🚀 go-pugleaf

In article <4jfmka$c9j@cville-srv.wam.umd.edu>, veakblad@Glue.umd.edu
(David T. Wang) wrote:

] 
] : Let's talk actual systems, not estimated processor benchmarks.
] : January 1996 Byte, Page 140 Graph "Raw Performance: The VIS-less Ultra
] : I".  Sorted in order of integer performance:
] : ----------------------------------------
] : DeskStation Raptor 3 266 MHz Aplha 21164:                3.75, 4.49
] : IBM Personal Computer Power Series 850, 133 MHz PPC 604: 2.90, 2.92
] : Sun Ultra I Creator 3D Model 170 167 MHz UltraSparc:     2.18, 1.54
] : Intel Reference System, 150 MHz Pentium Pro:             2.10, 2.60
] : Netpower Fastseries MP 200 MHz Mips R4400:               2.03, 2.08
] 
] : PPC 604 is faster than Pentium Pro, even though the PPro is running at
] : a higher clock speed.
] 
] The above are Bytemark scores.  Try again when either....
] 
] 1) Bytemarks become a commercial application.
] 
] 2) You or Joe Ragosta find an app that runs faster on a PPC604 box than
] a P6 box.  Then you may try to make that assertion again.
] 

Mathematica?

As far as benchmarks go, all interested parties should check out the HINT
benchmarks homepage:

http://www.scl.ameslab.gov/HINT/

I think HINT goes a long way towards removing a lot of biases from many
benchmarks since the HINT evaluates the system as the problem scales from
small to large, showing clearly the effect of caches, external bus speed,
etc.  Give it a look-see.

Cheers,

Mel Martinez
The Johns Hopkins University
Dept. of Physics
mem@jhu.edu

In article <4jfmka$c9j@cville-srv.wam.umd.edu>, veakblad@Glue.umd.edu (David T. Wang) writes:
|> Christopher C. Wood (christw@meaddata.com) wrote:

David T. Wang:
|> : |> > > I did. They did not compare P6 and 604 boxes. You keep on
|> : |> > > telling me to look at Byte, which I did, and there was nothing
|> : |> > > there. 604 is a nice chip, "faster" than a P6? no.

|> Chris, which part of my assertion was erroneous?  

"They did not compare P6 and 604 boxes".

Chris
-- 
Speaking only for myself, of course.
Chris Wood    christw@lexis-nexis.com   cats@erinet.com

Christopher C. Wood (christw@meaddata.com) wrote:
: In article <4jfmka$c9j@cville-srv.wam.umd.edu>, veakblad@Glue.umd.edu (David T. Wang) writes:
: |> Christopher C. Wood (christw@meaddata.com) wrote:

: David T. Wang:
: |> : |> > > I did. They did not compare P6 and 604 boxes. You keep on
: |> : |> > > telling me to look at Byte, which I did, and there was nothing
: |> : |> > > there. 604 is a nice chip, "faster" than a P6? no.

: |> Chris, which part of my assertion was erroneous?  

: "They did not compare P6 and 604 boxes".

"With a commercial Application as a benchmark" it was implied, since my 
quote above was a direct response to Mr Ragosta's claim......
"PPC604 is faster than a P6 running commercial applications"

Did you miss that part?

: Chris
: -- 
: Speaking only for myself, of course.
: Chris Wood    christw@lexis-nexis.com   cats@erinet.com

Mel Martinez (mem@jhu.edu) wrote:

: Mathematica?

sure, Mathematica, like Photoshop, if I recall correctly, were more "Mac
Friendly" Apps, and did see to perform nicely on most Macs.  However,
I haven't seen any standardized Mathematica benchmark which compared a 
high end PPC604 to a PPro box.  Please post again when you have something.

: As far as benchmarks go, all interested parties should check out the HINT
: benchmarks homepage:

: http://www.scl.ameslab.gov/HINT/

: I think HINT goes a long way towards removing a lot of biases from many
: benchmarks since the HINT evaluates the system as the problem scales from
: small to large, showing clearly the effect of caches, external bus speed,
: etc.  Give it a look-see.

