Some senseless babbling from Lynn Nash to Mike Ruskai
on 04-04-98  12:35 about Motherboards and os/2...
 MR>I am currently using a Pentium II/233 overclocked to 266MHz, with a 75MHz
 MR>bus.
 LN> I noticed that you have overclocked your bus, obviously to overclock
 LN> your PII. Did you include the increased memory speed in your
 LN> calculations?
I didn't do calculations of that nature.  Wouldn't really know how, either, 
since I don't have any useful numbers to compare DRAM to SDRAM, using any 
speed for the latter.
 
 MR>The Intel SpecWhatever benchmarks clearly show the Pentium II faster.
 LN> Single tasking benchmarks, raw speed is not everything. Several
 LN> magazines reported at the time that multi-tasking and server
 LN> performance showed dramatically different results with a real
 LN> application mix. Unfortanately when this debate was going on I did not
 LN> save any of these references. They were more in things like computer
 LN> reseller news, VAR report etc.  This same debate occured over the
 LN> 486DX-50 versus the 486DX2-66.  Intel's solution at that time was
 LN> simply to remove the DX-50 from their benchmarks.  Current advertizing
 LN> seems to show that they have done the same thing with the PPRO. I would
 LN> guess that the tiny nuances when doing real work must be difficult to
 LN> explain and the world has become used to using synthetic benchmarks for
 LN> comparisons. 
 MR>I'm certainly not above questioning Intel's integrity, but my own
 MR>experiences have confirmed this.
 
 MR>For example, with Java benchmarks I've done myself, my speed is very much
 MR>faster than 200MHz Pentium Pro machines (using Windows NT).
 LN> Does this mean that your Java benchmarks were also on NT or are you
 LN> comparing apples to oranges? I mean everyone fights over Java
 LN> benchmarks, it is the current spectator sport.
I don't have NT installed.  What I did do was read between the lines based 
on the benchmarks done with both the Pentium and Pentium II with the same 
JVM, compared to the Pentium Pro machine.  In point of fact, my Pentium did 
better than the Pentium Pro machine because the JVM (OS/2 1.1.4 JDK) is 
better, but I took that into account, since I know how a Pentium Pro is 
supposed to compare to a Pentium.
 
 MR>It's certainly true that Pentium II's are not optimized for 32-bit
 MR>performance, but that's not a sacrifice of it, so far as I can see.
 
 MR>I searched high and low for all speed information I could find, and 
othin
 
 MR>showed the Pentium Pro faster than a Pentium II at the same clock rate,
 MR>much less twice as fast.  In fact, a normal Pentium with MMX (which isn't
 LN> Mike simple logic at this point should tell you that if the cache
 LN> speed is cut in half, something should change if the clock rate is the
 LN> same or the benchmarks are not stressing cache/memory performance.
How important cache and memory performance is depends on what you're doing, 
naturally.  
If this makes the Pentium Pro top out over the Pentium II at some point, I 
didn't find it.
 
 MR>important - the larger internal cache is) at 200MHz is only marginally
 MR>slower than a Pentium Pro at 180MHz.  A Pentium/233 with MMX is faster.
 
 MR>Granted, there are other factors which would show a Pentium Pro to do
 MR>better, probably, due to its superior branching, prediction, etc.  But 
on
 MR>of these internal features are absent in the Pentium II, and in fact
 MR>enhanced.
 LN> At the time of introduction even PC Weak did not agree with that
 LN> assessment on initial testing. I believe what they said was that it
 LN> seemed to be enhanced to make Win95 perform better. One must then ask
 LN> the question, why did Win95 perform badly on the PPRO to start with
 LN> since the CPU portions of both chips are essentially the same
 LN> architecture all the way down to having the same bugs.
Because the Pentium Pro is optimized for 32-bit code performance, while 
Win95 is heavily imbued with 16-bit code on many levels.
 
 MR>While the L2 cache is decoupled from the same die, it sits right next to
 MR>the chip in the processor module, accessed at chip speeds, never touching
 MR>any motherboard bus.
 LN> True for the PPRO, not true for the P-II. On the P-II it sits on the
 LN> cartridge circuit board requiring receiver/driver logic and is the
 LN> reason access speed was cut in half. PPRO is CPU clock speed, P-II is
 LN> (CPU clock speed)/2
True.  I should have said chip/2 speed.
 
 MR>I upgraded to this chip from a Pentium/200 (non-MMX, with the normal
 MR>internal cache), and clocked at 266MHz with the 75MHz bus, it's right 
bou
 
 MR>twice as fast in raw processing power.  Encoding a MPEG-1 Layer-III audio
 MR>stream with a 44.1KHz sampling rate, joint-stereo, and a 128KBps stream
 MR>speed took about 2.6 minutes per minute of audio on the Pentium/200, and
 MR>takes about 1.3 minutes per minute of audio on this machine.
 LN> I have no problems with those comparisons to the Pentium as they seem
 LN> spot on.
Have you personally evaluated Pentium Pro machines?  I'm open to being 
wrong.
Mike Ruskai SA/AG #1106
thanny@home.com
... And the only thing the Borg left was this copy of Windows.
___ Blue Wave/QWK v2.20
--- Renegade v5-11 Exp
---------------