Original URL: http://www.theregister.co.uk/2005/10/29/intel_xeon_2009/
Second-class Intel to trail AMD for years
Parity prayers aimed at 2009
Analysis Two weeks ago, Intel looked poised to mount a serious challenge to AMD's server processor performance lead. Then the shockers arrived. Dual-core Itanium chip production slowed because of quality concerns. A sophisticated future Xeon processor was cancelled. Plans to unite the Itanium and Xeon lines around a common architecture to make servers more affordable and faster have been pushed out likely until 2009. In short, bad news for Intel customers.
Underneath all of these roadmap adjustments lurk some painful technology slips that must have customers concerned. In particular, it now appears that Intel will stay married to its front side bus dependency for much longer than previously expected and will fail to deliver integrated memory controllers on time. Where Intel had a very real shot at closing the gap with AMD in just 18 months on previous roadmaps, it now looks more likely to trail for close to four years. This should worry folks at Dell, HP and SGI, as they're most vulnerable to Intel's shortcomings.
We covered all of Intel's roadmap changes here, but will zero in on the "Tukwila" Itanium processor delay and "Whitefield" Xeon cancellation in this story.
Tukwila and Whitefield were both meant to arrive in 2007 and not only have integrated memory controllers but also share a high-speed serial interconnect called CSI. Both pieces would help Intel compete against AMD's direct connect and hypertransport technologies.
Intel, however, struggled to insert these goodies into Whitefield and will be forced to push out a replacement processor called Tigerton without the technology in 2007. Tukwila will have the goodies but will ship in 2008 instead of 2007 as hoped. And the common technology won't help Tukwila too much since Intel doesn't have a Xeon chip ready to share it. Such a Xeon isn't supposed to arrive until 2009, according to industry chatter.
So, Intel has two problems. On one hand, its chips will continue to suffer from performance issues as a result of the front side bus dependency and lack of integrated memory controllers. (Intel has some kind of workaround for this in Tigerton, but not a real competitor to AMD.) In addition, Itanium customers - the chosen few - won't benefit from the cost savings that a shared Itanic/Xeon server infrastructure promised to deliver.
Nathan Brookwood, an analyst at InSight 64, summed up the issues well in a research note issued this week.
"If Intel is ever to reclaim the performance lead from AMD, it must make the transition to an on-chip memory controller," he wrote.
"Intel slated Whitefield (a Xeon) and Tukwila (an Itanium) as its first processors to incorporate on-chip memory controllers. Tukwila will still use this approach, but Tigerton, Whitefield's replacement, will rely on a memory controller built into the chipset that supports the CPU, Intel's traditional approach, rather than a controller built into the CPU itself. Given the two year cycles that drive Intel's server roadmaps, this means that Intel will not be able to field a server processor with an on-board memory controller until 2009 at the earliest. Between now and then, we see little likelihood that Intel will be able to claim performance leadership."
Suspension of Dell-belief
Compare this scenario to one being espoused last week by Dell Chairman Michael Dell.
"Intel takes a very definitive lead in performance and power management at 65 nanometers," Dell said at a conference, according to a report from IDG News Service. "If we thought AMD was going to be supercompetitive in the spring and fall of next year, we'd be introducing AMD products right now."
(Such bravado seems laughable given that Dell's own statements indicate AMD has a supercompetitive position now and that Dell still chooses not to use the products.)
Dell's statement came before Intel reworked its roadmap. At the time, Dell's logic seemed plausible. Intel was set to release a new line of dual-core Xeon chips in 2006 that, while lacking an integrated memory controller, could well rival Opteron in performance. Then, the industry assumed, Intel would follow these chips with the 2007 dynamos.
In that scenario, Intel supporters could easily justify sticking with their man. The dual-core Xeons might not win on every benchmark and they might not be performance per watt gods, but they'd compete well enough against Opteron to stop massive defections. That would hold things steady for 18 months or so.
A more middle-of-the-road scenario painted by Xeon and Opteron supporter HP placed the dual-core Xeon and Opteron chips on pretty even ground in the two-way server market but gave a real edge to Opteron in the four-way server market where AMD's direct connect plays a more significant role.
"Dempsey, Intel's dual-core Xeon for servers with two processors, should be very close in performance to Opteron," IDG News Service reported, basing the remarks on comments from Mario Cooper, a manager of HP's ProLiant servers. "AMD's advantage over Intel's designs are more pronounced in servers with four processors, and the Opteron 800 series should have a clear advantage over Intel's Tulsa processor early next year."
Now, however, this tit-for-tat squabble has changed.
As stated, Intel will struggle to match Opteron in the next 18 months but instead of rolling out Opteron killers in 2007, Intel will introduce more processors tied down by its aging architecture dependencies. The company does not have a realistic chance of besting Opteron on typical server benchmarks until the new chips arrive in 2009. By that time, AMD will have new four-core designs of its own and who knows what other innovations.
Itanium customers meanwhile face near-term delays, and must use the same costly architecture they're stuck with today until 2009.
The great news for Intel is that its failings seem to have relatively little impact on its revenue. The money keeps rolling in as AMD tries to help customers certify its hardware and related software for corporate data center use. AMD gains a point of market share here and there but has yet to stump Intel's Xeon business in the massive way that many predicted.
Intel's missteps will certainly challenge this trend. It's hard to believe that Intel's customers won't start thinking "enough is enough." ®