Coppermine Bug Prevents... Booting?

mircea writes "Apparently, a problem with the wafers resulted in a PIII bug that prevents some machines from booting. ZDNet has the story. Dell has stopped shipping Optiplex GX110. So, what happens when you combine a PIII and a i820 chipset? " Let's be honest, how often do you boot anyway? I mean, its only gotta work once, right? *grin*

Comments

The Reason...

[jd]

...it was never found was that Intel's engineers are all using Linux and never saw the need for reboots.

Since then, "Windows Compatibility Enhancements" have been discussed and will be introduced, just as soon as the engineers stop laughing.

No sympathy for Intel...

[Wakko Warner]

After all the garbage they've done, I'm savoring this year. First the lawsuits against VIA, then the alleged pressure on Asus to cancel their Athlon board... I'm downright happy that they're getting what's coming to them now: the Coppermine chipset showstopper bugs, the fact that AMD not only has the fastest x86 CPU, clock-for-clock on the market, but also the fastest, MHz-wise, and is gaining more and more mindshare and OEM support, and the general lack of faith that I see people displaying on an increasingly regular basis toward them. This latest bug is just icing on the cake, in my opinion. Go AMD, go International Semi, go Transmeta (or so the rumors say)... competition is a *good* thing, and this latest twist in the CPU wars saga has been especially satisfying.

- A.P.
--

"One World, one Web, one Program" - Microsoft promotional ad

Catching bugs like this?

[Anonymous Coward]

I happen to have some background in the validation of processors. I don't speak for any company -- my opinions are solely my own.

Every processor sold is generally sent through two layers of testing, with random samples (fulfilling a statistical thing I'd rather not try to recall) going through even longer testing. There is a trade off -- the longer you test, the less processors go out the door. The shorter, the more chance you have off failure.

The first level would be as soon as you have viable silicon, you'd do your burn-in, and test the parts before breaking them into individual pieces. This filters out the total junk. This process simulates booting several times.

The second stage involves the finished product in a simulated system environment, running at shipping speeds. This is where they get the ability to 'bin' parts based on how they perform. This is typically done at temperatures beyond anything you'd see in the system level, to give some safety. This part typically involves booting up operating systems.

A longer test, on statistical sampling, would probably involve many applications, over a multi-hour period. (How many people overclocked their processors only to have something like Quake or Winbench fail -- it's a similar approach, run a lot of varied code.

While I don't have inside knowledge on this, I would speculate that it would have to (a) very rare, (b) only happens in a certain environment, eg, chipset or motherboard. I wonder if there is some significance to the term "glitch", which does have an engineering meaning.

It's almost a shame that Rob Collins has appearantly moved on, it would have been interesting to see his speculation on this.

Re:Bigger than fdiv bug?

[technos]

Not a big deal? These new Coppermine processors are to be installed in servers. What happens when you have a power 'blink' just too long for the UPS, or you're remote administering the machine and force a reboot? THE SERVER DOESN'T COME BACK UP. The system not coming up is FAR more tragic than a non-fatal fdiv bug. Why? I have to get out of bed in the wee hours of the morning, drive an hour, drink lots of coffee on the way, and press the power button again. All because some idiot paid a premium for on Intel's rushed-to-market 2nd place processor instead of buying a Athlon. If the system doesn't reliably power up, I don't want it or any of it's like brethren.

90% of PC misery caused by booting

[Greyfox]

If most PC Users would simply never boot their systems, they would be much happier and more productive.

Intel, feeling the pressure from AMD, releases a shoddy and under-tested product? Tell me it ain't so! And this differs from their previous releases, how? At least this time they can use the AMD excuse. In the past their shoddy and under-tested products had no excuse at all. Between them Microsoft and Intel are responsible for setting PC Users' quality expectations so low that any moron can peddle just about any piss-poor programming on the market indefinitely and expect to get away with it. It'd be nice to see them take a good hard dive.

I'm thinking Alpha chip here...

[weave]

I'm sitting on a budget of $200,000 to buy new servers at work and I've been wrestling which way to go. The 32-bit Pentium seems to be at the end of its useful life and the Itanium is still years away.

On the other hand, the 64-bit Alpha processor is still young, but old enough to be proven in the field. I'm wondering.

I'd like to grab a big Alpha-based muti-processor box with at least a gig of RAM and an external RAID unit with several hundred gigs of storage, all running Alpha Linux. I'm still a bit nervous about Compaq's future plans for the chip though.

Unfortunately, every single vendor and VAR I talk to actually laughs at me and asks me questions like "Can I ask you why you are not considering an industry standard NT/Intel-based solution?" and "Can you really count on Linux to survive the next few years? Those socialist coders are going to get bored and figure out they can make more money in the real world eventually." (actual conversations) :-(

Amazing, $200,000 and no one wants to help me spend it or at least listen to what I need instead of selling me their NT-based turnkey product. :(

It's no wonder Intel and Microsoft are #1 in the world. If you are not very thick-skinned, you get beaten into submission. It reminds me of the old days when getting approval to buy anything but a S/370 based IBM mainframe or S/36 box was near impossible.

Growing problems with slashdot (off-topic)

[Frac]

I submitted this story yesterday and it got rejected. I was surprised why it got rejected too. I guess someone at Slashdot originally thought a Coppermine Bug is insignificant.

I think it's time we need another "slashdot" thread just to discuss the growing flaws of the slashdot moderation system and etc. I'll list the ones off the top of my head:

* Stories get rejected for unknown reasons, only to show up again a day later.

(maybe rejected stories should be reviewed somewhere, so we can moderate the storie back up to be posted?)

* On the other hand, there are "new" stories that link to the same article as that other "new" story one month ago. How is linking to an article 9 days ago considered "news"? (IBM porting Linux to S/390, for example)

(have another queue that shows "approved" stories - members can go in and moderate interesting stories up, and repeated or dull stories down)

* news guys that are inserting too much of their own biased opinion into the news stories, shifting the opinion of the comments before anything gets posted! For example, I'm am on the verge of ignoring a certain section on slashdot, because one news guy is frequently jumping the gun on My Rights Online, ie. how a game software developer is tracking me down with my video card model. I appreciate the intention of that section (which is why i haven't dropped it yet), but is it really necessary for the article to acuse the developer having "unconvincing explanations", then later give some half-assed retraction with "oh, this isn't a big deal really, I just didn't really think before I gave out my conspiracy theory.

There are many more problems I (and I presume many of us) have with the current system. We don't want to bitch and moan, we just want a dedicated thread where we can all give out positive criticisms.

This will probably be moderated down, but I really hope malda/hemos/roblimo and etc. will see this.

Re:Growing problems with slashdot (off-topic)

[turg]

Here's a place to discuss this http://slashdot.org/article.pl?s id=slashdot-issues
-
<SIG>
"I am not trying to prove that I am right... I am only trying to find out whether." -Bertolt Brecht

Some folks in the HA group were working on this

[Greyfox]

IIRC some folks in the High Availability working group were actually working on a way to replace the kernel while the system is actually running. That would be pretty damn slick. Check around on the Linux High Availability pages and I'm sure you'll find it.