I never said that VIA chipsets were bad. I only said that they operate
differently from Intel chipsets, and that most BIOS code is cut and
pasted from Intel’s examples, which means that the BIOS tends to be more
broken on a VIA machine than on an Intel machine.
The Intel i815 is a fine chipset, but it has bugs just as all other
hardware does. (All software has bugs too. I don’t mean to
discriminate.) I’ve personally added five or six hacks to Windows XP to
make it run, which is about average for an Intel chipset. In general, I
don’t add as many hacks for non-Intel chipsets. But that’s just because
the design cycles for the other guys is shorter, and they don’t send me
machines with their chips in time to do anything about the problems.
All the desktop chipsets shipping today, with the exception of one from
VIA and one from SiS, have an I/O APIC. But not all motherboard makers
enable the feature.
I disagree with your notion of hack-job PC’s being as reliable as a Dell
or an HP. I spend a huge amount of my time debugging problems with
specific motherboards. And it really boils down to finding the
cross-product of bugs that create nasty hardware interactions and then
avoiding them. When Dell, HP, Compaq or IBM build a machine, they
choose a chipset and a few different component suppliers. Then they
test the machine extensively, shaking out the incompatibilities before
they settle on a final configuration. This process gets a much higher
quality product into the end-user’s hands. (Incidentally, this is ten
times as true with SMP machines. None of the raw motherboard makers
builds one that I would consider reliable.) Part of the problem with
building your own machine is just that you can’t personally afford to
test each peripheral in isolation, chosing the ones that don’t have bad
interactions.
I once debugged a machine that had a bad interaction between the SuperIO
chip and the CD-ROM drive. When the CD would put data on the IDE bus,
it would cause a signal to be asserted that would cause all bytes in the
top lane of the X-Bus to appear as all 1’s. If that occurred while the
OS was doing a 16-bit read to the X-Bus, then data returned would be bad
in the upper byte. The problem went away if you changed to a different
brand of CD-ROM drive, or if you used a different SuperIO chip.
In another example, you can’t get stand-by to work on a 440BX-based
machine unless you exclude one of the three supported types of RAM.
Using one of the other two types requires a massive BIOS hack to make it
work. But if you apply the hack with the wrong RAM type (and software
couldn’t tell the difference) then the machine will hang. When I told
ASUS about the problem, they couldn’t do anything about it because their
customers install the RAM themselves. When Intel told the large
companies about it, they were all able to successfully patch their
BIOSes.
My point is that the people who design the commodity motherboards don’t
control the peripherals that will be matched with them. So they can’t
guarantee that the machine, when viewed as a whole, will actually work.
The big guys can.
- Jake Oshins
-----Original Message-----
Subject: RE: ACPI Machines and IRQ 9 [was: Communicating with the NT
developers]
From: “Maxim S. Shatskih”
Date: Mon, 31 Dec 2001 14:56:33 +0300
X-Message-Number: 19
>This is a stupid story. And I’m embarrassed to tell you the truth.
But
>here it is.
Really an interesting story, Jake. Thanks for it!
>But if your machine has a VIA chipset, or if it has a BIOS that we know
>to be broken, then we fall back to the Win2K-style stacking behavior.
So, VIA is this bad. Will know this now.
>The unfortunate truth is that you guys on this list mostly build your
>own machines, rather than buying them from reputable manufacturers,
Yes, twice as cheap, and the same performance and reliability. All
depends on parts vendor, mainly motherboard vendor.
As about the brand names like Dell or HP - they are known to be not 100%
compatible with the usual PC, for instance, Dell required a
PCI card to be Dell-certified.
>which means that you guys own the machines with broken BIOSes and VIA
>chipsets.
Lots of motherboards on the market (and thus self-assembled machines)
use Intel chipsets like i815 or such.
Is i815 bad?
>One notable addendum is that any machine with an APIC interrupt
>controller, and thus more than 16 IRQs, will spread interrupts out,
even
>in Win2K.
And what chipsets will provide APIC on non-SMP machine?
Max
—
You are currently subscribed to ntdev as: $subst(‘Recip.EmailAddr’)
To unsubscribe send a blank email to leave-ntdev-$subst(‘Recip.MemberIDChar’)@lists.osr.com
