In article <a8464d$q18$1@news.net.uni-c.dk>,
Erik Corry <erik@arbat.com> wrote:
>Nick Maclaren <nmm1@cus.cam.ac.uk> wrote:
>
>> The point is that, if AMD can get the Sledgehammer out only 6 months
>> behind the McKinley and outperform it 2:1 on SpecInt, there is a
>> very strong pressure on Intel to announce the Skunkworks project
>> and thus kill IA-64.
>
>Why?
See below :-)
>As far as I can see, no major OEM is going to announce servers
>based on Sledgehammer. People don't buy 4 way servers and up
>from screwdriver shops. So for people looking for a server
>what is the competition going to be? For adventurous Linux types
>Sledgehammer may well be interesting (and Clawhammer will be a
>hit) but what proportion of that market are they?
Small. But that isn't the point. The main reason that OEMs are
playing down any AMD discussions, experiments and even plans is
that Intel bears serious grudge. As I said a LONG time back, no
OEM of any consequence is going to announce any Sledgehammer
plans until it has taken the decision to stop pulling its forelock
to Intel and possibly even drop Intel in favour of AMD.
AMD and its believers have told me that I can expect to be told
plans under NDA in the first half of 2002 (i.e. now). Well, I
thanked them and followed my own counsel. Reports will start
leaking in the third quarter of 2002 and become serious in the
last quarter. It is likely that there will be no announcements
until 2003. Yes, I am predicting hard figures again :-)
Note that this really does mean that I don't expect announcements
to lead products by more than 3 months - as distinct from the
year we have become accustomed to (and 2 years for IA-64). So
the current lack of statements proves nothing, either way. And
please note the "either way".
>I think McKinley's performance relative to SPARC and Power4 is
>going to much more important. Even more important are
>non-performance things like the OEM issues, commercial Unix
>support or lack of it, etc.
I agree completely. But here we do have some information. Intel
has managed to get domination of the small server market by some
very hard selling, but took ages to break into even that because
of its lack of attention to RAS and SMP and support for the software
primitives needed for those. This is the main reason that the large
server market is still solidly 'RISC'. The IA-64 was and is an
attempt to dominate that market as well.
AMD has traditionally sold into the 'happy hacker' market, but it
is VERY clear that the Hammer range is an attempt to break into
the Intel-dominated small-server market. Don't look at the mere
performance claims, look at the RAS and similar aspects. So AMD
is girding up its loins for an attack on Intel's heartland, which
is something it has never done before.
>> Would this cause the OEMs to hold off investing in IA-64 systems,
>> the way that they held off investing in RISC systems following
>> certain Intel Merced announcements?
>
>1) It's too late to hold off investing in IA-64. The investments
>are made and they are ready to try to roll when McKinley gets
>here.
Do you remember the PowerPC? EXACTLY the same was true of that.
Yes, the whole scale was smaller, but the proportions were much
the same. Only, then, Intel was AMD and IBM/Motorola were Intel.
>2) AMD isn't Intel
Not yet, but it is working on it :-) And remember when it was said
that the PowerPC would dominate the industry because Intel wasn't
IBM?
There are a couple of further points.
Even a year ago, all the OEMs were bowing and scraping to Intel,
and putting up with the most appallingly one-sided IA-64 contracts
because they were terrified of being left out. They were then
thoroughly shafted by Intel, who got them to develop products for
the Itanic which was then cancelled. Intel DID then reverse
slightly, and allow it to be released for development, but the OEMs
lost a LOT of money. And so did the ISVs.
In the past 6 months, it has been clear to me from talking to them
that they are no longer intimidated by Intel. Some are seriously
pissed off, but it is amazing how much sewage people will wallow
in when there is serious money at stake. So they are pursuing
their IA-64 lines, yes, but I get the impression that they are no
longer prepared to suspend other projects just to keep IA-64 on
schedule despite Intel's delays.
Hence my remark (again a while back) that we shall know what they
are thinking by the products that OEMs announce (WITH dates, not
ones to be delivered sometime) immediately following the launch
of the McKinley. Which I am pretty certain will happen, unlike
the Itanic, where I was suspicious it might not.
And, similarly, a year ago, Intel could do no wrong in the eyes of
Wall Street. The fact that us 'techies' knew that the Itanic was
sinking (and some of us had known that for 2 years) had not reached
the Messrs Moneybags. But there have been a couple of thoroughly
negative articles about the IA-64 project in places like the Wall
Street Journal since then. So cracks are appearing.
Now, AMD are no slouches at marketing, dirty tricks and all that,
and know the realities of the industry better than I do. We can
be almost certain that they are planning to get the Sledghammer
versus McKinley issue raised in such fora. And even financial
journalists will use soundbites like "Twice the transaction rate
(as measured by TPC)" - if that is convincing.
So watch this space. We are in for a really spectacular fight.
Godzilla versus King Kong and all that. Just be careful not to
get underfoot, as they aren't strong on rules, and there isn't
an arena around that will hold them :-)
Regards,
Nick Maclaren,
University of Cambridge Computing Service,
New Museums Site, Pembroke Street, Cambridge CB2 3QH, England.
Email: nmm1@cam.ac.uk
Tel.: +44 1223 334761 Fax: +44 1223 334679
groups.google.com
Thanks to Combjelly for pointing to other messages on the thread... |
