Hello Mike!
Monday March 30 1998 17:38, Mike Ruskai wrote to Jonathan Michaels:
 MR > Some senseless babbling from Jonathan Michaels to Lynn Nash
 MR > on 03-29-98  23:52 about Motherboards and os/2...
 MR > [snip]
 JM>> i saw a lot of pentium II's running win95 very badly might
 JM>> i add, well with 16 mb even os/2 or evem linux would be a
 JM>> bit jerky.
 MR > [snip]
 JM>> hardware is not as good as it used to be, example at point
 JM>> is the diference between the pentium pro and the pentium
 JM>> II's.
 JM>> from my calculations, a pentium II based processor would
 JM>> have to have about twice the clock rate to out gun a
 JM>> pentium pro if all else remains the same.
 MR > [snip]
 MR > This is very interesting to me.  What calculations exactly lead
 MR > you to believe that?
maybe calculation was not quite the right word, but the pII's run thier 
interal
caching at half buss speed, the ppro runs at full buss speed. i compared to
identical platforms, all bar the processors. starting as samish spped
precessors the pro was significantly faster than the pII this remaind the 
ase
untill the pII was about twice the 'clock' speed of the ppro.
observation would have been a better word to use.
also the pII have several internal tricks to make them more 'responsive' to 
s
windows 95 and this causes some obstructions tot he way they work in real 32
bit operting system mode.
thier was a very good technical disertation on the faults flaws and merits of
pII's, ppros and dec alphas .. i cannot remember where i saw the info, it was
over a year ago, it was at a technical seminar i attended i think. i knew i
should have photocopied the damn paper .. drats.
 MR > I can find nothing that leads me to believe that a Pentium II is
 MR > not faster than a Pentium Pro MHz for MHz, all else being equal.
from all the marketing info, of course, can you imagine the outcry that that
would cause .. its is in the technical literature that you will dind the
details in the the fine print .. where the elegance of the solution is
appreciated and who gives a dogs hind leg for how fast it runs.
hobby computer owners and marketing drivem purchasing officers are the only
people dumb enough to be sucked in by the numbers scratched on the ceramic. 
y
only concern for chipo speed was for the requiremts of buss timing and
interrupt trigering and for the actuall length of time the event took to 
tart
happen and to fionally return to its rest state.
all of this happend on processors running at 1 mhz ... we pitted a 1 mhz
mc68b09e running os9 aginst a i386dx33, no prizes for guessing which one
of the two got shamed.
 MR > I am currently using a Pentium II/233 overclocked to 266MHz,
 MR > with a 75MHz bus.
and this is supposed to prove, what ?
you might be a bit stupid to do that but then again its your processor, not
mine. or that it shows how conservative intel are about thier rating.
i'm stilll waiting to find out the real story behind the damning reports that
were authored by nasa schuttle engineers about the arm twisith intel used to
get nasa to swap from the far more reliable 6800 seris processors to the less
reliable 80xxx seris 'consumer grade' procesors.
 MR > The Intel SpecWhatever benchmarks clearly show the Pentium II
 MR > faster.
and this is supposed to prove, what ?
that intel can tailor a benchtest that makes its processor look good against
all comers
 MR > I'm certainly not above questioning Intel's integrity, but my
 MR > own experiences have confirmed this.
being in what fields .. home bbs running ?
 MR > For example, with Java benchmarks I've done myself, my speed is
 MR > very much faster than 200MHz Pentium Pro machines (using Windows
 MR > NT).
i can find easier ways to cripple a machine than that.
 MR > It's certainly true that Pentium II's are not optimized for
 MR > 32-bit performance, but that's not a sacrifice of it, so far as
 MR > I can see.
if some fifty perent clock cycle for clock cycle is not a sacrafice then what
is ?
 MR > I searched high and low for all speed information I could find,
 MR > and nothing showed the Pentium Pro faster than a Pentium II at
 MR > the same clock rate, much less twice as fast.  In fact, a normal
 MR > Pentium with MMX (which isn't important - the larger internal
 MR > cache is) at 200MHz is only marginally slower than a Pentium Pro
 MR > at 180MHz.  A Pentium/233 with MMX is faster.
whic intel publication did ths come from .. certain not any of the processor
specific documentation i have seen, recently.
 MR > Granted, there are other factors which would show a Pentium Pro
 MR > to do better, probably, due to its superior branching,
 MR > prediction, etc.  But none of these internal features are absent
 MR > in the Pentium II, and in fact enhanced.
for ms win95 .. problem is that ms win 95 is a ms dos shell and ms dos is, at
best a 16 bit system with 8 bit internals, os/2 is 32 bit largely as is nt 
nd
most if not all unicies.
because they are thier is noting to write home aboutm what has been done to
them to make thems more or less usable in 32 bit mode is what concerns me 
rom
an engineering point of view .. alas thes days only as a source of interest.
even when i was working i'd nver recommend any project the was responsible 
or
the saftey and lives of people be entrusted to intel silicon.
sure intel has goten more reliable over the last twent years, but it is still
not as reliable or predictable as motorola was twenty years ago.
inspite of apearences to the contrary i am open to 'reeducation' as the
chinesse call it, but you would have to put up a good showing.
