Tried a google search for current Itanic chips, and all I got were 5 yr
old articles.
You might want to look a little more, new Itanium chips are still
being released, though not as regularly or to as much fanfare as the
latest x86 chips.
The latest and greatest Itanium is the Itanium2 9050, running at
1.6GHz, dual core and with 24MB of cache. This chip was codenamed the
"Montecito" and was released a couple weeks after the first Core based
Xeons.
Just wondering how the current Itanics perform compared to the fastest
Core servers.
For it's intended tasks Itanium is still doing ok actually. If you
check out the SPEC CPU2006 pages, comparing the fastest Core based
Xeon chips to the fastest Itanium2 chips we have:
CINT2006_peak
Dell Precision 690 - Xeon 5160 = 18.1
HP Integrity rx6600 - Itanium2 9050 = 15.7
CFP2006_peak
Dell Precision 690 - Xeon 5160 = 15.6
HP Integrity rx6600 - Itanium2 9050 = 18.1
So for single-chip performance, according to SPEC CPU2000 at least,
the Core Xeon's are fastest at integer stuff while Itanium2 is faster
at floating point stuff, with the margin of difference being
essentially identical. On the CPU20006 Rate scores, the Itanium is
blowing all the others out of the water due ot it's ability to use up
to 64 sockets/128 cores vs. 2 sockets/8 cores for the fastest Xeons.
For TPC-C at the high-end it's still a battle between Itanium2 servers
(HP Superdome and Fujitsu Primequest) vs. IBM Power5 servers. However
looking at only 2-socket servers, we get a surprisingly close result:
HP Proliant ML370G5 - 2 x Xeon X5355 (8 cores) = 240,737
HP Integrity rx6600 - 2 x Itanium 9050 (4 cores) = 230,569
So on a socket-for-socket bases, the Xeons are coming out slighlty
ahead, though on a core-for-core basis the Itanium is way ahead
(rx6600 with 4 x Itanium 9050, 8 cores = 359,440).
As far as $/tmp-C they are nearly even, best 2-socket Xeon =
$1.80/tpmC, best 2-socket Itanium2 = $1.81/tpmC.
Ironic how we used to hear Itanic this Itanic that - the holy grail
this holy grail that..............now not a peep!!
I think it's pretty well acknowledged that the Itanium was WAY
overhyped and that it just hasn't turned out to be the end-all be-all
chip that Intel wanted it to be. That's been pretty clear for at
least 3 or 4 years now. However it is still competitive, particularly
in the tasks that it's good at. It's still seeing some use in
high-end servers, high performance computing and elsewhere.
The real question though is what is Intel going to do for the future
of Itanium? Right now Intel has at least the "Montvale" scheduled for
this year, which looks mostly just like a tweak of the current
"Montecito" chips, possibly bringing a die-shrink to 65nm as well as
some updated features. After that there is "Tukwila" planned for
2008. This was SUPPOSED to be the chip to bring Itanium to the big
time back in 2004/05, but has been delayed MANY times and had some of
it's more radical features removed, so now it looks more like just a
4-core version of "Montvale".
Same thing as when the p-4 came out and how great that chip was - all
"future-proof" with SSE and all.
The P4 WAS top-dog at Intel for 6 years, at least as far as x86 chips
are concerned. That's longer then the PPro/P2/PIII core lastest (5
years). The main difference was that AMD had MUCH more competitive
alternatives to the P4 with it's Athlon and Athlon64 chips vs. what
they had for most of the PPro core's life (mostly K5 and K6).
Funny that though, given a few years and what was once the "future
proof" solution isn't so future proof. Give it a couple years and
current Core 2 Duo and Xeon chips, as well as AMD's Athlon64 X2 chips,
will start to look pretty dated too.
