Hmm.. to be fair to Intel though, their 5,665 server units
No surprise here, they are counting whole server system prices, not
just the processors, Now check motherboard prices
Well even with that taken in account the number of processors on
itanium system is greater than on opteron.
Intel is pushing it no matter what, and AMD should hope intel won't
push it harder. If intel would just once release the Itanium for a new
procecess at same time as their x86 counter parts others would be in
deep trouble in server market...
And how is the performance difference ? whoops ?
Well not 17 times as average, but typically the price goes up
exponentially, from smallest to biggest systems.
4 Processor topend opteron will costs, over 4times as much as two
processor top end opteron, without giving 2 as much realworld
performance while 2 the peak.
The same goes from 4->8 etc... Itanium sales are in bigger
configurations, like 16 or 128 processor systems, so price/performance
isn't such a deal, because they give the performance points opteron
won't have, even at price points that opteron system vendors can only
hope for.
How much exactly R&D for Itanium was ? I remember something arround
$1B.
Thats SPEND money, but there is difference between sustainable and
already spend money. For instance itanium can sustain its current R&D
based on sales for this year on intel...So who cares what was the R&D
costs that it had on previous years, intel invested its x86 revenues
to kill 3 RISC families and take the processor market from them and
succeeded, and probably withing few years can get the investment back
in extra revenues on itanium platform. Itanium doesn't have to sell
millions of peaces to succeed. Even half a million per year is quite
profitable venture. But if it succeeds greatly and intel could sell
million or couple million per year its still some extra revenue for
intel that would of gone for other ventures without itanium.
If you doubt the cost difference then. Let take their low end of
itanium line...
Sells at 513$ has half the cache and 3rd less of cache so only usefull
for software developement platform, still intel is having nice little
markup on them also. And the 1.5MB cache itanium really has 6MB of
cache just most of it disabled. [The disablement is not for fixing
defects its just that engineer salary compared to volume makes em more
profitable to simply disable extra cache than design new layout for
smaller cache, especially with high yield process on large wafers that
intel has.] So intel is selling 4000$ for high end itanium because
many customers are willing to pay for that for their multiprocessor
systems, and is exacly same chip as they can sell for 500$ with nice
profit margins on them too. Now if constantly people talk that alpha
team costed 100M$ to keep alive. And itanium has 3 teams on it plus
compiler guys. So thats well under 400M$ that itanium revenue should
be this year, and after that rest is profit. [Or repaying the
investment intel made on itanium.] So ASP with 2000$ and sales of 200k
per year its something you should keep its alive, BUT there is still
more itanium sales should grow little bit.
I personally hope that AMD can survive, and power too on the strenght
of intel since if they die processor prices go up, we all have
itaniums...
BTW: Itanium2 core is much smaller than P4 core but the cache's thats
is redundancy protected take most of the area. So itanium should be
about 2x as expensive to make than P4. And intel seems to make great
profits on P4 so it wouldn't be far fetched that itanium COULD be made
as a desktop processor with new software x86 emulation layer that
already is onpar xeon on integer and beats it on floatinpoint.
Jouni Osmala
