From: ching@angelo.amd.com (Mike Ching) Subject: Re: AMD-40 Cycle Timing Problems Date: 18 Aug 1992 18:06:50 GMT
In article <1992Aug15.230511.14243@mintaka.lcs.mit.edu> knight@hal.gnu.ai.mit.edu (Knight of Tossing and Turning) writes:
>
>Sorry this took forever to jump into this conversation but I figured it
>wasn't going to freak so many people out.
>
>I was instructed in my PC Archetecture class that the AMD-40 CPU (and
>only that CPU) has a relatively minor problem with timing between
>certain cycles. I'm not sure which combination of instructions causes
>the problem, and I'm not in a position to find out.
>
>If someone looks this up, it is a documented bug. However, I don't
>have the AMD CPU manuals to look it up.
>
>The reason this is only for the 386-40 is because it is basically a
>superscaler CPU (like a 486-2DX50) consisting of something like
>2-25 MHz chips. I'm not really into the hardware aspect of personal
>computers, so don't ask me much more than this. But is basically
>timing between the two where the bug occurs.
>
>As the guy with the flat-hat says, "A PC that is MS-DOS compatible
>might not be UNIX compatible, but any that is compatible with UNIX
>will run probably anything."
>
>--- Eric Knight
No one has bothered to correct this so I guess I will. The information
in this posting is wrong. The Am386 does not have a documented bug in
the manuals. Neither the Am386 nor the 486DX2 are superscalar.
On a related topic, there is a design problem on the motherboard if
the Am386 is getting hot. If the core logic chip is also getting hot,
it's pretty likely that there is bus contention. On a properly designed
motherboard the Am386 is barely warm to the touch.
Mike Ching
AMD Field Applications