IBM Power-PC future
Issa El-Hazin
Hi all,
About a week ago, I went to an IBM show flanting their future PPC's
and I was moved enough to sit down and write this. Moved not out of being
impressed, but rather due to the lack of it. Here's my $0.02 worth of what
I got out of the show.
I actually have been hoping the best for the IBM PPC's for the
last two year or so; enough to hang an IBM Power PC printout in my
place of work. Now, this printout is going down and will probably be
replaced with an Intel/HP pro printout. I think we are about to see
another 'PS/2' / 'Micro Channel' deal from IBM - that's if we/they
are lucky.
Without any further delay, here are few things that I was left
wondering about as I left the show:
* Hype/speed and bragging rights.
The 601 chips will be used instead of 604s initially in the
IBM version of the Power PC's (for short, I'll call them PPS's.)
At the time IBM PPS's will finally ship, Apple will be selling
604 based PowerMacs (PM's/PPC's). I think this is going to be
embarrassing for IBM. Imagine, after all of the delay, IBM
PPS's will be about half as fast as Apple's PPC's which will be
offered at the same time. IBM still insists on selling 601 based
computers initially, and there isn't much news when they'll have
604 based ones. (in another prospective, IBM's future PPS's will
be _about_ as fast as todays Pentiums.)
* Natvie applications.
The IBM PPS's will have no, or next to no, native applications
when they come out around Q4/94 or Q1/95. (OS/2, didn't make it
big even though that it's superior to Windows in many ways mainly
because it lacked native software.) Another no no for IBM. How many
people will buy a [fairly expensive] new computer to run things slowly
in emulation - even if that? Many users will find that DX4-100/P5-60
PCes with a sound card and a CD-ROM to be a much better value and
very fast.
* OS/2 and NT.
Microsoft's Windows NT will probably become the OS of choice
for the IBM PPS's and the Intel PCI local-bus will be the only
bus offered with IBM's new machines. Now wasn't a main idea with
the new architecture is to compete/get ride of the MS/Intel
dominance so IBM/Apple can start making a good buck again! OS/2
for the PowerPC (previously known as WorkPlace OS) keeps on getting
delayed and when it's finally released, I don't think it will compete
with Windows NT 3.5. Beside being a very robust OS, NT is also
available for Intel, MIPS, DECs, and other workstations and its'
been out for a while. Many users and software developers value
this. Software is a big factor on a platform. If the user is given
a choice between MS Word/Excell for NT or OS/2 for the PowerPC with
no software, guess which most are going to pick?
* Video sub-system.
What is the thing with IBM making a big deal out of 60-100
win-marks that their future machines will achieve? (thanks to
the Weiteck 32bit 9100 video chip.) My 90MHz Pentium can,
today, do over 100 winmarks and the Weiteck 9000 chip have
been coming on video cards for the Intel platform for over 2
years and the 9100 chip is here too (ever heard of the Viper
and the Viper Pro video cards!) By the time the IBM PPS's make
it to market, 128bit video cards will be a common place on the
Intel platform. Another strike for IBM.
* Price.
Starting price of about $3,000 and up for the lowest-end 601
60MHz IBM PPS is not going to be very competitive considering
that Dell is selling fairly loaded 60MHz Pentiums for under $2,000
today. By the time we get to see the PPS's, 60MHz Pentium PC's
will be going for $1,500 and less.
* Speed.
IBM future 604 PPC chip is going to be about as fast as
Intel's P6 chip coming out next year. The first P6 (a 133MHz
chip) is supposed to have an Integer SPEC of 200 or higher and
the initial 604 (a 100MHz chip) has an Integer SPEC of 160.
The 604 performance is supposed to improve (up to 230 SPECints)
by the time the P6 comes out, but this is not enough to kill
the P6. The P6 (at 200MHz) will be a close match to the 620
chip (which we wont see in personal computers for a while).
When the next chip from Intel/HP comes out (around 97), it
will probably tip the advantage scale toward Intel. The biggest
selling point for the PPC based computers is that they are much
faster while being cheaper. This will not be the case especially
when the whole system's price/performance is considered and where
Intel based PCes are still, and will be, considered to be very
competitive.
* Windows 4.0.
Microsoft Chicago (Windows 4.0) and Chicago's apps. will not run
on the IBM PPS computer - not even in emulation, at least not for
some time. Microsoft, and many computer analysts, predict that
Chicago will ship around 50 million copies within the first year
of its' release (the most pessimistic figure I've read put Chicago
at over 12 million copies in the first year.) This is very significant.