HINT is not a "Commercial Application" either...

but sure, are you sure that it's the best that you can do? You might as 
well use SPEC, which, with the release of a new compiler for PPC chips, 
now have pushed the PPC numbers up higher against the Pentium, but not
in the P6 range.   Here are some numbers....

SPEC...  (these numbers are used to scale comparison with the HINT numbers)
(all with optimizing compilers running on native OS  i.e. no excuses)

IBM 43P  PPC604 @ 133 MHz. W/512K L2 cache
http://www.specbench.org/osg/cpu95/results/res9509/p036.html
gets a SPEC95INT of 4.45  SPEC95FP of 3.31

Dell Dimension  Pentium@120 Mhz W/512K cache.
ftp://ftp.cdf.toronto.edu/pub/spectable
gets a SPEC95INT of 3.53  SPEC95FP of 2.26

Intergraph TDZ300 PPro@200 Mhz. 256K internal L2
http://www.specbench.org/osg/cpu95/results/res9512/p040.html
gets SPEC95INT of 8.04 SPEC95FP of 5.82

Now let's look at HINT with numbers obtained from page above.
(Powermac 9500 using metroworks c 6.0  Pentium using gcc with 386 optimization
fairly outdated optimizations)

Apple PowerMac 9500 @ 132 MHz 3.47 QMIPS
Pentium	@120 MHz 2.70 QMIPS 
No PPro numbers.

Hint says that a system using a 120 MHz Pentium is approximately 23% slower
than a 132 MHz Apple PMac 9500.  SPEC says that Integerwise, a system using
the 120 MHz Pentium is 20% slower than a PPC604@133 Mhz. (floating point,
the PPC604 is ~35% faster at this frequency comparison)  Remarkably, they
seem to agree to a certain extent, about the relative speed of a PPC604
chip (@132/3 MHz) vs a P5 @ 120 MHz.

There have been rumblings from Mr Ragosta on this precise topic about MacWay's
private mailings that with a new compiler, the PMac9500's numbers jumped to 
something like 5.7 QMIPs, and that the best they could obtain on the PPro
was a 4.0  This is simply evidence which shows that with better compilers,
these QMIPs score will no doubt climb, and HINT, like SPEC, are benchmarks
that are heavily compiler dependent.  As I've said, perhaps someone would
like to get a whole list of cross platform applications together, and 
users could just look at the subsets of the applications that they're likely
to use and look at the speed comparisons.  That Mathematica site, whose
URL I have now lost, seems like a good start. How about someone keeping 
track of all the application benchmark scores, from Photoshop to Word.

(personal side note. from looking at all those numbers above, I have come 
to the conclusion that the PPC604 is approximately 10% faster at the same
clock speed than Pentium, given roughly equivalent configurations, i.e. 
same L2 cache size.  However, Pentiums are currently available at a slightly 
faster clock speed. Incidently, PPC604's have 3.6 million transistors, and
Pentiums have 3.3 million, so the performace per transistor-MHz for the 
two chips are very very roughly, the same. However, Intel appears to have
an advantage at this time, due to difference in process technology. Intel's
133, 150 and 166 Mhz Pentiums are all fabbed on Intel's .35 micron process
technology, and IBM's PPC604 100, 120 and 133 MHz [the 150 Mhz process 
technology is under dispute, no reputable source for clarification] are
all fabbed on the .5 micron process. Bottom line is that the Pentium die
is 90 mm^2, and the PPC604 die is 196 mm^2 and so Intel could get 2 
pentiums out of the same piece of silicon for every 604 that IBM could
make.)

: Cheers,

: Mel Martinez
: The Johns Hopkins University
: Dept. of Physics
: mem@jhu.edu

In article <4jjusu$484@cville-srv.wam.umd.edu> David T. Wang,
veakblad@Glue.umd.edu writes:
>Hint says that a system using a 120 MHz Pentium is approximately 23% slower
>than a 132 MHz Apple PMac 9500.  SPEC says that Integerwise, a system using
>the 120 MHz Pentium is 20% slower than a PPC604@133 Mhz. (floating point,
>the PPC604 is ~35% faster at this frequency comparison)  Remarkably, they
>seem to agree to a certain extent, about the relative speed of a PPC604
>chip (@132/3 MHz) vs a P5 @ 120 MHz.

