Quote:> While the Pentium Pro sacrafices 16 bit performance (poor old Windows)
> for 32 bit performance, I am curious how well the Cyrix 'clones' (the
> so called '+') chips work with Linux. My gut feeling is that they
> were looking at the Windows/95 market and optimized 16 bit
> performance, perhaps sacrificing 32 bit performance. Does a so called
> P166+ deliver P166-like performance in a true 32 bit OS? If not, what
> level of performance does one realistically get? I do not care about
> floating point performance...
The P-ratings are computed using mostly 16-bit applications, if I
recall correctly, but the 6x86's performance ratio on 32- vs. 16-bit
code is similar to that of the Pentium, so the P-rating is as
reasonable a guide to CPU speed (on 6x86 vs. Pentium) for 32-bit OSes
like Linux as for older code. Relative to a Pentium Pro, of course,
neither the Pentium nor the 6x86 holds up in 32-bit code as well as
they do in 16-bit code.
Quote:> Actual PCI and system board speed on these so called over 'clocked'
> Cyrix system boards may also be big issue.
Overclocking is overclocking, and there's no fundamental difference in
this with the 6x86 than with any other chip, though of course how much
you can overclock may vary, and I've no information on this.
Quote:> I assume a Cyrix 166+ actually runs at 133 and probably uses a 66MHZ
> system board achitecture, whereas a Cyrix 133+ maybe runs 120 on a
> 50MHZ system board.
Correct, except that the 120MHz CPU speed translates into a board
speed of 60MHz.
Quote:> Certainly, a true Intel P120 (system board speed actually 50MHZ)
> offers little to no benifit over a P100 (system board 66MHZ).
The Intel P120 also uses a board speed of 60MHz.
Remember that the P-ratings used on the Cyrix CPUs are based upon
empirical data, not reasoning about how the performance "should"
compare. Performance will vary from subsystem to subsystem,
obviously, but overall performance of something with a P-rating of,
say, 150 will be pretty close to that of an Intel P150. It's probably
simplest if you take the ratings in that way, unless you want or need
to get into performance of specific subsystems, in which case there
are benchmarking results available on various sites on the web (sorry,
I've no URLs handy).
+--------------------------------------------------------------------+
| Rod Smith Author of: |
| http://ezinfo.ucs.indiana.edu/~rodsmith "OS/2 Soundcard Summary" |
+--------------------------------------------------------------------+
by