Not being compatible here will deter many users/developers from
jumping on the PPC bandwagon. Many companies already invested heavily
in Intel and Windows and they wont be very thrilled to dump their
existing hardware, software, and the know-how in favor of a new
computer that lacks many of the native productivity applications,
even if this computer has an IBM tag on it. (You might ask, what
about native applications? Lotus just annonced that they will not be
porting/writting any software for the PPC until they see a bigger
market and WordPerfect stopped poring their apps to OS/2...if this
gives you any ideas.)
* Intel/HP alience.
Most analysts think Intel/HP are more than likely to come out
with their heads high against IBM/Apple/Moto. Sure, analyst are
wrong many times, nonetheless; they have a great influence on
the decisions of the new computer buyers. This sounds like a
vain reason, but don't under-estimate the power of this. Besides,
may be Intel/HP do really have something great cooking. At worst,
we should see a very capable chip from the Intel/HP alience.
So, IBM's claim about the superiority of their chips will not be
the case - at least not for long. (also consider that a possibly
slightly better chip doesn't make or break a system.) Intel PC's
will probably still have an advantage where the I/O, video, and
bus sub-systems are considered.
* Competition and commitment.
There are many reports coming out from different magazines
(example: Wall Street Journal) claiming that IBM is starting
to down-play the importance of their PPC line and that IBM is
re-emphasizing its' Intel x86 line (PS/2s, Value Points,
Ambras... .) Me think, IBM is re-thinking their strategy
about butt-heading Microsoft (the largest software company in
the world), Intel (the biggest processor maker in the world),
and now HP (the 2nd biggest computer company in the world).
IBM is in business to make money for their stock holders.
Competing against the other giants on the block for something
this big, isn't probably the simplest way to get IBM out of the
red. I think IBM is probably having second doubts. Is IBM in a
shape good enough to compete against Intel/MS/HP. After all,
Intel/MS/HP are fat pigs (so to speak). They are the ones that
have been making the billions in profit for the last few years.
* Etc. etc. etc... .
I do know that there will be followers for the IBM PPS's, but I don't
think there will be a significant number of them. Microsoft/Intel/HP have
the momentum and it will take a lot more to change this.
A positive thing: competition is [usually] good. This is not all
going to waste for us, the computer users. All of this is forcing Intel
and Microsoft (and Apple, HP, DEC among others) to better their products
and lower their prices fast. PCes will have Chicago (with true pre-emptive
multitasking and threads - something the Mac will have to do without for
at least 2 more years), Plug and Play (PnP), be Multi-Media ready, and
have enough power to do many new great things (telephony, video...) soon.
Forgive me if I sound like I am [unfairly] trying to take
something away from IBM's future PPS's. I do think that the whole
thing is nice, but I think this is a little too late (Apple might
pull this one through mostly due to its' loyal users), but I don't
know about IBM. What comes to mind right now, IBM's new CEO was
caught saying, "the last thing IBM needs right now is vision." I
think he proved to me that this is the case. IBM either has no
vision or too much, but not just right. These usually don't work.
In conclusion, I don't think the current market share of about ~80%
Intel, ~10% Apple, and ~10% others will change much in the near future.
Of course, anything can happen. My best bet right now is a wait and see
approach. May be also cheer or boo whatever team if desired. I know that
I will be cheering IBM (I always picked the under-dog - if I may call a
60+ billion dollar company an under-dog.) Until I see otherwise, I'll be
very skeptical. I believe, a niche market is where the PPS's will be
for at least few years to come. Intel/Microsoft/HP are more than ready
to fight anybody for what they perceive as theirs. They've done it
successfully many times before and probably will do it again.
All IMHO of course.
Regards,
Issa
(P.S., the IBM spoke-person made a number of mistakes during the show.
Two of which I recall at the moment, are:
* He said IBM demo'ed a _150MHz_ PowerPC based computer at some
PC Expo show not long ago. The known truth is that it was only
_120MHz_ one. (it over heated and they had to take the cover
off and spray compressed air on it to cool it down.) Intel is
the one that demo'ed a 150MHz P5 PC long before that.
* He also claimed that Intel (with the help of a small drill),
cut the wire connecting the fpu to the rest of the 486 chip to
come up with the 486sx. I'll spare you the details, but if this
was the case then why don't IBM show us one of those 486sx chips
that they are talking about - if they exist that is. (anyway, the
sx chips aren't supposed to have an fpu.)
