HP PA-RISC based Amiga
sp...@drycas.club.cc.cmu.edu
For all of you critics of CBM's decision to go with the HP PA-RISC
architecture instead of PowerPC here's some info on HP's new workstation
using their new 7100LC Risc chip. From the specs its looks good. The
next generation Amiga should be like this. This info was gathered
from comp.sys.next.advocacy.
_______________________________________________________________________________
HP Gecko
Champagne workstation at beer prices!
It's been a long wait, but the world's most advanced RISC workstation
is here! The HP 712 series, or Gecko, is a major advance in desktop
workstation technology. Using the "bi-endian" PA7100LC as its main
CPU, the Gecko can adapt to whatever operating system you choose.
Onboard MPEG hardware compression/decompression will display your
realtime news and video feeds smoothly and quickly. And with a
starting price of $3995--including 15" color monitor, 16 MBs of RAM,
Ethernet, and a 260 MB disk--advanced workstation technology is
affordable to anyone!
With the new Gecko, you can choose the level of performance that's
appropriate, from the base 60 MHz model to the integer-tuned,
number-crunching 80i. Here's a look at HP Gecko performance
benchmarks.
712/60 712/80i
---------- -----------
SPECint92 58 84
SPECfp92 79 79
MPEG (640 x 480) 30 fps 30 fps
The HP Gecko base configuration includes Desktop HP-UX, a slimmer
version of HP's Unix that leaves you 100 MBs for your files. You also
get CD-quality, 16-bit audio for clear, crisp playback of sound and
music. You can choose from expansive 17" or 19" color monitors, or
even an active-matrix color, flat panel display for Energy Star
compliance. If you require higher storage capacity, choose from an
available 525 MB or 1.0 GB SCSI disk. And you can decide on the
amount of memory upgradeability that's right for you--choose from 16
MB up to 128 MBs of RAM!
But the Gecko's revolutionary design includes more than mere
blistering performance. With the introduction of the Gecko, HP
proudly announces TeleShare--the ability to tie your Gecko into phone
systems without expensive components. With TeleShare, you can
incorporate voice technology into your applications today! TeleShare
eliminates the need for costly external modems, with data
capabilities built right in your desktop.
The base configuration includes an RS-232C serial port, an AUI port,
one SCSI-2 port, one Centronics port, X.25 and telephony ports, and
two slots.
HP 712/60
16 MBs DRAM 3879
260 MB SCSI disk
Ethernet
CD-quality audio and MPEG
15" color monitor
Add-on products
-----------------------
ADDITIONAL MEMORY
---------------------------------
8 MBs RAM (2 x 4 MB SIMMs) 620
16 MB RAM (2 x 8 MB SIMMs) 1,170
32 MB RAM (2 x 16 MB SIMMs) 4,340
64 MBs RAM (2 x 32 MB SIMMs) 8,740
INTERNAL STORAGE DEVICES
---------------------------------
260 MB SCSI-2 470
525 MB SCSI-2 1,120
1 GB SCSI-2 1,995
OTHER
---------------------------------
HP Token Ring/9000 for Model 712 645
Software driver on DDS tape 2395
Second serial port 265
Second LAN AUI and second serial port 645
Second monitor port 1,395
X.25 port with second serial port 795
Bohuslav Rychlik
based Amiga by sp...@drycas.club.cc.cmu
> architecture instead of PowerPC here's some info on HP's new workstation
I wasn't aware that Commodore has chosen a RISC CPU for its next
generation of machines, yet. As far as I know, the only thing they are
doing with HP so far is supplying them with AGA chips for possible
set-top units and possibly using some HP technology in their future
graphic system.
Excerpts from netnews.comp.sys.amiga.advocacy: 19-Jan-94 HP PA-RISC
based Amiga by sp...@drycas.club.cc.cmu
>number-crunching 80i. Here's a look at HP Gecko performance
>benchmarks.
