back to article Mystery Intel bug halts shipments of some Sapphire Rapids Xeons

Intel's 4th generation Xeon Scalable processors arrived behind schedule when the silicon, codenamed Sapphire Rapids, debuted in January 2023. Now the x86 giant has paused shipments of some chips in that family due to a fault with the components. In a statement to The Register, Intel said the issue specifically affects medium- …

  1. Gary Stewart

    OK, but...

    Anybody know how the CPU determines that it is running "commercial software" so as not to trigger the bug? Since Linux is mostly not commercial software, does the bug affect Linux?

    1. Justthefacts Silver badge

      Re: OK, but...

      Commercial doesn’t mean paid-for. The article says that the chip contains new “slew of dedicated accelerators for everything from data analytics and cryptography to machine learning”. The bug is probably in one of these. Unless your code accesses one of those, you won’t trigger the bug. That’s not even a compiler issue: you’d have to explicitly call an API to frob registers that Intel hasn’t yet officially released the existence of.

      1. CrackedNoggin Bronze badge

        Re: OK, but...

        Well then, If it's just cryptography there is nothing to "see". Move along please.

  2. Gary Stewart

    All of those extra cost special processors are for commercial use. But the bug doesn't affect commercial use so the bug is not in one of those processors. Sorry, I still can't figure this one out. Does anyone else have any ideas?

    1. Version 1.0 Silver badge
      Boffin

      How many similar bugs do we have recorded in an 8080 or a Z80 environment? I worked with a range of 8048's for years and never saw any problems, sure they were much slower and 32-bit calculations were big but I saw very few coding issues back then. Ever since then we have developed far more "features" and "abilities" but these issues described by El Reg are not rare anymore because we have "evolved" the computing environment ...

      It's like thinking that we eliminated the dinosaurs because we were furry little animals stealing and eating their eggs, but then something else happened.

      1. Atomic Duetto

        6510 = 3,510 transistors

        8080 = 6,000 transistors

        4th gen Xeon = 100 billion transistors (says internet/Intel)

        Quite likely somebody/something made a mistake when drafting it (out by hand on what was possibly a very large piece of paper…..)

      2. Henry Wertz 1 Gold badge

        Agreed, but...

        Agreed, but... the Z80 did have several bugs. Apparently, several bugs were fixed in a later revision... then reintroduced after they found some widely-shipped software relied on the buggy behavior, later Z80s were made bug-for-bug compatible and just listed that as the documented behavior. Per google, the 6502 originally had a buggy ROR (Rotate bits Right) instruction, it worked except the "end' bit did not end up in the overflow register as it was supposed to; so they just said it didn't have one and that it would be added in a later revision. These CPUs tended to have at least several "undocumented" instructions -- on the 6502, generally bits of hardware were selected based on the bits in the instruction so all the undocumented combinations did something -- sometimes just locking up the chip, or selecting unusual and usually useless combinations of functionality from other documented instructions. But on later revisions, they made them NOPs, then on even later variants used those to add additional instructions.

        That siad, I do agree -- it'd be great to have some "middle ground" -- CPUs that are not as ungodly slow as the Atom-based E-Cores, but far less complicated pipelines, instruction reordering, etc. than the "performance" Intel and AMD cores (and probably no hyperthreading) so there's less scope for bugs, corner cases, and security flaws to creep their way in. Ahh well. At least the modern ones have microcode so you can usually patch around the "stepping 0" bugs with microcode updates.

      3. Bitbeisser

        Well, there were times when humans used sharp stones to shape sharp sticks for their tools. Nowadays, tools are created by sophisticated CNC machines/robots.

        The evolution of processors over the last 50 years is pretty much equivalent to the evolution of human tools from the stone ages to today....

  3. iron Silver badge

    Did Intel forget how to do maths again?

    1. Tom Chiverton 1

      Foof! I hope not...

      1. Anonymous Coward
        Anonymous Coward

        Back side bus?

        Is that a sly reference to Pentium F00F bug?

    2. Korev Silver badge
      Windows

      "Pentium inside, Can't divide"

      The old jokes are the best -->

      1. OhForF' Silver badge

        Pentium of Borg: Mathematics is futile. You will be approximated.

  4. CowHorseFrog

    The article doesnt even remotely say what the bug is ...

    1. Richard 12 Silver badge

      'Cos Intel aren't saying, and presumably the customers who found the issue have agreed not to publicise it in exchange for some "consideration".

  5. Aqueiroz_BR_BA

    CVE RELATING

    Dear all,

    RELATED CVE

    Dear all, does the BUG detected by INTEL at the end of June, as described in the article, have any relation to CVE-2022-40982?

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like