For some reason, there’s been some great desire to compare dual-core processors lately. There’s a fairly poor one (spare me the Dick Clark pretending to be a teenager routine) here comparing Intel’s and AMD’s duallies, and a better one comparing a new Intel Yonah to the x2s.
In both cases, the real issue is somewhat to completely ignored.
In general, as a rule of thumb, an Intel Pentium IV design needs to have a MHz rating 50-60% greater than that of a Hammer design for roughly the same performance. In the case of duallies, it looks to be about 55%.
So a 2GHz Hammer works out to roughly a 3.1GHz PIV, and a 2.4GHz Hammer works out to about a 3.7GHz PIV.
In the CNet review, the AMD flagship 2.4GHz 4800+ was going against a 3.2GHz Extreme Edition (a PIV with more cache). That’s a mismatch, and it should have come to no one’s surprise that the AMD chip consistently won.
Now it’s hardly AMD’s fault Intel’s fastest CPU (which costs even more than the 4800+) isn’t/can’t be revved up enough to be competitive, but it’s important to realize that relative clock speed is the major reason for the outcome.
If you compare the AMD X2 3800+ to the Pentium D 830, which is a 2GHz Hammer vs. a 3GHz PIV (both of which costing about the same), which is the real-life comparison most people would make, we find that the Pentium D 830 is much more competitive (it had a 2-3-1 record against the 3800+).
Of course, we are certainly not suggesting that any overclocker buy a Pentium D 830 over an X2 3800+. Why? Because you can overclock a 3800+ without cryogenic technology a lot more than an 830.
But isn’t that saying “Intel can’t crank it up enough” a different way?
In short, AMD’s biggest advantage over Intel right now is not any design superiority, but the ability of its SOI-based chips to crank up somewhat more than Intel.
And that’s the basis on which you ought to judge Intel’s next generation of CPUs.
The review of Yonah indicates that clock-for-clock, a Yonah does roughly about the same as a Hammer. Maybe a tiny bit less, but that can be ignored for our purposes.
The true test to determine if Yonah will be any good or not is not what one of them does at 2GHz, but how far you can push it.
If Opteron duallies are doing 2.7-2.8GHz without too much fuss, Yonahs are going to have to hit at least those speeds to compete on the desktop for overclockers’ money. If they miss to any real degree, then forget about them.
Nor is this a question for just overclockers. If Yonahs fall far short of AMD dually frequency, that’s going to crimp Intel’s ability to offer faster processors later in 2006, and when AMD finally gets to 65nm. . . .
This may be of limited interest in the notebook world, but tomorrow’s Yonah will probably look an awful lot like Intel’s late 2006 offerings.