>
> 712/60 712/80i
> ---------- -----------
>SPECint92 58 84
>SPECfp92 79 79
Well, I'd still prefer the PowerPC. The low power consumption (3 W max,
1.5 W typical) PowerPC 603 is expected to hit 75 SPECint and 85 SPECfp
when running at 80 MHz, and 60.6 SPECint and 72.2 SPECfp at 66 MHz. In
any case, here is the RISC Comparison chart from 1/94 issue of BYTE:
Item DECchip 21066 PowerPC 603 MicroSparc II MIPS/NEC R4200
---- ------------- ----------- ------------- --------------
Transistors 1.75 1.5 2.3 1.3
(million)
max Clock (MHz) 166 80 85 80
max Power (W) 20+ 3 5 2
Size (mm^2) 209 85 233 81
SPECint92 70 75 57.2 55
SPECfp92 105 85 49.5 30
Price ($/1000) 424 n/a 500 75
(Note: Perfomance figures except for MicroSparc II based on simulations.)
-------------------------------------------------------------
Bohuslav Rychlik (Bob), ryc...@cmu.edu, br...@andrew.cmu.edu
-------------------------------------------------------------
Bjorn Reese
> Excerpts from netnews.comp.sys.amiga.advocacy: 19-Jan-94 HP PA-RISC
> based Amiga by sp...@drycas.club.cc.cmu
> > For all of you critics of CBM's decision to go with the HP PA-RISC
> > architecture instead of PowerPC here's some info on HP's new workstation
>
> I wasn't aware that Commodore has chosen a RISC CPU for its next
> generation of machines, yet. As far as I know, the only thing they are
> doing with HP so far is supplying them with AGA chips for possible
> set-top units and possibly using some HP technology in their future
> graphic system.
> Well, I'd still prefer the PowerPC. The low power consumption (3 W max,
Neither have I heard anything official, but read the signs on the wall!
PowerPC have been out of question for quite some time; not because of
lacking performance, but because IBM has too much control over it - I
don't remember the exact details, as I wasn't paying much attention
(I was never interested in contract issues anyway.) The spokespersons
from C= have said several times that it was down to either MIPS or
HP PA-RISC. Now, go to comp.arch and read how well informed and dedicated
Randall Jesup (C=) seems to be about the PA-RISC... makes you wonder.
BTW, I'm getting a bit bored with different people advocating their
favorite RISC. The simple fact is that no matter how good arguments
we manage to come up with, it isn't going to change a damn thing about
C='s decision - whatever that may be. And you can be sure that the
engineers at C= will examine each RISC very carefully, before making
their decision.
--
+-------------------------------+--------------------------------------+
| Bjorn Reese | Email: bre...@imada.ou.dk |
| Dexion Design | (bre...@bohr.ot.dk) |
| Certified Developers | Odense University, Denmark |
+-------------------------------+--------------------------------------+
"It's getting late in the game to show any pride or shame" - Marillion
Mike Rogers
>HP PA-RISC. Now, go to comp.arch and read how well informed and dedicated
>Randall Jesup (C=) seems to be about the PA-RISC... makes you wonder.
Yes, but C= will have to work hard to compete with the HP Gecko.
Mind you, the peripheral costs *are* outrageous.
--
Mike Rogers,#3,44Westland##EveryoneHasTheRightToFreedomOfOpinionAndExpressionT
Row,Dublin2,Ireland Perl ##hisRightIncludesFreedomToHoldOpinionsWithoutInterfe
##############################renceAndToSeekReceiveAndImpartInformationAndIdea
sThroughAnyMediaAndRegardlessOfFrontiers#10 UN Declaration Of HumanRights Kibo
Gregory R Block
: workstation technology. Using the "bi-endian" PA7100LC as its main
Ha! Even RISC processors are coming out of the closet. Rot in hell,
fundies.