FWIW, I've used Pentium's of all speeds, Pentium Pro 150's, PPC601's and
PPC604's. Based on everyday use, I've found the 601 corresponds to the P5
and the 604 to the P6. Each CPU's excels in some areas and not in others.
None of the cpu's has a decisive advantage over its contemporary. As
usual, the best CPU depends on what you want to do with it.

My personal feeling is that the PMac 9500/120 and 132 pretty much trounce
any of the Pentium systems I have. At home my 8100/80 seems on par with
my P5/90 (the Pentium has faster video with a Matrox Millennium, and
using a 2940 SCSI card gives it comparable raw disk speed).

My gut feeling is that most of the general business apps I have on our
net run just fine on 601's and Pentiums. For these, a 604 and P6 are
overkill. CPU intensive apps (SPSS, CodeWarror, Visual C++, PhotoShop,
various 3D tools) cook on the 604 and on the P6, but a very small
percentage of my users run them. 

Either way, I'm waiting for PPCP 604e systems and debugged P6/200+ MMx
systems before I buy my next personal box. I'll make the choice then. 

WT

In article <4jjusu$484@cville-srv.wam.umd.edu>, veakblad@Glue.umd.edu
(David T. Wang) wrote:

Wow, a reasonable respndent. So:

> Mel Martinez (mem@jhu.edu) wrote:
> 
> : Mathematica?
> 
> sure, Mathematica, like Photoshop, if I recall correctly, were more "Mac
> Friendly" Apps, and did see to perform nicely on most Macs.  However,
> I haven't seen any standardized Mathematica benchmark which compared a 
> high end PPC604 to a PPro box.  Please post again when you have something.

MMA benchmarks are available on http://smc.vnet.net/mathbench.html. There
is nothing that I've seen on P6; the fastest Intel chip represented is a
166 mhz Pentium. It loses, although not by much, to the 132 mhz 604.

I would suggest that the PPro advocates post to the MMA list! If you want
to say that PPro is so much faster than the 604, please feel free to post
results on the MMA site. I think it speaks *volumes* about the true
availability of PPro machines that there is still no PPro result on the
MMA test site! The results are mailed in by users, but so far no results.
That is strange, since I would think that the power users who want P6
machines would also include at least a few MMA users.

Ironically, this thread was originally titled, over a year ago(!)
"PowerMac trounces Pentium with Mathematica". At the time, it was
certainly true: PowerMacs were running MMA tests TWICE as fast as the best
Pentium systems. The comment is no longer true; the P5 and P6 machines
have gained a lot of ground. A tribute to Intel. Also, the thing that is
never stated in the MMA tests is that MMA does so well on PMac because the
kernel is compiled with the IBM RS/600-based xlc compiler! How's that for
truth in advocacy?

> There have been rumblings from Mr Ragosta on this precise topic about MacWay's
> private mailings that with a new compiler, the PMac9500's numbers jumped to 
> something like 5.7 QMIPs, and that the best they could obtain on the PPro
> was a 4.0

I would like to hear more about that, something better than "private
communication" for a reference.

> (personal side note. from looking at all those numbers above, I have come 
> to the conclusion that the PPC604 is approximately 10% faster at the same
> clock speed than Pentium, given roughly equivalent configurations, i.e. 
> same L2 cache size.  However, Pentiums are currently available at a slightly 
> faster clock speed. 

I would quibble over the 10%...it's more like 20%, at least. This is
pretty well settled...the debate is over 604/604e vs P6.

