COMMENTARY--First we had Intel announcing the new Itanium 2, now with bleach, borax, and brighteners (or at least with a faster clock speed than the original Itanium, 3MB of on-die L3 cache, and an increased system bus speed--all of which make it a genuine world-class CPU in design).
Then we have a note from Gartner, part of which stated, "Some observers wrongly compare the raw performance of Itanium to that of Intel's desktop processors, optimized for clock speed and low cost. However, clock speed doesn't offer the proper comparison. Itanium was designed to deal with three critical issues: Long-term memory performance, integrity, and server performance. These factors must all be balanced carefully for enterprise applications, so the nature of a server's performance differs from that of a desktop 'speed demon.'"
Why bother with that at all? The Itanium 2 is faster than the original Itanium. Comparably, it's a speed demon and there's no substitute for more megahertz, right? The answer to that question is found in the verbal shots Intel took at AMD in the desktop space.
Quick rewind: AMD's Athlon XP processors are rated by relative performance, not clock speed. For example, the Athlon XP 2200+ is a 1.8GHz CPU which, according to AMD claims, performs with gusto of 2.2GHz (old-style AMD) processor. Of course, there is no old-style 2.2GHz AMD processor anymore, so the obvious obfuscated target of this claim is the Intel Pentium 4 line. Perhaps the source of Intel's agitation was the fact that the 2200+ does indeed perform much like a 2.2GHz Pentium 4 despite the clock speed difference.
In defense of the Pentium 4--and in an attempt to shatter the credibility of AMD's claims--David Mitchell, Head of Marketing, Intel UK, in a BBC World "click" piece, said: "I think megahertz and gigahertz are really the only way of accurately assessing what you're buying, because it's a clear measure in terms of processor performance and it clearly states to the consumer what they're getting for their money." Obviously, Mr. Mitchell hadn't yet tried to run an older Willamette Pentium 4 against a similarly clocked, newer Northwood Pentium 4. Differences in the chips' cores--the guts of the processors--make the Northwoods faster.
Now, however, Intel is facing a problem with the 64-bit Itanium 2: Its 1GHz clock speed will carry half the rating of AMD's forthcoming 2GHz eighth generation Hammer 32/64-bit processor. Using Intel's rationale, I guess we should assume the Itanium 2 will be a dismal failure in the server space as soon as Hammer is released. So Intel has to change the way it reckons, or at least someone at Intel has to take AMD's argument and apply it to the Itanium 2 as if Intel knew this all along.
Other clues that Intel is attempting to hedge its Itanium 2 bets lie in the chip's monumental 3MB integrated L3 cache. Historically, Intel has defended small internal cache amounts. When the company's ground-breaking 25MHz 80486 debuted with a miserly 8KB of internal L1 cache (and Intel was lambasted for it by the technology press), the company dragged out cache hit/miss statistics that seemed to indicate that 8KB was an ideal amount--until the 100MHz 80486 showed up with 16KB. L2 cache--external at the time--was also juggled around over the course of the 80486's existence, with less always giving way to more, and performance always rising with the cache size no matter what the statistics said to the contrary.
All of this will probably leave you in a quandary. As usual, things tend to boil down to the bottom line. Whether or not you jump on the Intel Itanium bandwagon probably has more to do with how high your costs may climb as you replace your existing 32-bit applications with 64-bit equivalents (if or when they're even available). AMD's Hammer will not run IA-64 code, so there'll be no shuffling of application environments--as is the case for Athlons and Pentiums on the desktop--so we're talking about a serious commitment. Bad as it might be for the economy, I'd avoid choosing between Itanium 2 and Hammer until I came up with some honest comparative data for the production versions of the two platforms--and not just rely on the praise or pans of pundits.
What's your take on Intel's and AMD's CPU performance theories? Speak your mind in our TalkBack below.
Stay focused: Sign up for Tech Update Today, the daily e-mail newsletter for those who need to know.