Greg
--
(: (: (: (: Have you overdosed on smileys today? Why NOT!?! :) :) :) :)
(: "He uses statistics like a drunkard uses a light-post; :)
(: for support, not for illumination." -Anonymous :)
(: (: (: (: (: (: (: (: (: (: (: (: :) :) :) :) :) :) :) :) :) Wubba :)
Dan Higgins
Gregory R Block <gbl...@csd4.csd.uwm.edu> wrote:
>
>In article <1994Jan19...@drycas.club.cc.cmu.edu>,
>sp...@drycas.club.cc.cmu.edu wrote:
>:
>: workstation technology. Using the "bi-endian" PA7100LC as its main
>
>Ha! Even RISC processors are coming out of the closet. Rot in hell,
>fundies.
>
>Greg
HEE HEE!
Yo, clue me in on this 'fundies' gag. I want in on the fun: Die, scum.
- Dan
+--------------------------+--------------------------------------+
| Daniel Higgins | Associate Tech/Software Engineer |
| da...@ccwf.cc.utexas.edu | Texas Research Institute, Austin |
| | and lowly Computer Science Undergrad |
| // +-------------------+--------------------------------------+
| \X/ | "Politics - who gets what, when, and how." - unknown |
+------+----------------------------------------------------------+
Francis Hsu
: Excerpts from netnews.comp.sys.amiga.advocacy: 19-Jan-94 HP PA-RISC
: based Amiga by sp...@drycas.club.cc.cmu
[...]
: >number-crunching 80i. Here's a look at HP Gecko performance
: >
: > 712/60 712/80i
: > ---------- -----------
: >SPECint92 58 84
: >SPECfp92 79 79
There has been speculation that HP purposefully compiled some of the SPECfp
benchmarks with less optimization on the 712/80i. The 80i uses the exact
same CPU clocked at 80MHz and in addition has a 256k cache instead of the
64k one found in the 712/60. A propbable reason for this marketing ploy is
that the 712/80i could well draw away sales from the more expensive 715/735
models. The "real" SPECfp92 is probably more on the order of 110.
: Well, I'd still prefer the PowerPC. The low power consumption (3 W max,
: 1.5 W typical) PowerPC 603 is expected to hit 75 SPECint and 85 SPECfp
: when running at 80 MHz, and 60.6 SPECint and 72.2 SPECfp at 66 MHz. In
If C= was into laptop computers, perhaps I would agree. But for a desktop
computer, I think the 8W consumption (don't know if that's average or
max) of the 7100LC is quite reasonable.
Also, from what I've read the 603 isn't expected to ship until 3Q.
: any case, here is the RISC Comparison chart from 1/94 issue of BYTE:
I'm appending the PA-7100LC column based on data from Electronic News and
EE Times (thanks Andre) SPEC figures based on information by HP:
(also editing the columns to provide room)
: Item DEC 21066 MPC 603 uSparc II R4200 PA-7100LC
: ---- --------- ------- --------- ----- ---------
: Transistors 1.75 1.5 2.3 1.3 .8
: (million)
: max Clock (MHz) 166 80 85 80 80 or 100?
: max Power (W) 20+ 3 5 2 8+
: Size (mm^2) 209 85 233 81 196
: SPECint92 70 75 57.2 55 84
: SPECfp92 105 85 49.5 30 79+
: Price ($/1000) 424 n/a 500 75 120+
A few notes:
Electronic News/EE Times says the PA-7100LC is scalable to 100MHz, but
the fastest Gecko is the 712/80i and those are where I drew the SPEC
figures from. They also say the expected price is $120 in "volume" and
that Microprocessor Forum has speculated the manufacturing costs to be
1/3 that of the Pentium.
Once again, the "true" SPECfp of the 7100LC @ 80MHz is probably more like
110.
: -------------------------------------------------------------
: Bohuslav Rychlik (Bob), ryc...@cmu.edu, br...@andrew.cmu.edu
: -------------------------------------------------------------
Francis
Andre Yew
>: > 712/60 712/80i
>: > ---------- -----------
>: >SPECint92 58 84
>: >SPECfp92 79 79
>There has been speculation that HP purposefully compiled some of the SPECfp
>benchmarks with less optimization on the 712/80i. The 80i uses the exact
>same CPU clocked at 80MHz and in addition has a 256k cache instead of the
>64k one found in the 712/60. A propbable reason for this marketing ploy is
>that the 712/80i could well draw away sales from the more expensive 715/735
>models. The "real" SPECfp92 is probably more on the order of 110.
