Monday, December 26, 1994 by Dave Winer.
DaveNet is back!
I'll have some comments of my own after Scott speaks...
Intel's problems go way back. The processor that put them on the map for most people was the 8088 (also available as the 8086.) People ranting and raving over Intel's handling of the Pentium floating point bug don't realize that Intel's method of dealing with problems in this fashion dates all the way back to this first processor.
The 8088 was released with a serious flaw in it. This flaw would cause your computer to destroy data and/or freeze up. For the techinically inclined I'm talking about the problem of the missing interlock on loading the SS register.
Intel never replaced any chips I know of and to this day anyone coding software that might run on a 8088 class computer has to put a software work-around in their code. I've had to put this code in many times myself.
But perhaps the biggest sin that Intel is finally paying for was their next x86 processor: the 80286. Most people screaming about the Pentium flaw don't realize that it was incredibly serious flaws in the 80286 processor that cemented DOS as the OS for PC's. Bugs in the 286's protect mode resulted I believe in DRI being finally taken off the scene as a serious OS vendor. They had gambled on being the protect mode OS for PCs, but alas the 286 couldn't run it. Many people have written about a magic plane flight the decided that DOS would be the OS for PCs. I think instead it was a group of Intel engineers that made DOS the OS for PCs.
After the fiasco with the 286, Intel did much better with the 386. Here they wound up with only one serious public bug, the infamous "Double Sigma" stamp came to be an all important factor when purchasing 386 processors for many folks.
Then we have the 486. Alas I'm in the biz. As such I'm under NDA to Intel. I can't tell you what I know about 486 processors. All I can do is refer you to Andy's Internet posting and nod my head up and down quickly.
And then we have the Pentium. The bill finally comes due and it's shaping up as a $1 billion bill. One thing that the general public should note though is that Intel has only gone public on one flaw in the Pentium. If you go back and read what Intel has said about the Pentium you'll notice something interesting, there are other bugs being fixed in the new chips. Those other bugs are still under NDA protection.
Andy may have appeared to have come clean of late, but they're still hiding problems with current Pentiums behind their NDAs.
Dave, I think they still haven't gotten it yet. I'm expecting big things from the P6 if you know what I mean. :) I'm also just waiting with baited breath for the press to finish what they've started by asking about the other Pentium bugs and "Hey what about those 486 chips you folks make?"
The Intel 486 bugs really piss me off. Intel played stupid until we were able to prove it was the Intel part causing the problem. They then sent us a document under NDA which revealed they had known of the problem since April of 1993 (it was then November 1994.)
The fascinating thing was that we had all the major BIOSs. Intel's "fix" for the problem was a software hack in the BIOS. Neither AMI, Phoenix, or SystemSoft knew about the problem. One of my tasks right now is to hack the work around into our SystemSoft BIOS (and hopefully sell the hack back to SystemSoft.)
How in God's name can Intel sleep at night I ask? They've got a serious bug in the chip that causes it to crash and they don't even make sure the BIOS suppliers put the work-around into the BIOS code! (or evidently even make them aware of it...) Perhaps the computer maker fixes it, or maybe they just write it off to unstable Microsoft Windows code. We didn't find the problem right away ourselves, it's kinda hard to provoke due to the nature of it.
But wait, it gets worse!
Intel "fixed" the problem in a newer stepping of the 486. This is what allowed us to prove it was an Intel problem. We started soldering some of the newer chips into our units and when we went back to take another wack at fixing the problem we discovered units that worked a lot better than when last we had looked into the problem.
We can still make the newer chips fail though. But it's real hard, your usual Windows user, who spends most of his/her time playing Solitare won't encounter the problem. Compile 10 megabytes of source code though and boom! You basically have to get the box busy.
But wait, it gets even worse!
Intel still has both parts in the pipeline! We got a shipment of chips in November 1994 and got both flavors. Intel is purging the buggy parts by selling them mixed in with the good chips. Are they hoping to spread the bad chips around so no one notices the problem??? Enquiring minds want to know! :)
All I can tell you is to tell yer press friends to ask Intel for their problem reports on SB80486SX-25 parts with the following S-SPECs: SX765, the bad part, SX854 the better part. (S-SPEC is the Stepping Specification, sort of like a version number.)
