Intel puts on multi-core peep show
IDF Intel's main revelation of a new chip architecture stole the show on the first day of IDF. The chip beasty, however, did dish out a couple more tidbits that will interest the enterprise crowd.
On the more immediate front, Intel will start shipping its "Sossaman" low voltage version of Xeon next week. IBM announced plans to pick up this chip in its blade products.
By the end of the month, Intel will ship "Dempsey," which is the follow on to "Paxville" aimed at dual-processor servers. In the third quarter, Intel will then ship a follow on to Dempsey called "Woodcrest" that uses the new "Core" architecture. The Woodcrest chips will slot into the same servers using Dempsey, which should make for a quick transition, according to Intel.
In early 2007, Intel will back up these two-socket products with "Clovertown" – a four-core dandy. The company demoed Clovertown today for the first time in front of a large audience.
Also demoed for the first time was "Kentsfield" – a four-core desktop chip due out in 2007.
Intel has been big on selling "the future," which isn't a bad strategy since AMD has been cleaning its clock – excuse the pun – in just about every benchmark known to man. AMD's massive performance edge should erode somewhat with the introduction of Intel's "Core" designs in the second half of this year.
For those looking for earlier signs of hope, Intel championed a new TPC benchmark score for an IBM box running on four of its "Paxville" MP chips. The IBM x460 box has taken the top TPC score for a four-socket x86 system.
The victory, however, wasn't all Intel claimed. IBM's system did put up a TPC score of 273,520 compared to an Opteton-based box from HP that churned out a TPC score of 236,054. Of course, the IBM system cost $1.3m, while the HP kit cost just $476,000, leaving HP with a more than 2x price/performance edge. You've also been able to buy HP's box for the last few months, while the IBM system doesn't ship until May.
In addition, the IBM server uses the company's own specialized X3 chipset for Xeon-based servers – a fact seemingly lost on Intel as it promoted the IBM score as evidence of its Truland server platform strength. But why bother with the niggling details.
Hoping again to draw attention away from Opteron, Intel convinced an HP executive to appear on stage and profess his love for Xeon. HP plans to launch an entire line of new ProLiant boxes with Intel's fresh Xeon chips inside. Together, Intel and HP took a shot at Sun Microsystems's Opteron servers with another astonishing benchmark. Oddly, the Sun server – already shipping – was running Windows Server 2003, which seemed sacrilegious even though Sun does support Redmond's hairball these days. Anyway, the HP box trounced Sun's kit, but you can't buy it just yet.
To prove its Xeon passion, HP today also announced that the ProLiant refresh will include new boxes using Opteron. "HP will make available worldwide on March 20th the high-performing HP ProLiant DL145 G2, DL385 and DL585 rack-based, density-optimized servers along with its full-line of Opteron-based server blades – the HP ProLiant BL25p, BL35p and BL45p" with the speedy, new 2.6GHz Opterons, the company said in a statement mailed to reporters.
Of course, no IDF would be complete without a minute or two spared on Itanium.
Intel executives said that "the top server vendors" have come out in support of the chip. That's the top vendors minus three of the biggest sellers – IBM, Sun and Dell. We'll let you do the math on that one.
"Six of the eight" top mainframe vendors have also committed to moving customers onto Itanium. That, of course, excludes the one mainframe vendor – IBM – that matters. But, again, why bother with the details.
For those not convinced by Itanium's success, Intel pointed to "industry expert" IDC, which recently predicted that sales of Itanic servers would hit $6.6bn in 2009. It was pretty gutsy to call out that report in public given that IDC's past Itanium server predictions have been more than 1,700 per cent too high! After seven years of missing Itanium forecasts by revolting margins, IDC has only managed to prove that it's an expert at being wrong. ®