It should also be noted that some systems (eg. DG's 88110
machine) have produced better SPECs with less optimization. This
may be the result of a strange interaction with the cache (which
is onchip, to boot). BTW, my main workstation is a DG 88110 Aviion,
and it gets SPECs of about 30-something/40. Running UNIX and X,
it is a very responsive system --- there is little doubt in my
mind that something as unbloated as the AmigaOS would be wonderfully
fast. I think that all this carping about who has the higher
SPEC ratings (do you even know what the individual tests mean,
especially with respect to what you want to do?), is all just so much
posing and oneupmanship with very little bearing on what the machine
will do. Even worse, these SPEC marks are off by less than 10%
in many cases, and SPEC tests have been known to vary by over 3 times
that much on consecutive tests with the exact same machine and
machine code! BTW, this is not aimed directly at you, Francis,
just a general fan of flame for those people who like to compare
SPECs :-).
I think the only people who truly know today are perhaps
some hardworking engineers somewhere in Pennsylvania.
--Andre
--
PGP public key available
Bohuslav Rychlik
based Amiga by Bjorn Re...@imada.ou.dk
> > generation of machines, yet. As far as I know, the only thing they are
> > doing with HP so far is supplying them with AGA chips for possible
> > set-top units and possibly using some HP technology in their future
> > graphic system.
> [...]
> > Well, I'd still prefer the PowerPC. The low power consumption (3 W max,
>
> Neither have I heard anything official, but read the signs on the wall!
> PowerPC have been out of question for quite some time; not because of
> lacking performance, but because IBM has too much control over it - I
> don't remember the exact details, as I wasn't paying much attention
> (I was never interested in contract issues anyway.) The spokespersons
> from C= have said several times that it was down to either MIPS or
> HP PA-RISC. Now, go to comp.arch and read how well informed and dedicated
> Randall Jesup (C=) seems to be about the PA-RISC... makes you wonder.
May make me wonder, but doesn't really tell me anything concrete. If a
RISC CPU hasn't been choses yet officially, it hasn't been chosen yet,
unless anyone has more definite info. I never heard anything about the
PowerPC being "out of the questions" except by PA-RISC advocates here.
I don't understand why using the PowerPC would create problems when IBM
& Motorola hope that the PowerPC will be the CPU-of-choice in the next
decade for system vendors currently thriving in the PC clone market -
though that's not to say that these PowerPC machines will be "clones."
If anything, it seems that HP is more closed about it's PA-RISC line,
though that probably wouldn't be a problem for Commodore due to the two
companies' "relationship."
Excerpts from netnews.comp.sys.amiga.advocacy: 21-Jan-94 Re: HP PA-RISC
based Amiga by Bjorn Re...@imada.ou.dk
>BTW, I'm getting a bit bored with different people advocating their
>favorite RISC. The simple fact is that no matter how good arguments
>we manage to come up with, it isn't going to change a damn thing about
>C='s decision - whatever that may be. And you can be sure that the
>engineers at C= will examine each RISC very carefully, before making
>their decision.
I understand what you're saying, but I find the RISC discussion one of
the best in .advocacy. Plus, you never know - it could change things.
It's something I'm concerned about so I like to talk about it. Way I
see it, if Commodore blunders on this decision, that's the last blunder
for the Amiga line of personal computers (though that might not mean the
end for Commodore).
David Meiklejohn
>I understand what you're saying, but I find the RISC discussion one of
>the best in .advocacy.
Nearly as good as beer and football, yes. :-)
> Plus, you never know - it could change things.
Highly unlikely. Commodore have some very talented engineers, who have
probably thrashed out most of the issues long before they get discussed
here. This is for our amusement only.
