Kap,
This turn of events is rather telling.. here is my
analysis for what it is worth.
In May of 1998 HP did a "Merced climbdown" telling everyone
they still loved Unix had several spins on PA-RISC
in the works or shortly to come...
Through the end of 1998 SGI didn't have the inside track.
We collectively must feel sorry for SGI as they didn't
"know" and didn't get to do a "Merced climbdown" until
March of 1999. If you go back to even August? of 1998
SGI's plans according the the rags were to drop MIPS and
go with Merced. Unfortunately, Merced wasn't going to
show up on time and the SGI "Merced climbdown" in March
of 1999 was very awkward indeed. I will drop all this
from a prior go round at the end of this note for reference.
Now turning to HP's climbdown of today ... the timing
is very suspect wouldn't you say? HP -- lead Intel partner
and all the glory that goes with that -- "stomps" on IDF
by "pre-announcing" a "Merced climbdown"... Hummm strange
indeed. But not so strange actually. If you go back to
a day or so ago where Uncle Andy got to demo the Merced
simulator actually doing something for the first time,
he makes the cryptic comment:
"We will see if it works when it goes first silicon".
That was strange and Intel spokespeople got their knickers
in a twist when poked and prodded by the press on that one.
Maybe reading too much on this but Paul sure is
*very quiet*, wouldn't you say? Besides, you tell me
why HP is raining on Intel's IDF parade... are they
softening the blow?
The climdown begins..
Rob
-----
Refernce for SGI/HP climbdown history (sorry for the
format... a posting of mine to a forum 6/24/99):
>> Why did SGI respin the MIPS if Merced is so good?
>
> Because they cut off their MIPS line way too early, _waiting_ for
> Merced?
>
Because they were the last to get the "very bad" news regarding Merced...
check this out from April 1998:
news.com
One reason SGI is cozying up to Intel is because its own MIPS architecture
has begun to lose its grip on the market. SGI will phase out the MIPS
processor used in current workstations and servers probably by 2001 as 64-bit
Merced chips from Intel become more pervasive and powerful, as previously
reported by CNET's NEWS.COM.
SGI will continue to boost processor speeds and incorporate the chip in its
high-end computers, but the last major MIPS release for servers and
workstations is scheduled for 1999.
It wasn't a month later that HP climbed down from their
Merced perch, did a "we love Unix" booster campaign, talked 8500, 8600, 8700
and possibly 8900. SGI was left flapping and
finally got to climb down from their Merced perch in the spring of 1999. In a
warm Intel gesture, (givin' the dog a bone) John Mashey of SGI gets to rub
elbows with Intel and pronounce: "Merced at 6 Gigaflops ... I've got
goosebumps ALL over me body" or some such. Reference:
news.com
Quote of the day from that URL:
"The Merced slowdown in many ways can be seen as the result of a collision
between expectations and reality."
Retro Roadmap Revamp
from URL above:
As part of this effort, SGI is releasing new MIPS chips every six to nine months,
said Mashey. SGI is currently building servers around its R12000 chip and will
come out with a faster version of the chip toward the end of the year, according
to Mashey. In the middle of 2000, SGI will release the R14000, which will run at
410 MHz and go up to 450 MHz.
The R16000, which will run at 600 MHz, will follow in 2001. It will be followed by
an 800-MHz chip in 2002, which could be sold as either a faster version of the
R16000 or as the R18000. Earlier, SGI's chip roadmap went through the R12000
only
Want a good read, go click on that...
Here we are 3 months later.
But that's my view, others have theirs. However, let's embrace more recent
history.. WinHEC April 1999 .. "Merced is on track" and Craig Barrett publically
states:
"Merced WILL sample in 3 or 4 months"
Oh?
eet.com
Let me relate a little story I've been hearing. A well-placed microprocessor
executive walks into the office of his boss, a capable leader who's known for his
penchant for the outdoors.
The exec tells the boss that, based on his best technical assessment, there is
no way a certain chip will be ready by its target date, and he suggests moving
the target date. The boss replies that there will be no delays: The chip must be
ready on time and it will be ready on time, he says. End of meeting.
But not yet the end of the story.
Time for a climb down?
theregister.co.uk
Craig Barrett, CEO of one of the biggest multinational corporations in the world,
Intel, said today that when he heard Merced was delayed "nine months ago" he
got very irritated.
But, said Barrett, Merced was on target to deliver in 12 to 18 months time.
Oh, so that's it. We are hearing a story that is nine months
old? But that can't be it as Merced slipped to Y2K quite a while ago. Why not
just come out and say: "Merced is on track and we expect to have samples
before Labor Day."
Maybe not, eh?
Merced goes volume in December 2000. Craig Barrett says so.
|
