LN >  The P-II 333MHz is the
LN > first chip from intel that returns to the internals of the PPRO.
LN > All the previous versions ran the cache at half speed.
JM> miserable failure on the 32bit clean pentium pro and why intel
JM> were bulldowsed by microsoft into producng a chip that would "make
JM> win95 look good" in essance the pentium II's are now inextricable
JM> linked to ms win 95 one without the other
JM> I have heard a lot of people who upgraded to pentium II's very
JM> shortly 'upgraded' from os/2 to ms win95 and started ahving allsorts
JM> of problems but that is another story.
JM> from my calculations, a pentium II based processor would have to
JM> have about twice the clock rate to out gun a pentium pro if all else
JM> remains the same.
PG>What is the deal with the Pentium II vs. Pentium Pro?  I thought that
PG>the 333 Mhz Pentium II would run 32 bit code (like OS/2) as well as a
PG>Pentium Pro?  I have just purchased a 333 Mhz Pentium II, would I have
PG>been better off with a Pentium Pro?
PG>What is the highest clock rate on a Pentium Pro?
There are two separate issues.
1) The Pentium Pro is a 32 bit chip, optimized to run pure 32 bit
instructions.  No steps were taken in the design or in the chip
microcode to enhance running legacy 16 bit code.  Win95 (and DOS) is a
pig, figuratively, compared to the same code running on a pentium, based
on the percent improvement projected.  It made Win95 and the PPRO look
bad because Win95 was pushed as true 32 bit with no DOS legacy, and the
PPRO was supposed to be substantially better than the Pentium.  It
caused tremendous tension between Intel and Microsoft during the Win95
betas.  Several trade mags predicting a breakup much as what is again
predicted with the new 64 bit chip.  Remember OS/2 still had a 50/50
chance of upsetting the apple cart at that time.  OS/2, WinNT, Unix,
Linux show the expected improvements.  People used this as proof that
Win95 was a DOS/WFWG scam with a new interface.
The PPRO did not take off for the masses because of the above.  Intel
recovered by simply making faster and faster Pentiums; Cyrix and AMD
added in to make the PPRO a niche and server chip.  It became a self
fulfilling prophecy because as we all know chips get cheaper the more
that you can make and sell.  Win95 performance stopped the selling.
Making them is difficult because of the on-board high speed static RAM
cache, which lowers yields.  Even if the CPU works, if the cache fails
speed tests after assembly in the dual well package the whole thing has
to be tossed.
2) The L2 cache in a PPRO runs at processor clock speed giving no extra
wait states, so you end up with a 150, 180 or 200MHz L2. The L2
cache on a pentium motherboard runs at 60 or 66MHz. Even the
overclockers with VIA etc only get to 75 or 83MHz.  Of course 100MHz is
right around the corner.  The PII does two things that makes it not a
one to one match for the PPRO. The PII is the same general
architecture as the PPRO, However, the microcode was detuned to allow
better 16 bit performance and the internal L2 cache was removed from the
die well and placed on the little circuit board in the package.  The
added delay meant that the clock speed for the L2 had to be cut in half.
The end result was that the PII runs Win95 better than the PPRO. However
several real world multi-tasking tests using Unix, Linux and OS/2 show
that a 200MHz PPRO with 256K L2 can almost keep up with a 266MHz PII,
The 512K cache PPRO comes close to matching the 300MHz PII.  Notice that
Intel no longer includes the PPRO in their benchmark charts, they did
this once before with the true 50MHz 486DX and waited for it to fall off
the map. Just like the 486DX50 with its faster memory bus, increasing
the MHz alone is not enough to offset faster main memory performance.
Until the PII's go to a 100MHz memory bus the clock speed alone is a
fleeting performance indicator against a faster cache.
I haven't seen reports on the 1MB PPRO, it is so expensive that it is
only affordable for servers.  Intel had lead people to believe that the
333MHz PII would return to a full speed L2 cache but it did not come to
pass.  It still runs at half speed but the timing has been shortened to
increase access performance.  Intel hasn't really clarified this but
some think that they included RAMBUS technology to speed things up in
the L2.  I am withholding final judgement until I see something more
definitive in writing.  I am way behind in my hobbyists research.
Anyway a 180MHz PPRO easily overclocks to 200MHz and a 200MHz one
has been run by many at 233MHz reliably with the L2 speed automatically
increasing in speed at each of those steps. Because of this, lots of
PPRO owners have felt the 333MHz PII was the first one to offer a
reasonable performance upgrade path.  The PPRO is still the only option
for server market 4-way SMP until the next chipset BX ?  is released.
--Lynn
 * SLMR 2.1a * That's not an oil leak! It's a self undercoating option.
--- DB 1.39/004485
---------------