>It's something I'm concerned about so I like to talk about it. Way I
>see it, if Commodore blunders on this decision, that's the last blunder
>for the Amiga line of personal computers (though that might not mean the
>end for Commodore).
Blunder? No, it's not that serious. The whole reason we've been able to
talk about this for so long, in an intelligent way with good arguments on
all sides, is simply that there are several good choices.
Whether C= go with PowerPC, PA-RISC or even MIPS, they'll end up with a fast
line of CPUs that can run NT. Assuming the contenders are PPC and PA, they
are both backed by big companies and industry consortiums. Given that
the machine architecture at this level has as much to do with benchmarked,
real and perceived performance (three different things) as the cpu does,
the difference between, say 50 and 70 SPECints is irrelevant. When you get
right down to it, they're both damn fast. And how fast it *seems* will,
for ordinary use, depend much more on video and hard drive speeds.
Sure, only one choice is "best", but that doesn't make the other options
far off it.
--
David Meiklejohn (dav...@qdpii.ind.dpi.qld.gov.au)
- No longer working for the Queensland Department of Primary Industries
Francis Hsu
: fra...@red-branch.MIT.EDU (Francis Hsu) writes:
: >: > 712/60 712/80i
: >: > ---------- -----------
: >: >SPECint92 58 84
: >: >SPECfp92 79 79
: >There has been speculation that HP purposefully compiled some of the SPECfp
: >benchmarks with less optimization on the 712/80i. The 80i uses the exact
: >same CPU clocked at 80MHz and in addition has a 256k cache instead of the
: >64k one found in the 712/60. A propbable reason for this marketing ploy is
: >that the 712/80i could well draw away sales from the more expensive 715/735
: >models. The "real" SPECfp92 is probably more on the order of 110.
: It should also be noted that some systems (eg. DG's 88110
: machine) have produced better SPECs with less optimization. This
: may be the result of a strange interaction with the cache (which
: is onchip, to boot). BTW, my main workstation is a DG 88110 Aviion,
[...]
: fast. I think that all this carping about who has the higher
: SPEC ratings (do you even know what the individual tests mean,
Personally? I don't care about the individual benchmarks or even the
overall figure save for bragging rights. But that's because closest I
come to doing any serious work on computers is compiling applications.
Fact is though that SPEC sells. I know, having sold workstations to
grad students and faculty here. Knowing SPECint and SPECfp figures has
convinced customers to pick an Alpha workstation over a Sparc for instance.
Then I note to them that application availability on the Alpha platform is
much less than Sun and only then they begin to think a little more carefully.
I argue SPEC numbers because for C='s sake, they need all the freebie
marketing tools they can get. If they claim that the next generation Amiga
is 25% faster than PowerPC/Pentium machines, whatever.. at least they'll
have the numbers (numbers that alot of people give credence to) to back
their claim.
Anyways, here's a table of individual SPECfp results for the 712/60 and
712/80i:
712/60 712/80i
Benchmark SPEC SPEC
No. & Name Ratio Ratio
__________ _____ _____
013.spice2g6 48.3 70.0
015.doduc 58.9 71.0
034.mdljdp2 89.0 80.1*
039.wave5 66.2 61.4*
047.tomcatv 108.2 94.6*
048.ora 149.6 116.3*
052.alvinn 88.2 112.1
056.ear 132.4 125.5*
077.mdljsp2 52.9 37.9*
078.swm256 73.8 52.8*
089.su2cor 75.5 100.2
090.hydro2d 73.5 77.9*
093.nasa7 59.8 70.0
094.fpppp 82.4 89.1*
SPECfp92 78.5 79.0
Followed by an interesting post:
Other than it being a subtle slam at HP (when you look at the last paragraph),
the key is that the compiler switches were different only on the benchmarks
where the 80i turned surprisingly low results.
