations which interfere with our detection scheme (the bug was
admitted
by VIA just days before release of 8.0) and go on unprotected, and a few VIA MVP3
chipsets get slowed down unjustly because they manage to look like the deadly
configuration.
The beta 8.1 kernel, kernel-2.4.8-5mdk I just tested and we are back in business,
but still at VIA
performance levels instead of intel, now up to 20.82Mb/s and 23.7 after running
drakopt, but still
off the benchmark of 31.89 with the same drive on an i810 based board. This is of
course
a benchmark differnce and may and probably does have no significance in real life use.
But better chipsets are on the way for the AMD enthusiast. Remember the NVidias are
coming and we
hope mostly with open source drivers, and they won't be for Intel. The SiS 730 is a
pedestrian
chipset which goes on a lot of low quality boards but get a reliable board, as say a
Shuttle with
a 730 or even the mediocre ASUS A-7S, and it will probably never hiccup.
Civileme
But the performance of winME probably indicates you are simply not protected though
your hardware configuration to some extent protects you.
----------------------------------------
Content-Type: text/plain; charset="us-ascii"; name="message.footer"
Content-Transfer-Encoding: 8bit
Content-Description:
----------------------------------------
Want to buy your Pack or Services from MandrakeSoft?
Go to http://wwww.mandrakestore.com