>Incidently, PPC604's have 3.6 million transistors, and
> Pentiums have 3.3 million, so the performace per transistor-MHz for the 
> two chips are very very roughly, the same. However, Intel appears to have
> an advantage at this time, due to difference in process technology. Intel's
> 133, 150 and 166 Mhz Pentiums are all fabbed on Intel's .35 micron process
> technology, and IBM's PPC604 100, 120 and 133 MHz [the 150 Mhz process 
> technology is under dispute, no reputable source for clarification] are
> all fabbed on the .5 micron process. Bottom line is that the Pentium die
> is 90 mm^2, and the PPC604 die is 196 mm^2 and so Intel could get 2 
> pentiums out of the same piece of silicon for every 604 that IBM could
> make.)

Maybe. You forgot YIELD. Nobody knows the relative yields on 604 vs P5.
Conventional wisdom is that yields are supposed to be higher on RISC
chips?
I am no process engineer, so perhaps someone who is can provide some real
information. We sure aren't gonna find out from the companies...it's
information that they don't reveal. The only thing that I think can be
said fairly safely is that P6 yields are probably a LOT lower than P5 or
604...it can't be easy to get that huge full speed cache working on every
chip.

On Mon, 1 Apr 1996, Charles Bouldin wrote:

> MMA benchmarks are available on http://smc.vnet.net/mathbench.html. There
> is nothing that I've seen on P6; the fastest Intel chip represented is a
> 166 mhz Pentium. It loses, although not by much, to the 132 mhz 604.
> 
> I would suggest that the PPro advocates post to the MMA list! If you want
> to say that PPro is so much faster than the 604, please feel free to post
> results on the MMA site. I think it speaks *volumes* about the true
> availability of PPro machines that there is still no PPro result on the
> MMA test site! The results are mailed in by users, but so far no results.
> That is strange, since I would think that the power users who want P6
> machines would also include at least a few MMA users.

Well, I don't think that running Mathematica benchmarks would be of 
extremely high priority to those who own P6-200's, since I don't know too 
many people who purchase machines specifically to run Mathematica.  If I 
recall correctly, PPC604@133 boxes were made available last March (April?)
PPC604@150 boxes (PowerCC) were out last October, and P6@200 boxes were 
out in January96, (about same time as P5@166)  So I think that yes, P6@200 
boxes aren't as prevalent as P5@166's
 
> Ironically, this thread was originally titled, over a year ago(!)
> "PowerMac trounces Pentium with Mathematica". At the time, it was
> certainly true: PowerMacs were running MMA tests TWICE as fast as the best
> Pentium systems. The comment is no longer true; the P5 and P6 machines
> have gained a lot of ground. A tribute to Intel. Also, the thing that is
> never stated in the MMA tests is that MMA does so well on PMac because the
> kernel is compiled with the IBM RS/600-based xlc compiler! How's that for
> truth in advocacy?

Wow! truth in advocay indeed.  I wasn't aware of this.

> > (personal side note. from looking at all those numbers above, I have come 
> > to the conclusion that the PPC604 is approximately 10% faster at the same
> > clock speed than Pentium, given roughly equivalent configurations, i.e. 
> > same L2 cache size.  However, Pentiums are currently available at a slightly 
> > faster clock speed. 
> 
> I would quibble over the 10%...it's more like 20%, at least. This is
> pretty well settled...the debate is over 604/604e vs P6.

okay, whatever.  As SPEC showed, in floating point intensive applications,
the PPC604 runs about 20+ % faster, at the same clock rate (damn, I'm 
begining to sound like those who chant same clock rate, same clock rate)
I would think that Mathematica, would probably be heavily laden with 
floating point operations.  However, as I've said, doing mainly integer 
operations, the difference probably ought to be in the 10% range. 
Lastly, as I've also pointed out, Pentiums are available at higher clock 
rates than PPC604's.

