* ***** charles:
intel has annouced much more:
Itanium2 9010, FSB400/533, 1 Core, 1.6GHz, 6MB L3, 696$
Itanium2 9015, FSB400, 2 Core w. VT/HT, 1.4GHz, 12MB L3, 749$
Itanium2 9020, FSB400/533, 2 Core w. VT/HT, 1.42GHz, 12MB L3, 910$
Itanium2 9030, FSB400/533, 2 Core w. VT/HT, 1.6GHz, 8MB L3, 1552$
Itanium2 9040, FSB400/533, 2 Core w. VT/HT, 1.6GHz, 18MB L3, 1980$
Itanium2 9050, FSB400/533, 2 Core w. VT/HT, 1.6GHz, 24MB L3, 1552$
So that means not only that Montecito is delayed for one year but also
doesn't reach the 2GHz that were promised, or a fast FSB (FSB667 which
already is a bottleneck for two processors, 533MHz or 400MHz is even worse).
First SPEC benchmarks show that the top of the line Itanium2 9050 with
Linux 1474 CINT2000 and 3017 CFP2000 does (every core). A 4-cpu system
with 8 cores does does 134 int_rate_base_2000 and 186 fp_rate_base_2000.
As a comparison:
A Dell PowerEdge 2950 with the only two of new XEON 5160 does 123
int_rate_base and 83 fp_rate_base2000. A HP Proliant DL585 with 4x
Opteron 880 does 136 int_rate_base_2000 and 131 fp_rate_base_2000. So at
least in the very important integer arena the not-top-of-the-line
Opteron 880 already beats the fastest new Itanium2. The IBM POWER5 in
the eServer p5 570 which is already on the market for some time now has
a much better fp_rate_base_2000 score of 241.
So again the fp performance of Itanium is good (but due to the strong
competition not that great as it has been before) but the integer
performance is just average.
Of course YMMV as it heavily depends on the applications one is running...