>From: nei...@bier.kar.dec.com (Burkhard Neidecker-Lutz)
>Newsgroups: comp.arch
>Subject: Re: SPECfp92 of HP 9000 712/60 & 712/80i
>Date: 22 Jan 1994 13:08:19 GMT
>In article <1994Jan21.2...@icaen.uiowa.edu> dsie...@icaen.uiowa.edu (Doug Siebert) writes:
>>
>>I'd have to assume some FP instructions were removed and are done in software
>>on the faster chip having seen the table of results.
>No. It's the same chip.
>> Don't want to steal market share away from the
>>715/75 (which this would beat by 8% or so were it a full implementation) or
>>the 735 (which ths would fall short of by only 25%)
>That's it. A marketing game. If you look at the full SPEC disclosure, you
>will find that all machines (regardless of whether theya are based on the
>7100LC or the 7100 use the same compiler switches for almost all of the
>benchmarks (i.e. they run the same code). The results for the 80i in the
>SPECfp92 suite are generated with different switches for the following
>benchmarks:
>034.mdljdp2 039.wave5 047.tomcatv 048.ora 056.ear
>077.mdljsp2 078.swm256 093.nasa7 094.fpppp
>for which it happens to come out slower than the 60 Mhz machine. To give
>you an idea what they did, the options for swm256:
> 80 Mhz switches: -O
> 60 Mhz switches: OPT=+OP3 +O2 EXTRA_FFLAGS=-WP,'-nofuse -srlcd'
>Sort of reminds me of the bad old days at Digital in the late 80's when
>we played such positioning games between our machines. Summary: The real
>SPECfp92 of the 712/80i is more like the clock rate difference to the
>60 Mhz machine, hence forget about buying the 735 and 715 machines.
> Burkhard Neidecker-Lutz
>Distributed Multimedia Group, CEC Karlsruhe
>BERKOM II Project
>Digital Equipment Corporation
>nei...@nestvx.enet.dec.com
: machine code! BTW, this is not aimed directly at you, Francis,
: just a general fan of flame for those people who like to compare
: SPECs :-).
Doesn't matter to me.. I like to compare numbers. :)
: --Andre
: --
: PGP public key available
Francis
Andre Yew
>Personally? I don't care about the individual benchmarks or even the
>overall figure save for bragging rights. But that's because closest I
>come to doing any serious work on computers is compiling applications.
>Fact is though that SPEC sells. I know, having sold workstations to
>grad students and faculty here. Knowing SPECint and SPECfp figures has
>convinced customers to pick an Alpha workstation over a Sparc for instance.
>Then I note to them that application availability on the Alpha platform is
>much less than Sun and only then they begin to think a little more carefully.
Yes, I agree completely because I've done it, too. However,
people here were arguing technical merit, which is, unfortunately,
separate from marketing value :(. Ah, now it's my turn to throw an
article excerpt at you, and it's from one of my past ones to boot! :-)
You'll find the results quoted from comp.arch interspersed.
In reality, for home use, SPECfp isn't terribly important.
If HP were out to protect their market, it may have been for the
7100-based machines in their scientific market, where FP means quite
a bit more. It's also interesting to see that the places where
the 80MHz chip won in places where there's quite a bit of integer
math going on like spice (for graphs and matrices) and nasa7 (matrices).
SPECFP92 suite:
> 712/60 712/80i
>Benchmark SPEC SPEC
>No. & Name Ratio Ratio
>__________ _____ _____
>SPECfp92 78.5 79.0
>013.spice2g6 48.3 70.0
013.spice2g6 ... Circuit simulation package that simulates a typical
circuit - grey code. Written in Fortran using Double
Precision floating point.
>015.doduc 58.9 71.0
015.doduc ... A Monte Carlo simulation of the time evolution of a
nuclear reactor component. Written in Fortran using
Double Precision floating point.
>034.mdljdp2 89.0 80.1*
034.mdljdp2 ... A chemical application program that solves equations of
motion for a model of 500 atoms. This is similar to
modeling a structure of liquid Argon. Written in
Fortran using Double Precision floating point.
