AMD confirms a CPU bug found by Dragonfly BSD's Matt Dillon

[Freedom Kira]

Posted on Slashdot. This could be bad...

[Lupin]

Posted on Slashdot. This could be bad...

Did you even read the mails linked in that post? The bug was found on discontinued products and occurs in a very specific configurations. This isn't like that old Pentium bug that can be easily replicated.

Hardware bugs aren't uncommon. All CPUs have errata. Even Sandy Bridge CPU has errata. What is rare is that this bug was discovered externally. What will AMD do with the bug? Will they fix it? No. They will add the bug to the list of errata for models that are affected.

[Freedom Kira]

I skimmed his report, yes. I don't follow AMD's product line closely enough to know their Operon line by heart, but I know that Phenom II was discontinued only recently, like a couple months ago. It's still a widely used processor regardless.

Doesn't the fact that the bug was actually discovered make the whole story more interesting? It takes a lot of effort to track down a CPU bug, and a lot of time gets wasted as a result.

[AnimeJanai]

AMD has dropped to 12th rank in semiconductor sales now whereas Intel has widened its lead and of course remains at #1.   The other ranks are filled by up-and-coming semiconductor/processor makers creating processing chips for everything from mobile devices to media controllers to internet switches to tablets. 

[kureshii]

This affects the 10h (K10 Phenom & Opeteron) and 12h (Fusion) microarchitectures.

Frankly it’s not really interesting to find a new CPU bug these days (except for the discoverer, perhaps). They’re not everyday occurrences, since the obvious bugs would’ve been quashed before production, and easier bugs would’ve been spotted soon after release (re. Intel’s Cougar Point chipset recall).

This one is getting attention because it was /.ed. Considering the bug was found more than 5 years after K10 launch, I’d say it isn’t really serious. Opterons and Phenom IIs around the world aren’t going to suddenly start failing like Schrödinger boxes.

[TMRNetShark]

This affects the 10h (K10 Phenom & Opeteron) and 12h (Fusion) microarchitectures.

Frankly it’s not really interesting to find a new CPU bug these days (except for the discoverer, perhaps). They’re not everyday occurrences, since the obvious bugs would’ve been quashed before production, and easier bugs would’ve been spotted soon after release (re. Intel’s Cougar Point chipset recall).

This one is getting attention because it was /.ed. Considering the bug was found more than 5 years after K10 launch, I’d say it isn’t really serious. Opterons and Phenom IIs around the world aren’t going to suddenly start failing like Schrödinger boxes.

So your saying the CPU ate the poison? Quick! Call Poison control! Hang in there CPU! I'm gonna rub aloe vare on your neck! 

[per]

This one is getting attention because it was /.ed. Considering the bug was found more than 5 years after K10 launch, I’d say it isn’t really serious. Opterons and Phenom IIs around the world aren’t going to suddenly start failing like Schrödinger boxes.

Whether or not the bug is serious is debatably.

Since it (basically) triggers random stack-pointer, albeit very infrequently, it is
actually way more serious than as an example the FPU multiplication error bug.

This kind of error causes the program to crash and burn, which is usually
considered more serious.

However, especially for windows-based workstations, people are not all that
surprised if a program crash randomly every few months. So perhaps it is not
all that serious after all.

Then again, the opterons are server CPU:s, and usually paired with ECC memory etc
to increase reliability. Crashes in servers can be more serious.

Then again, all large scale systems handles random crashes rather well, including whole computers
(or cluser of nodes) going down. So, again, it is probably mostly an annoyance.

[kitamesume]

the crash only happened on the most precise scenario, to make it more meaningless the scenario occurs very rarely that makes the bug too rare to occur, making it ignorable.