Skip to main content

View Diary: Smart Economy: Crazy Market "Logic" & the 486 Computer Chip (32 comments)

Comment Preferences

  •  I recall Intel math problems (0+ / 0-)

    but I don't recall that being the cause of the 486 disabled math processors. Are you saying Intel began 486 manufacturing with two separate manufacturing areas - one for chips with no math processor at all and one with an included math processor?  If not, the disabling of the math processors was not just something that happened after they discovered flaws in the math processing.

    I believe I heard some Unisys mainframes were also manufactured to be sold at different capability levels depending on whether only parts or all of it was enabled.

    "We all too often have socialism for the rich and rugged free capitalism for the poor." - Martin Luther King Jr.

    by workingwords on Mon Jul 15, 2013 at 11:39:33 AM PDT

    [ Parent ]

    •  No, you're right in that they made all the chips (3+ / 0-)

      in one place but there was more to the motive than just crippling chips for profit. They new there were going to be problem with the FPU (there always had been on the x87 series) and rather than just throw chips away they found a way to profit from it.

      What I don't recall is what the failure rates were and if that alone filled the need for the lower end chips. So it wouldn't shock me to find out they crippled some on purpose. But in some modern chips the failure rates approach 90% so anything they can do to salvage those failures can really help their bottom line.

      The tone of the diary strikes as "evil company being evil" and while Intel may be the Devil's own in this case I think it was the right thing for them to do. Especially given the fact that the practice was widely known.

      Sorry conservatives, but Occam's Razor isn't a beard trimmer for jihadists. What it means is I don't have to accept your crazy-assed theories as an alternative to reality.

      by ontheleftcoast on Mon Jul 15, 2013 at 11:54:06 AM PDT

      [ Parent ]

      •  I Googled (0+ / 0-)

        asking for "Intel math processor errors" got plenty of results for Pentium FPU problems, but not 486.  So, I tried adding "486" (then next "487") to the query and got some Pentium results and some results that were 486/487 but not about FPU errors.  It doesn't look like it was 486.

        "We all too often have socialism for the rich and rugged free capitalism for the poor." - Martin Luther King Jr.

        by workingwords on Mon Jul 15, 2013 at 12:37:51 PM PDT

        [ Parent ]

        •  Ahh, the Pentium FPU glitch (3+ / 0-)

          That was monumentally bad. That was bad because Intel discovered the problem after the chips had been sold. The team working on Excel discovered it when users started reporting failures with spreadsheets run on Pentiums that worked just fine on 386/486 machines. I forget the exact details of the problem but it was something as silly as (1.4 - 1.4) != 0.0. In effect that shit storm is what would've happened in the market if Intel hadn't disabled the onboard 487 for those chips, weird math failures (or crashes). With the Pentium they thought they had all the production problems squared away for the FPU was always enabled. It turns out the problem was with their microcode instead. Eventually they made it so their chips could have the microcode reprogrammed. Which leads to the other diary about security in a world where you can't trust the hardware. Once you can change how the CPU functions at an instruction level you can't trust any software on the box to be secure.

          Sorry conservatives, but Occam's Razor isn't a beard trimmer for jihadists. What it means is I don't have to accept your crazy-assed theories as an alternative to reality.

          by ontheleftcoast on Mon Jul 15, 2013 at 12:46:00 PM PDT

          [ Parent ]

Subscribe or Donate to support Daily Kos.

Click here for the mobile view of the site