>039.wave5 66.2 61.4*
039.wave5 ... A two dimensional electro-magnetic particle-in-cell
simulation used to study various plasma phenomena.
The benchmark solves equations of motion on a cartesian
mesh involving 500,000 particles on 50,000 grid points
for 5 time steps. Written in Fortran using
Double Precision floating point.
>047.tomcatv 108.2 94.6*
047.tomcatv ... A vectorized mesh generation program. Written in
Fortran using Double Precision floating point.
Hmm, this next one could be important to diehard, multitasking
Amiga owners everywhere especially when they're formatting three
floppies at once and downloading at 14.4 kbaud :).
>048.ora 149.6 116.3*
048.ora ... Traces rays through optical systems composed of
spherical and plane surfaces. Written in Fortran
using Double Precision floating point.
(Oops, accidentally lost this one in the edits. The 80MHz chip won here.)
052.alvinn ... Trains a neural network using back propagation.
This is written in single precision C.
>056.ear 132.4 125.5*
056.ear ... An inner ear model which filters and detects various
sounds and generates speech signals. Written in
C using Single Precision floating point.
>077.mdljsp2 52.9 37.9*
077.mdljsp2 ... (Same as mdljdp2 but in single precision) A chemical
application program that solves equations of motion for
a model of 500 atoms. This is similar to modeling a
structure of liquid Argon. Written in Fortran
using Single Precision floating point
>078.swm256 73.8 52.8*
078.swm256 ... A shallow water model that solves shallow water
equations using finite difference equations with
a 256X256 grid. Written in Fortran using Single
Precision floating point
>089.su2cor 75.5 100.2
089.su2cor ... Computes masses of elementary particles in the framwork
of the Quark-Gluon theory. Written in FORTRAN using
Double Precision floating point. It is vectorizable.
>090.hydro2d 73.5 77.9*
090.hydro2d ... An astrophysics application program which solves hydro-
dynamical Navier Stokes equations to compute galactical
jets. Written in Fortran using Double Precision floating
point.
>093.nasa7 59.8 70.0
093.nasa7 ... Consists of 7 application level routines that do matrix
manipulation, FFTs, Gaussian elimination, create
vortices. The routines are MXM, CFFT2D, CHOLSKY,
BTRIX, GMTRY, EMIT and VPENTA. Written in Fortran Double
Precision floating point.
>094.fpppp 82.4 89.1*
094.fpppp ... A quantum chemistry application program used to
calculate two electron integral derivative which occurs
in the Gaussian series of programs. Written in Fortran
using Double Precision floating point.
>Other than it being a subtle slam at HP (when you look at the last paragraph),
>the key is that the compiler switches were different only on the benchmarks
>where the 80i turned surprisingly low results.
You never know, sometimes those are the fastest achievable
results. It's happened before.
Bohuslav Rychlik
based Amiga by Francis Hsu@red-branch.M
> : > ---------- -----------
> : >SPECint92 58 84
> : >SPECfp92 79 79
>
> There has been speculation that HP purposefully compiled some of the SPECfp
> benchmarks with less optimization on the 712/80i. The 80i uses the exact
> same CPU clocked at 80MHz and in addition has a 256k cache instead of the
> 64k one found in the 712/60. A propbable reason for this marketing ploy is
> that the 712/80i could well draw away sales from the more expensive 715/735
> models. The "real" SPECfp92 is probably more on the order of 110.
This must be strangest defense of a processor's performance I have yet
heard. Although the similarity in the numbers is disconcerting between
the 60 and 80 MHz parts, I prefer to go with the numbers as posted,
myself.
Excerpts from netnews.comp.sys.amiga.advocacy: 23-Jan-94 Re: HP PA-RISC
based Amiga by Francis Hsu@red-branch.M
>Also, from what I've read the 603 isn't expected to ship until 3Q.
BUt PowerPC 601 is shipping now and according to Motorola the numbers
for that microprocessor at 80MHz are in the range of SPECin 85 and
SPECfp 105+.
