> * Faults in chips generally involve things like shorts from power to
> ground, or miss-alignment over large areas. These kind of problems
> will cause the entire chip to be unreliable, especially over the
> long term life of the chip. It just isn't worth your reputation to
> release chips that fail after a few months/years.
Yes, I never thought it was a fault thing.
> * Without doing some major redesigning of the chips, it isn't easy
> (read cheap) to disable a portion of a chip after it is has been
> made and completely tested. So, shipping working 486DX's as 486SX's
> isn't profitable, and if you are going to redesign the chip, why not
> redesign it with out the FPU?
Sure it is. One of the 486 pins (where is my databook) is a FPU enable
pin. Just cut it in the final steps of making the package (or sever the
pin)
> * Major cost savings can be had by reducing the chip size, which
> removing the FPU would do. Major cost savings can also be had by
> reducing the amount of time it takes to test a chip to make sure it
> completely works. Removing the FPU does this also.
Major profit can be made by having two 'lines' one for the people who can
afford expensive things, one for everyone else. Two assembly lines, and
designs are expensive though.. So, you make the better one (because they
all sell for 100x their manufacturing cost, and it costs about the same
for both) and handicap some of them depending on demand.
> There may have been a few 486SX's that were carefully modified 486DX's
> for prototype purposes, but it is unlikely that any of them made it out
> Intel, and if they did, they went to a very small set of key vendors.
I've opened up 486 chips and between a SX and a DX of about the same age..
There is not any visiable difference in the die..
The design of a FPUless one would have blown away their extra profits from
the cheaper chip making better sales.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/