"All Chips Are Not Created Equal"
Ed Stroligo - 12/16/00
--------------------------------------------------------------------------------
When Thomas Jefferson said, "all men are created equal," he wasn't talking about CPUs, too.
Some think that. Here's a typical comment I got just this morning:
"Many people say that they could only o/c there P3 800 to 972. Why?? My friend o/c his CPU to 1049."
All chips are NOT created equal. Just because your friend or neighbor hit a certain speed doesn't mean you're going to, too. Just because you did doesn't mean everyone else can, either.
CPU-making is like cookie-making, you make batches, and not all batches come out the same, nor does each cookie in any particular batch.
We get quite a few comments from people who overclock chips for a living as to how chips are doing. Over the past six months or so, we've seen more variation than we've been accustomed to from Intel.
Until the last six months, Intel chips more-or-less followed what I call the 80/20 rule; they either work 80% or more of the time at a certain speed, or less than 20% at that speed. This made overclocking pretty predictable.
Lately, though, that hasn't been the case. The later cB0s and now the cC0s show more variation than in the past, and that variation lies right over that 1Ghz point a lot of people are aiming at.
A few months back, I had said cC0 chips were likely to hit a 1.0Ghz-1.05Ghz range with high-end air cooling. The upper proportion is holding very well, but it doesn't look like we're getting 80% or more of these chips to 1Ghz, around half is more like it.
Here's some comments from a reseller that illustrates the point:
"We sell P3-700 cC0 FCPGA OEM CPUs, which we receive in batches. In some of these batches, 95% of the CPUs run at 1Ghz+ with only modest voltage boosts, and all chips were able to run at 933Mhz.
"However, we have received batches of CPUs, coming from the same fabrication plant and with the same stepping, that didn't come close to that. Of those, none of the CPUs would even approach 1Ghz, and at least half of them will not run at 933Mhz. The only difference we were able to detect between the batches is that the "good" CPUs were manufactured during the early summer, and the "bad" CPUs were manufactured somewhat later than that."
(Editorial note: When you pay more for a pretested chip, this is a big part of what you're paying for; the privilege of not getting one of these "bad" chips, which have to sold off at a loss.)
"Perhaps Intel's yields are not as high as usual, and they are skimming off the better ones to meet their 1Ghz quota. Perhaps they're making them less overclockable. Perhaps they've expanded production a bit too far, and lost a little quality in the process. Or perhaps we have just had bad luck.
"Just thought you would like to know.
"There are no guarantees in overclocking, even if the stepping is the "right" one and a particular chip's track record is particularly good."
We gotten similiar comments from other resellers. That (along with looking at the databases) is why we're so iffy about PIII hitting 1Ghz. It's not a sure thing, even with cC0.
Facebook
Twitter