> >Incidently, PPC604's have 3.6 million transistors, and
> > Pentiums have 3.3 million, so the performace per transistor-MHz for the 
> > two chips are very very roughly, the same. However, Intel appears to have
> > an advantage at this time, due to difference in process technology. Intel's
> > 133, 150 and 166 Mhz Pentiums are all fabbed on Intel's .35 micron process
> > technology, and IBM's PPC604 100, 120 and 133 MHz [the 150 Mhz process 
> > technology is under dispute, no reputable source for clarification] are
> > all fabbed on the .5 micron process. Bottom line is that the Pentium die
> > is 90 mm^2, and the PPC604 die is 196 mm^2 and so Intel could get 2 
> > pentiums out of the same piece of silicon for every 604 that IBM could
> > make.)
> 
> Maybe. You forgot YIELD. Nobody knows the relative yields on 604 vs P5.
> Conventional wisdom is that yields are supposed to be higher on RISC
> chips?

"RISC" is an architectural design concept. it could be implemented 
independent of process technology.  If I had Bill Gate's and Howard 
Hugh's money and insanity, I could build a mechnical "RISC" machine (the 
Babbage thing)  I don't know where the conventional wisdom you've cited 
comes from.  I would most certainly have to say that it's not true. 

> I am no process engineer, so perhaps someone who is can provide some real
> information. We sure aren't gonna find out from the companies...it's
> information that they don't reveal. The only thing that I think can be
> said fairly safely is that P6 yields are probably a LOT lower than P5 or
> 604...it can't be easy to get that huge full speed cache working on every
> chip.

I don't know about "a LOT lower". Caches in general are easier to fabricate 
and test than a chip full of random logic.  Yield might even be higher than 
an equivalent sized chip used to build a CPU. (I'm no cache designer, but 
from the looks of it, you probably wouldn't need all the levels of 
routing and interconnects that you would need on a cpu, so you can 
*probably* get away with less masking layers in the fabrication step. 
simpler process==higher yield.)  So you can make fewer P6's than P5's off 
of the same line?  yes, the dies are physically bigger, so you get less 
chips.  "a LOT lower" due to process faults? no, I wouldn't believe so.

bouldin@enh.nist.gov (Charles Bouldin) wrote:

> Maybe. You forgot YIELD. Nobody knows the relative yields on 604 vs P5.
> Conventional wisdom is that yields are supposed to be higher on RISC
> chips?

There is no reason to assume that defects/square inch on a wafer
care about instruction sets.

> I am no process engineer, so perhaps someone who is can provide some real
> information. We sure aren't gonna find out from the companies...it's
> information that they don't reveal. The only thing that I think can be
> said fairly safely is that P6 yields are probably a LOT lower than P5 or
> 604...it can't be easy to get that huge full speed cache working on every
> chip.

The CPU and the cache in the PP are two separate chips, even when
they are mounted on the same chip carrier. ( But since the number
of available transistors on a chip doubles every 18 months or so,
prediction of the single chip PPro arrival date is left as an
exercise to the reader.)

Gerhard

In article <4jsaco$fe9@unlisys.unlisys.net>, Gerhard Hoffmann
<ghf@berlin.snafu.de> wrote:

> The CPU and the cache in the PP are two separate chips, even when
> they are mounted on the same chip carrier. ( But since the number
> of available transistors on a chip doubles every 18 months or so,
> prediction of the single chip PPro arrival date is left as an
> exercise to the reader.)
> 
> Gerhard

The funny thing is that PC Week says that Intel is going to sell the PP
and cache as separate devices in the future. Apparently, the decided that
selling them on the same mounting wasn't such a great idea.

-- 
Regards,       Joe Ragosta

Copyright Joseph M. Ragosta, 1996. Non-exclusive, royalty free
license to distribute this post granted to any service provider 
except Microsoft. By posting this, Microsoft agrees to pay $1,000 per 
posting.

In article <4jjusu$484@cville-srv.wam.umd.edu>, veakblad@Glue.umd.edu
(David T. Wang) wrote:


> 
> Now let's look at HINT with numbers obtained from page above.
> (Powermac 9500 using metroworks c 6.0  Pentium using gcc with 386 optimization
> fairly outdated optimizations)
> 
> Apple PowerMac 9500 @ 132 MHz 3.47 QMIPS
> Pentium	@120 MHz 2.70 QMIPS 
> No PPro numbers.