)
--
=======================================================================
Issa El-Hazin (is...@unomaha.edu)
University of Nebraska at Omaha
Technical Specialist (C&DC)
Adam Nash
> PC Expo show not long ago. The known truth is that it was only
> _120MHz_ one. (it over heated and they had to take the cover
> off and spray compressed air on it to cool it down.) Intel is
> the one that demo'ed a 150MHz P5 PC long before that.
Just a correction -
The 120MHz PPC 601 came nowhere near to overheating, as @120MHz it runs at ~4W.
Remember this is the new geometry 601 that is only 74mm^2.
What happened was one of the specialty ASICs overheated.
-Adam
Bryan Butler
the ball on PPC. But I had to take this opportunity to vent some frustration.
Issa El-Hazin (is...@cwis.unomaha.edu) wrote:
> The IBM PPS's will have no, or next to no, native applications
> when they come out around Q4/94 or Q1/95. (OS/2, didn't make it
Pentiums don't have much native sw either (code has to be optimized to take
advantage of Pentium speed). Of course, x86 code will run faster on a Pentium
than comparable emulated code on a PPC.
> Microsoft's Windows NT will probably become the OS of choice
> for the IBM PPS's and the Intel PCI local-bus will be the only
Not my choice. I want a good Unix.
Almost everyone is migrating to PCI, so
it makes sense to support it. Although an Intel product, it hopefully will
end up like the VMEbus, designed by Motorola but widely supported
throughout the industry.
Actually I think PCI is a much better choice than their previous MicroChannel.
IBM can't compete making expansion cards for their machine. Better to use
a market standard and let the low cost outfits provide expansion for your
machine.
> delayed and when it's finally released, I don't think it will compete
> with Windows NT 3.5. Beside being a very robust OS, NT is also
You're probably right, everyone will get NT just because it's Microsoft. Ignore
the fact that it's huge, slow, and brain-damaged.
> When the next chip from Intel/HP comes out (around 97), it
> will probably tip the advantage scale toward Intel. The biggest
Assuming IBM/Motorola stand still. I doubt if they will. They already have
a 4 year head-start. Apparently, the Intel/HP chip will not be based on the
Precision Architecture (is this really true? This is my impression from the
trade rags) so they will be designing a new architecture. THe PPC was derived
from the IBM architecture.
And, when it comes out, it won't have any native software either. So, if no
one will buy a machine with native software, Intel/HP will have a hard time
selling them, right? :-)
> Microsoft Chicago (Windows 4.0) and Chicago's apps. will not run
> on the IBM PPS computer - not even in emulation, at least not for
I could care less.
> even if this computer has an IBM tag on it. (You might ask, what
Ironic, isn't it, that the reason the Intel/Mess-DOS platform is so popular
today is because when the IBM PC came out, everyone had to have one just
because it was IBM.
> * Intel/HP alience.
> Most analysts think Intel/HP are more than likely to come out
> with their heads high against IBM/Apple/Moto. Sure, analyst are
Most computer alliances fail in the long run. (was it the ACE consortium
that had Microsoft, Compaq and MIPS teaming to produce the next-generation
PC? I think that lasted about 2 weeks). The IBM/Apple/Motorola alliance is
lucky.
> the decisions of the new computer buyers. This sounds like a
> vain reason, but don't under-estimate the power of this. Besides,
You are right, unfortunately. In the computer world name means everything. It
has for a long time ('No one ever got fired for buying IBM'). Only the names
change.
> may be Intel/HP do really have something great cooking. At worst,
If Intel is involved, I doubt it.
> slightly better chip doesn't make or break a system.) Intel PC's
> will probably still have an advantage where the I/O, video, and
> bus sub-systems are considered.
Yea, ISA buses are quite an advantage.
> I do know that there will be followers for the IBM PPS's, but I don't
> think there will be a significant number of them. Microsoft/Intel/HP have
> the momentum and it will take a lot more to change this.
How can they have the momentum? They are at least 4 years away from producing
a chip, much less a computer.
> * He also claimed that Intel (with the help of a small drill),
> cut the wire connecting the fpu to the rest of the 486 chip to
> come up with the 486sx. I'll spare you the details, but if this
> was the case then why don't IBM show us one of those 486sx chips
> that they are talking about - if they exist that is. (anyway, the
> sx chips aren't supposed to have an fpu.)
