Three quick notes.
1) The ASUS P4P800 board I have uses a new method of BIOS recovery. I
have edited my BIOS just to play around and then used this to recover
it. There is nothing dangerous about it:
http://www.asus.com.tw/support/english/techref/mbfeatures/cfbios2.aspx
http://www.asus.com/products/mb/crashfreebios2.htm
2) Don't forget that Prescott CPU's introduced a 1 MB L2 Cache and SSE3
instructions. I am going to assume that MS tried to optimize the use of
these two things from the Prescott CPU's as well.
3) Most of the Dell's that have shipped since about a month after the
release of the Prescott core that are configured with a P4 processor use
the E Revision CPU's. They have loaded up their Outlet market with the
C Revision (Northwood) CPU's.
Finally, what I noticed through this whole process was that MS did
excellent testing on SP2. What they failed to do was to release the
potential RTM to the testers, allow them to test, then seeing that there
were no huge problems, then tag it RTM and release it. This may have
saved quite a bit of heartache.
Also, just FYI, Cari Miller (MVP) wrote the KB Article pertaining to the
Prescott issue:
http://support.microsoft.com/default.aspx?scid=kb;en-us;555185
----
Nathan McNulty
One thing that is nice is most modern computers have either a backup for
the BIOS or a tool that makes it almost impossible to screw up
Well, it's like backups - for most of us, the only true test of a
backup is to scorch your data then try a restore.
Those dice are way too big to roll, IMO.
What happens isn't so much that Intel has problems with their CPU's and
the BIOS fixes those problems or that Windows is dependent on the type
of CPU. It is more that Windows is being designed to take advantage of
new features of CPU's such as we saw with NX. When you start adding new
capabilities for new CPU's to an OS, this is where you begin to run into
problems.
Rod asked whu SP2's Update.sys was bigger (implying it must include
microcode) and I mentioned extra detection logic. All that AMD64 NX
stuff may be a factor, plus evaluating post-MMX support.
One reason SP2 might care about Intel CPU details (noting that these
CPUs don't do NX, and suppressing NX via Boot.ini doesn't fix) is for
DirectX 9c's new floating point pixel shaders; they may use SIMD3.
I can't believe that MS released SP2 through Automatic Updates even
after knowing about the Prescott issue. That seems like the worst thing
they could have ever done.
Yep, IMO that was a bad call. At the time I'd been involved in some
private email stuff with MS (that's where my NDA concerns came from)
and I was polling WU from my test PC to spot when SP2 push started.
By this time, no-one was still saying "865 and 875 chipsets", it was
now correctly identified as Prescott, so what I'd heard about other
(non-Intel) motherboard chipsets for Prescott wasn't news that had to
be formally broken. Still, I was relieved when MS posted most of the
detail (Update.sys, etc.) in the newsgroups.
SP2 started being pushed at around 02:00 (that I could tell) and so I
wrote it up and webbed it by 04:00. By now, both Intel and MS have
formally documented it, tho I dunno if MS's /kb article is linked off
the "what you need to know about SP2" front door.
I pity those who purchased a Dell or something
You see, the big brands weren't the problem; they are still shipping
Northwood. The target market doesn't think beyond "oooh Dell, good
brand name" so there's no reason to fret specs; Dell will ship
Prescott when they've cleared thier old stocks first.
So maybe in the US's big-brand-dominated market, this looked like a
small deal. Whereas here, we started shipping Prescott immediately,
from June; all new builds since then have been Prescott.
before the microcode was updated and now have a useless system
and don't know enough to fix it.
Exactly - it's a disaster, and unforgivable if known in advance.
It's also ironic, given that SP2 sets up WU patches to be downloaded
*and* installed without user initiative. That looks like a bad idea,
given that SP2 itself could render a PC unbootable - and if MS knew
that and pushed it anyway. On what basis should we be confident they
won't do that in future, with other patches that break "a few" PCs?
I've seen assertions made here that the issue was known during the RC
phase, and I have no personal knowledge whether this is so or not - it
may be thet the number of affected PCs would be under-estimated, as
Intel's spin made this look like a minority issue that would be
confined to a few lame brands and should be fixed by now ("just get
the new BIOS" - assumes one exists to be gotten).
That's why I'm chasing the details behind this. For example, looking
at Intel's documentation of thier own BIOSs, which versions ship with
what mobo revisions, which of these are still on sale, the date at
which Intel's BIOS became properly Prescott-OK, etc.
And I see Intel mobo revs too old for Prescott (even BIOS aside) on
sale through distributors as late as August, and "updated for Celeron
D" Intel BIOSs dated June 2004 (the month Prescott shipped here).
So I really have my doubts about Intel's presentation of this issue.
It could be that PCs were put at risk because Intel played fuzzy with
the truth, which they may have done to FUD up sales of their own
motherboards, as well as play down the scope of the problem.
It's deep politics to ponder on whether MS took Intel's story at face
value, or were complicit in the spin. I think the former, as I can't
see why MS would want to go out on a limb to cover up for Intel's
dirty washing (why have ppl shout at SP2, when the real issue is
Intel's stealth bug-fixing?) and I don't think Intel has that much
leverage over MS anyway. After all, it seems as if it's AMD who are
on the cutting edge with MS, considering NX etc.
It must be depressing for MS, having thrown intense resources at SP2's
development, only to have to face ITW problems anyway. The hard
lesson is that good intent isn't enough to ensure success; not when
hitting the reality wall of complexity and the entropy of human error
rates - not just within MS, but across the entire industry.
-------------------- ----- ---- --- -- - - - -
No, perfection is not an entrance requirement.
We'll settle for integrity and humility
-------------------- ----- ---- --- -- - - - -
[/QUOTE]