David, I've answered this before. If you're really so interested in
getting the truth, you can get it simply by sending an e-mail to the HINT
people at Ames Lab.

They are slow updating their web site, but they are currently getting a
HINT value of 5.9 on the PowerMac 9500. The best they've been able to do
with Pentium Pro is around 4.

> There have been rumblings from Mr Ragosta on this precise topic about MacWay's
> private mailings that with a new compiler, the PMac9500's numbers jumped to 
> something like 5.7 QMIPs, and that the best they could obtain on the PPro
> was a 4.0  

Which of course you could verify if you bothered to send them an e-mail.

-- 
Regards,       Joe Ragosta

Copyright Joseph M. Ragosta, 1996. Non-exclusive, royalty free
license to distribute this post granted to any service provider 
except Microsoft. By posting this, Microsoft agrees to pay $1,000 per 
posting.

Joe Ragosta (jragosta@dca.net) wrote:
: The funny thing is that PC Week says that Intel is going to sell the PP
: and cache as separate devices in the future. Apparently, the decided that
: selling them on the same mounting wasn't such a great idea.

I didn't see the article you were referring to, is this going to happen 
to all P6s or are they referring to a low cost version of the P6 that 
will have an external L2. I would suspect the latter, onboard L2 has a 
significant performance advantage. Basically, your "wasn't such a great 
idea" argument may be true when looking solely at price, but it is false 
when looking solely at performance.

Tony
-- 
------------------
Tony Tribelli
adtribelli@acm.org

In article <jragosta-0404960805470001@ppp-1008.dca.net>, jragosta@dca.net 
says...
>
>In article <4jsaco$fe9@unlisys.unlisys.net>, Gerhard Hoffmann
><ghf@berlin.snafu.de> wrote:
>
>> The CPU and the cache in the PP are two separate chips, even when
>> they are mounted on the same chip carrier. ( But since the number
>> of available transistors on a chip doubles every 18 months or so,
>> prediction of the single chip PPro arrival date is left as an
>> exercise to the reader.)
>> 
>> Gerhard
>
>The funny thing is that PC Week says that Intel is going to sell the PP
>and cache as separate devices in the future. Apparently, the decided 
that
>selling them on the same mounting wasn't such a great idea.
>

That's one conclusion, but not the only one. It may be that they want
to target 2 different markets. The single package version is best
suited for high end applications. The off-package level 2 version
would be a better choice for the desktop. Intel has made it quite
clear that they are going at the highend server market and they will
continue to make and upgrade the single package version.

Elmo

In article <bouldin-0104961111580001@dafs.nist.gov>, bouldin@enh.nist.gov
(Charles Bouldin) wrote:

> Maybe. You forgot YIELD. Nobody knows the relative yields on 604 vs P5.
> Conventional wisdom is that yields are supposed to be higher on RISC
> chips?

Yield depends on die area, process technology -- a lot of things. If all
else is equal, the smaller chip will have a higher yield.
-- 
Philip Machanick             Department of Computer Science
University of the Witwatersrand,   2050 Wits,  South Africa
phone 27(11)716-3309                     fax 27(11)339-7965
http://www.cs.wits.ac.za/~philip/       philip@cs.wits.ac.za

elmo@ns.nst (Elmo D. Rothrock) wrote:


>That's one conclusion, but not the only one. It may be that they want
>to target 2 different markets. The single package version is best
>suited for high end applications. The off-package level 2 version
>would be a better choice for the desktop. Intel has made it quite
>clear that they are going at the highend server market and they will
>continue to make and upgrade the single package version.

Just been to an intel conference.
They push the PPro on chip L2 cache to allow multiprocessors to run
without  sharing cache, therefore no big performance hit when adding
processors.

Also they pushed cache on a stick, which allows you to change l2 cache
on board like simms ram.

PPro for multiprocessors with l2 on chip makes sense, with a cut down
version without l2 on chip for home users maybe.  But then such a chip
would be too much the cyrix 686 for intels comfort.....