Don't know about the drill or the exact technique used, but supposedly the
SX and DX dies were the same when they first came out, but the SX had the
FPU disabled somehow (the FPU was there, you just couldn't use it). THis is
hearsay, and it might not be true anymore, given the volume of SX and DX
chips.
I heard about a line of minis that DG sold back in the 70s. The high end
machine was identical to the low end machine. If you bought the low end \
machine and wanted an upgrade, they would sell you a tape with new microcode
that deleted extra wait states to make it run faster. Again, hearsay.
-------
Bryan Butler
bbu...@netcom.com
R S Rodgers
Switching sides, briefly --
> * Hype/speed and bragging rights.
> The 601 chips will be used instead of 604s initially in the
> IBM version of the Power PC's (for short, I'll call them PPS's.)
Price counts. Nothing wrong with shipping 601 systems if they price
them properly. Since Pentiums are hitting $2000 or so, "properly"
seems pretty low. But since IBM is manufacturing now, they can't use
604s -- and they don't really need to since the initial machines will
be bought mostly for evaluation purposes.
No 604's == non-issue in 1994.
[...]
> * Natvie applications.
MS Word and Excel for NT should be out by the time these systems ship.
Micrografx already has an NT version of their photo editing software.
[...]
> Many users will find that DX4-100/P5-60
> PCes with a sound card and a CD-ROM to be a much better value and
> very fast.
The PowerPS systems will certainly ship with sound (perhaps CPU
driven, unfortunately) and I can't imagine IBM will ship NT systems
without a CDROM drive.
> * OS/2 and NT.
> Microsoft's Windows NT will probably become the OS of choice
> for the IBM PPS's and the Intel PCI local-bus will be the only
> bus offered with IBM's new machines. Now wasn't a main idea with
> the new architecture is to compete/get ride of the MS/Intel
> dominance so IBM/Apple can start making a good buck again!
You're confusing the antics of Team-OS/2 and the like with IBM's real
strategy, which is to make money, and IBM makes money manufacturing
the hardware. Since, as you say, NT will at least have some software,
the no-native-software issue is only an issue if IBM force-bundles
OS/2-PPC.
Apple appears not to be going with PReP, so the new "open, friendly"
Apple isn't an issue, either. They don't seem to be having much
problem getting native software -- I was impressed by the number of
"Accellerated for Power Macintosh" stickers I saw in my last catalog
from Mac Connection.
[...]
> * Video sub-system.
> What is the thing with IBM making a big deal out of 60-100
> win-marks that their future machines will achieve? (thanks to
> the Weiteck 32bit 9100 video chip.)
Dunno. Agree, here -- the P9100 isn't nearly as impressive in
comparison to the competition as the P9000 was. The 64bit S3 chips
are faster -- the Stealth 64 is faster than the Viper Pro. Chances
are, though, that NT will have drivers for other cards -- NT/Alpha and
MIPS support Mach32 and other cards (but not nearly the number as
x86). It's not that big an issue -- the systems are PCI.
> * Price.
> Starting price of about $3,000 and up for the lowest-end 601
> 60MHz IBM PPS is not going to be very competitive considering
> that Dell is selling fairly loaded 60MHz Pentiums for under $2,000
> today. By the time we get to see the PPS's, 60MHz Pentium PC's
> will be going for $1,500 and less.
Probably. Agree.
> * Speed.
[lots of stuff about the 604 vs. the P6 -- not interested in arguing
the merits of comparing it..]
> * He also claimed that Intel (with the help of a small drill),
> cut the wire connecting the fpu to the rest of the 486 chip to
> come up with the 486sx. I'll spare you the details, but if this
> was the case then why don't IBM show us one of those 486sx chips
> that they are talking about - if they exist that is. (anyway, the
> sx chips aren't supposed to have an fpu.)
I wonder if this is the same IBM spokesman that was at Comdex last
winter claiming that OS/2-SMP scales linearly while NT doesn't (and
cited an IO-bound benchmark as evidence).
--
Previous .sig deleted because some people couldn't parse "from email
_received_ re: a post on comp.sys.powerpc."
R S Rodgers
bbu...@netcom.com (Bryan Butler) wrote:
> > The IBM PPS's will have no, or next to no, native applications
> > when they come out around Q4/94 or Q1/95. (OS/2, didn't make it
> Pentiums don't have much native sw either (code has to be optimized to take
> advantage of Pentium speed).