Excerpts from netnews.comp.sys.amiga.advocacy: 23-Jan-94 Re: HP PA-RISC
based Amiga by Francis Hsu@red-branch.M
>Once again, the "true" SPECfp of the 7100LC @ 80MHz is probably more like
>110.
I wouldn't bet on that kind of speculation.
Francis Hsu
: 013.spice2g6 ... Circuit simulation package that simulates a typical
: circuit - grey code. Written in Fortran using Double
: Precision floating point.
[...]
Yeah.. I've read the file on the different SPEC benchmarks.. all those
tests are definitely on a level (many levels ;) above what I use
computers for.
: You never know, sometimes those are the fastest achievable
: results. It's happened before.
Just curious, what other CPU has exhibited strange results SPEC results like
this when clocked at a higher rate?
: --Andre
: --
: PGP public key available
Francis
Bohuslav Rychlik
> line of CPUs that can run NT. Assuming the contenders are PPC and PA, they
> are both backed by big companies and industry consortiums. Given that
> the machine architecture at this level has as much to do with benchmarked,
> real and perceived performance (three different things) as the cpu does,
> the difference between, say 50 and 70 SPECints is irrelevant. When you get
Not that this is related to the Amiga, but business (and even personal)
buying decisions are to some degree based on these SPECmarks so it is
relevant. Of course, I think the most relevant thing right now will be
the OS.
Excerpts from netnews.comp.sys.amiga.advocacy: 23-Jan-94 Re: HP PA-RISC
based Amiga by David Meiklejohn@qdpii.i
>Sure, only one choice is "best", but that doesn't make the other options
>far off it.
Unfortunately, I think the PA-RISC has very little chance of becoming
the mainstream CPU in the next decade whereas the PowerPC has a lot of
chance, so I do think that the PA-RISC is far off it. Not performance
wise, no, but in terms of mainstream mass-market CPU choice, yes. The
PowerPC architecture is positioned by Motorola/IBM/Apple to become the
x86 of the next decade or two and that counts. In theory, it counts a
lot less than in the past with all these platform-independent operating
systems, but I think in practice, it will still matter whether your
local computer store has shrink-wrapped apps compiled for YOUR CPU or
not. With PowerPC, definitely. With PA-RISC, unlikely.
Thomas Holland
(br...@andrew.cmu.edu) wrote:
>
>Excerpts from netnews.comp.sys.amiga.advocacy: 23-Jan-94 Re: HP PA-RISC
>based Amiga by David Meiklejohn@qdpii.i
>>Sure, only one choice is "best", but that doesn't make the other options
>>far off it.
>
>Unfortunately, I think the PA-RISC has very little chance of becoming
>the mainstream CPU in the next decade whereas the PowerPC has a lot of
>chance, so I do think that the PA-RISC is far off it. Not performance
>wise, no, but in terms of mainstream mass-market CPU choice, yes. The
I basically agree with you on the PA-RISC.
>PowerPC architecture is positioned by Motorola/IBM/Apple to become the
>x86 of the next decade or two and that counts.
But what happens if those three companys decide to split up next week?
When copyrights, licenses, internal knowledge, microcode, actual
masks, the name etc. are tied up in the courtroom?
The PowerPC will be stuck and quickly fade away.
What happened to TrueImage or ACE?
Actually I can't think of any project that came out of a cooperation
of big companys that is still supported by all its founders.
Somehow every cooperation seems to fall apart either before or shortly
after releasing the Product. And the costumer can be lucky if one
Company continues the development (like OS/2).
I won't say that it will happen to PowerPC, but it can happen!
>In theory, it counts a
>lot less than in the past with all these platform-independent operating
>systems, but I think in practice, it will still matter whether your
>local computer store has shrink-wrapped apps compiled for YOUR CPU or
>not. With PowerPC, definitely. With PA-RISC, unlikely.
>
>
>-------------------------------------------------------------
>Bohuslav Rychlik (Bob), ryc...@cmu.edu, br...@andrew.cmu.edu
>-------------------------------------------------------------
-Thomas