The problem revolves around the STPCLK pin causing a stack corruption (we figured this much our before we went NDA so I can tell you that much.) The stack corruption will cause interesting things to happen, DOS windows under Microsoft Windows 3.11 will spontaneously close with no error message for example. Borland C++ Compilers will crash under MS-DOS 6.22.
But only if you stumble into the right timing on the STPCLK pin, i.e. it occurs randomly. The compile works one time, croaks the next.
Please tell your press friends that if they want to use my name I need to check their article to make sure they got it right. This is a highly technical problem and I'd hate for them to miss the target with my name all over it.
As for the 8086/8 and 286/386 problems, they're common knowledge. No NDA required. The 8086/8 screwup is even documented in learning-how-to-write Intel-assembly-code books. One of their fuckups becomes something you teach younger programmers about...
The press can also talk to companies like Quarterdeck, the makers of QEMM. This product for example checks your 286 to see if it's one of the "bad" 286's. If it is you can't do certain things with QEMM (Intel did eventually make good 286 chips.) And I'm sure Microsoft has code in Windows 3.1 to check for "bad" 286's as well...
And of course there might still be some folks to interview at Novell who worked for DRI. I'm sure they'd be happy to talk about how Intel pretty much killed their company (DRI never really did recover from the 286 fiasco.)
Anyway, all the above is why I laughed so hard the first time I saw the Pentium bug featured on Headline News. Intel finally got nailed, but it was for a stupid stupid stupid problem. I just keep seeing this mental movie of Andy and folks sitting around going "Hell why are they busting our chops about this problem? This is nothing compared to what else we've done to our users!" I still get a chuckle just thinking about it. Puts a smile on my face every time I overhear someone talking about how they've gotta get their Pentium chip swapped out.
Thanks Scotty! I have a few things to add...
First, I don't buy the idea that it's OK to ship inferior hardware to people (such as myself) who know how a computer works on the inside. There's been a sea-change in the computer business, sure we have a lot of Moms and Dads buying computers now, but product quality is an issue, a big one, no matter who you're selling to.
Second point: inevitably, this story raises questions about other chip vendors, clonemakers, system software vendors, even application and utility software companies. Everyone should take a very detailed look at their product quality issues. It seems like now is the time to come clean.
And Intel should change their policies, if appropriate, to provide full public disclosure of any problems with any of its products, at the time the problem is first discovered. This new policy should be implemented immediately, and it should be communicated, on a personal level, by the top guy at Intel.
A couple of stories --
My grandfather, Rudy Kiesler, who used to run a garment business (a big one, a very rough industry) gave me some advice when I started my first company back in the early 80s. He said "pay for your sins." At first I didn't understand what he meant. Then I did, when customers started having problems with our product, and we reacted arrogantly. You can't avoid paying for mistakes, but you *can* postpone the paying. But when the bill finally comes due, you end up paying a much bigger price.
That's what's going on at Intel right now. They're paying the bigger price. How much bigger? If Scotty is right, a much, much bigger price than anyone could have imagined. Potentially a Watergate-style price.
I read a story last week in the Washington Post about a quality problem that AT&T had. Service had gone out in the NY metropolitan area. At a press conference, the chairman of AT&T was asked who was to blame. He said (not an exact quote) "As far as our customers are concerned, I am." What a beautiful attitude! He put all AT&T employees on notice -- if we screw up, I'm going to take the heat personally. If I worked for AT&T, I'd be damned sure to take care of the customers first. And he went a step further, he reduced the rates for all long distance calls on Valentine's Day to pay for their sin. They took their money and applied it to love! So much for big companies not having big hearts.
Intel has shown that its customers don't come first. Intel does not have a big heart. Very reluctantly, gradually, indecisively, it is willing to pay for its sins. Listening to all the noise on the airways, it's not clear what or who comes first for Intel.
It's time for Intel to speak, clearly, decisively, own up to whatever sins it has committed, and pay for them, now. Clean it up. Say what Intel stands for. That's the only way to put this Pentium stuff behind them and be part of the future of this industry.
Intel's philosophy is in question. And so is the philosophy of this industry.
PS: As always, if you aren't interested in this kind of stuff, send me email and I'll happily delete your name from the list. And it's OK to forward it or repost it anywhere you like. The list is expanding -- I'm always happy to add new names. Send email to email@example.com.