A Tom Bomb

The Tom’s Hardware article on the Pentium IV 3.4 claims that a new D0 stepping for the chip.

They don’t.

Unfortunately, the authors of the piece are either unable to read a CPUID correctly or chose to ignore the evidence that they didn’t have a D0 processor.

I’m sure it’s the first, but . . . .

The Proof

CPUID consists of three hexadecimal numbers (hexadecimal means base 16 rather than the usual base 10).

When you count hexadecimally, the numbers are 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, A, B, C, D, E, F (sometimes you’ll see 0A, 0B, 0C, etc. for the latter numbers).

The CPUID is a string of numbers coded into the processor. Programs like CPUID or CPU-Z simply read the string of numbers and break it down into its categories.

For the current Prescotts, Intel says the CPUID is 0F33h.

Programs like CPUID/CPU-Z read that as:

Family: 15 (or F)
Model: 3
Stepping: 3

Not at all surprisingly, that’s just what CPU-Z reports, you can find one of many examples of that on this page.

Intel announced a while back that it was coming out with a D-0 stepping.

Part of that announcement read:

CPUID will change from 0F33h to 0F34h

So any D-0 stepping processor is going to report itself as:

Family: 15 (or F)
Model: 3
Stepping: 4

The Tom’s Hardware chip does not. It reports the same 15-3-3 as a current Prescott.

Despite this, the authors of the piece call it an 0F34h, even though CPU-Z proves that it isn’t.

3=4????

That’s essentially what the piece is trying to tell you. Rather bizarrely, they show you 3 and then tell you it’s 4.

It’s not like CPUID-type programs need to be revised to read the CPUID data correctly. They do get revised to provide the correct explanations for CPUID, but the numbers get pulled the same way.

3.4 Will Not Be All D0 Steppings

Not only is the chip tested not a D0 stepping, most if not all of the first 3.4s won’t be, either.

Who says so? Intel.

Just look at the Intel Processor Spec Finder. Right on top, you’ll see an SL7B9, which is very clearly a 0F33, C0 stepping processor.

Don’t Run Out Looking For A Low-Speed D0 Now, Either

There’s a section of the article which lists future D0 processors. The article then procedes to tell you “if you plan to upgrade to a Prescott processor now, make sure that you get a D0 stepping chip.”

The listing from that article was clipped from this Intel document we referenced earlier. This document also says when you might expect such creatures to emerge.

It forecasts “Date of First Availability of Post-Conversion Material” as May 7, so looking for one in the store now is completely futile.

More importantly, if the pattern of prior Intel releases apply to this release, May 7 really means, “Date we hope to start shipping to Dell,” which is much earlier than “Date you can expect to buy one at Newegg.” The latter has been at least six weeks, and as long as six months after the “official” date.

I suspect it will be sooner rather than later, since I doubt Intel has been making a lot of these things and doesn’t have a big inventory to sell off, but even the most optimistic scrounger will be wasting his or her time for at least the next six weeks.

I got rather sensitized on this point when we were awaiting the C stepping for Northwood in order to crack 3GHz. I told people at the time to expect them about six weeks after the “official” date, and that time, it took closer to six months for the critters to show up.

That put a lot of people (including myself) off on wild goose chases looking for one (you might recall people buying Dells just to get the chips at the time).

Wild goose chases are never fun. Completely avoidable ones are even less so.

So don’t even bother looking until May, and it will probably be rather later than that.

The Effects of Consequences

There are errors, and then there are errors. If an error doesn’t have any real-life consequences on people, it’s not as important as one that does.

Not getting the stepping right in-and-of itself isn’t a big deal. Telling people that they’re going to get something for their $400 when they likely won’t is, at least to those buying.

Similarly, listing future D0 steppings without a correct ETA doesn’t hurt anyone except the people going out to look for one.

It’s not the stuff of tragedy, but errors and omissions like these hurt those who read and believe what they read, at least a little. Sometimes it’s due to circumstances beyond your control, or simply not knowing something, but that ought to make you more careful about doing your homework and avoiding the avoidable errors and omissions. That is within ones control.

Ed