[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Don't buy Fixed Pentiums



Rick,

I guess you know already that Microsoft has officially slipped the date for
Windows95 now to August.  Most experts quoted in PC Week believe we will now
not see Windows95 for the general public before January 1996.  I really hope
the special features of CADKEY Architect (or whatever) do not require either
NT or Win95.

I do have tremendous respect for programmers.  I just know that the average
architect doesn't want to do it (even if they know how).

PC Week today confirms the suspicion that you, Phil, Howard, etc. had, that
the calculation problems have more to do with algorithms than with any of the
5 unrelated Pentium flaws.  Apparently people all across the world are trying
all sorts of calculations on their systems, including Excel and even the
Windows Calculator.  

The latest is that if you subtract 2.10 from 2.11 in the Windows Calculator,
you get Zero instead of .01, and the rounding off error goes all the way back
to Windows 3.0 some 5 years ago.

The other one circulating is that in Microsoft Excell;
 1000.1 + 1000.2 + 1000.3 + 1000.4 - 1000.1 - 1000.2 - 1000.3 - 1000.4 =
5.68E-13

Apparently all sorts of rounding (or not rounding) errors have been with us
for all Intel chips for many years with many software algorithms.  The first
Pentium flaw just has made the general public aware of it.  For years
virtually all software comes with all sorts of fine print saying that you
have to check your results and that no one can sue anyone over inaccurate
computer results.

Most of the other Pentium flaws apparently won't really surface until the SMP
version of NT comes out, and when people start trying to upgrade  with
OverDrive processors.  Hopefully, by then there will be lots of inexpensive
alternatives.

Geoff

[New Search][DataCAD Home]