 MR > While the L2 cache is decoupled from the same die, it sits right
 MR > next to the chip in the processor module, accessed at chip
 MR > speeds, never touching any motherboard bus.
to an engineer that says a lot about the significance of the systemic issues,
and how intel viewes its production values.
 MR > I upgraded to this chip from a Pentium/200 (non-MMX, with the
 MR > normal internal cache), and clocked at 266MHz with the 75MHz
 MR > bus, it's right about twice as fast in raw processing power.
 MR > Encoding a MPEG-1 Layer-III audio stream with a 44.1KHz sampling
 MR > rate, joint-stereo, and a 128KBps stream speed took about 2.6
 MR > minutes per minute of audio on the Pentium/200, and takes about
 MR > 1.3 minutes per minute of audio on this machine.
ok, now you have me at a disadvantage, i am on new turf, i know very little 
how
precicely mpeg is processed.
these numbers are impressive, but what do they realy mean and what sort of
assistance is the 'raw' data stream getting form peripheral circuit 
rocessing
and off buss processor emulation in the operating systems itself, or more
precicely the applications, that are handling the tasks .. whch are buss
mastering and whic are slving the processor .. thier is a lot of task 
pecific
questions that need to be answered before you can make the statement that you
have dome, by just presinting the 'statistics' as an fait acompli.
over my years i have don this as well, particilarly when i knew the client
wouldn't understand the technical details and it was important that the right
hardware be used .. i'd presnet the case for the prefered end run and let the
'statistics' speak for themselves.
i am not saying that this is what you are doing, atempting to do here. 
ather,
that you are just letting the sattistics 'speak for themselves' to which i
reply, yes i see the statistics but how do you achieve them and how was the
system altered at the critical peripher process/tasking level to aggregate 
thes
results. now unless you are the author of the software in question it is
pointless to continue this line of questioning because we fall into the trap 
of
having to assume how and what is happening ..  assumption, is never a
replacement for accurate measurement . my old professor used to say
 MR > The memory subsystem on this one is also better, using SDRAM at
 MR > the 75MHz bus speed, as opposed to 60ns 72-pin DRAM SIMMs on the
 MR > other machine, but given the synchronous nature of a streamed
 MR > file encoding, I'd wager than the L2 cache of each processor
 MR > essentially nullified the effect on performance of the memory.
of cousre .. but the fact that thier is less l2 cache and that it runs at 
alf
buss speed, that is it runs at the speed the motherbaprd is running, in your
case that woult be atclck/2 with would be some 75/2 ..  .37.5 mhz
lets get one thing straigt you do not have a 75 mhz motherboard, at best you
have a system that runs at 37.5 mhz and parts of the pci bus run at 75 mhz
ther is currently a great deal of debate about the efficiencies of this sort 
of
design structure .. the jury is out and not expected to be back in for at 
least
4 or so years .. that is how long it is going to take to 'prove' the
differences tween the intel ideas embodyied in thier merced line, dec's alpha
structures and the sun microsystems sparc architectures.
untill then yor guess is a s good as mine, as the okld saying goes.
 MR > That's my say.  I'm very interested to here what kind of
 MR > calculations you did.  And if there's any kind of rough speed
 MR > testing I can do for you here (I've got XFree86 installed here,
 MR > and am using a Matrox Millenium card), let me know.
this sounds good to me .. i;ve been looking for a way to do some definitieve
do comparisons. sofar i have looked at this issue with an old engineers eye. 
as
i put at the foot of one of my earlier posts regarding this, or a similar
issue. i am a besoted motorola bigot, have been for twenty someting years, i
like the clean lines and clear architecural delineations in the 6800 seris
processor family
i also like to discuss processor theory, thoug as you may readily see i am
getting rusty, i am not making excuses for my self, god only knows i need to
cut myself some slack, i have been on full social security benefits for some 
10
year and i'm not ecpected to recover enough to work again .. so far i';ve
regaind the ability to talk and to walk, much to my freinds disgust .. tey
wanted a life sized doll to play with (sort of grin ok). i got up one morning
and found my brain shut down .. it eventually took me 6 years to get out of
bed, showered and dressed the longest moring i've ever had .. grin.
sorry, i am easly distracted these days, back to the system .. what sort of a
matrox do you have, mine, ok as ordered is a millenium II with 16 mb . yours
being ? how do you plan to account for the differences in cache size and 
spped.
also the 440fx chipset dosent support dimms (any body want 128 mb of dimms ?)
so i had to get 128 mb of ecc simms, well the reseller did and im not hs
faverite customer at the moment .. he forgot the differences in the chipsets
and ordered the wrong stuff..
can you thinkof any other differences that mich have to be accounted for ?
you also mention xfree86, is this a unix xfree86 or the german os/2 port ?
 MR > ... But how do we know your the REAL Angel of Death?
i just know, and yes, i learnt this very early in the piece.
Jonathan
ps, mike, i am tired and hurting more than usual, i may have been a bit more
jonathanesque than usuall, i've just reread my post and it strikes me as more
that usuall .. please read between the lines if you find the going tough.
maybe i should have put this at th front .. ummm
... i do what i can, with what i have, are you able to say the same ?
--- GoldED/2 3.00.Beta33022 UNREG
(3:712/808)
---------------