JC Criticizes Intel's Increasing P4 Voltage
If you'd care to download this Product Change Notification pdf, you'll find that Intel has decided to increase the voltage of the Pentium 4 processor. This is rather, well, shocking, a bit. It's been long considered that raising voltage is the sign of a "last ditch effort" to be competitive in chip ramping performance, although this really has become commonplace with the Athlon/Duron and Pentium III processors over the last year.
But the real oddity here is that Intel has undergone this increase rather early in the life of the P4 core. Can I put this into perspective for you?
Back about a year, a month, and a week ago, AMD made public their first voltage hike. They did this at 800MHz. The K7 (Athlon) entry speed was 650MHz. In its first iteration (the cs44e7 "Classic" Athlon), the processor ran pretty much at the same voltage throughout. Remember this. This is important. I will also point out that the transition was made to cs50 at 750MHz (though, reportedly, some quantity of same-voltage cs44e7 parts were produced at that speed grade). This is less important, though notable.
Now, how does this equate to the P4? Essentially, in my opinion, it is a sign that Intel has likely fatally mispredicted their level of yields (or binsplits). We're something like halfway towards the second speed grade introduction of the Netburst (nee Willamette) core, and the core voltage has already been hiked. This is the important bit I wanted you to remember: the K7 core was conservatively spec'd with respect to voltage, while the P4 had apparently been more liberally rated.
Mistakes like these are annoying. And they might have repercussions in the future. Example: Intel had originally predicted that the P4 would reach 2.00GHz in their current process (P858?). Doubtless, they were presuming that a voltage hike would be required to get there to ensure arrival at the target speed grade. But the first voltage hike already happened. Did the Intel engineers' projections for 2.00GHz at 180nm assume clean fabbing for the initial speed grades at the initial voltages?
Clearly, there is much to ponder here.
jc-news.com |