No, it doesn't. To get the _maximum_ benefit, it has to be compiled
with Pentium optimization, but only a fool or someone who has never
used a Pentium thinks that the Pentium offers nothing but the
clockspeed advantage over a 486 when running non-optimized software.
Try running Windows 3.1 (*386* code) on a 486dx4-100 and on a P60, and
the latter will be noticibly faster -- at a little more than half the
clockspeed.
Timothy Jehl~
In article <bbutlerC...@netcom.com>, bbu...@netcom.com (Bryan Butler) writes:
>
> And, when it comes out, it won't have any native software either. So, if no
> one will buy a machine with native software, Intel/HP will have a hard time
> selling them, right? :-)
>
code natively, so that in fact it will have a huge existing software database.
>
> > may be Intel/HP do really have something great cooking. At worst,
> If Intel is involved, I doubt it.
>
Ah. Now we see the colors flying!
TJ
--
Mark Rogowsky
> Hi all,
>
> About a week ago, I went to an IBM show flanting their future PPC's
> and I was moved enough to sit down and write this. Moved not out of being
> impressed, but rather due to the lack of it. Here's my $0.02 worth of what
> I got out of the show.
And no offense, but not a penny more.
>
> I actually have been hoping the best for the IBM PPC's for the
> last two year or so; enough to hang an IBM Power PC printout in my
> place of work. Now, this printout is going down and will probably be
> replaced with an Intel/HP pro printout. I think we are about to see
> another 'PS/2' / 'Micro Channel' deal from IBM - that's if we/they
> are lucky.
And no offense, but you have killed the horse before he even has *entered*
the starting gate.
>
> Without any further delay, here are few things that I was left
> wondering about as I left the show:
>
> * Hype/speed and bragging rights.
> The 601 chips will be used instead of 604s initially in the
> IBM version of the Power PC's (for short, I'll call them PPS's.)
> At the time IBM PPS's will finally ship, Apple will be selling
> 604 based PowerMacs (PM's/PPC's). I think this is going to be
> embarrassing for IBM.
Look for 604s right on the heels of the 601s. I'd guess: 601s in
November/December. 604s in June/July.
> * Natvie applications.
> The IBM PPS's will have no, or next to no, native applications
> when they come out around Q4/94 or Q1/95.
OS/2 hasn't not made it (I deleted that). It's already sold 3-5 million
copies -- hard to know which number is real -- and with OS/2 Lite (the 4MB
version) and OS/2 for Windows, things are picking up -- even without
native apps. Don't you think IBM will differentiate this systems on more
than just price to counter the lack of native apps? Don't you think
evangelizing NT developers to recompile for PPC is high on the list to.
>
> * OS/2 and NT.
> Microsoft's Windows NT will probably become the OS of choice
> for the IBM PPS's and the Intel PCI local-bus will be the only
> bus offered with IBM's new machines. Now wasn't a main idea with
> the new architecture is to compete/get ride of the MS/Intel
> dominance so IBM/Apple can start making a good buck again! OS/2
> for the PowerPC (previously known as WorkPlace OS) keeps on getting
> delayed and when it's finally released, I don't think it will compete
> with Windows NT 3.5. Beside being a very robust OS, NT is also
> available for Intel, MIPS, DECs, and other workstations and its'
> been out for a while.
OS/2 is availble for Intel already and will be available for PPC. NT is
available for Intel and will be available for PPC. This is two sides of
the same coin. Also, you know nothing about PCI apparently. It was
invented by Intel and is now an open standard. You can build PCI
buses/chipsets without paying Intel a nickel.
> * Video sub-system.
> What is the thing with IBM making a big deal out of 60-100
> win-marks that their future machines will achieve? (thanks to
> the Weiteck 32bit 9100 video chip.) My 90MHz Pentium can,
> today, do over 100 winmarks and the Weiteck 9000 chip have
> been coming on video cards for the Intel platform for over 2
> years and the 9100 chip is here too (ever heard of the Viper
> and the Viper Pro video cards!) By the time the IBM PPS's make
> it to market, 128bit video cards will be a common place on the
> Intel platform. Another strike for IBM.
Your 100 million WinMarks is trash stats. PC Magazine tells us that all
those overblown numbers are just that. Apparently, a more accurate figure
for high-end video cards and high-end machines is 40 million WinMarks (or
some such).
>
> * Price.
> Starting price of about $3,000 and up for the lowest-end 601
> 60MHz IBM PPS is not going to be very competitive considering
> that Dell is selling fairly loaded 60MHz Pentiums for under $2,000
> today. By the time we get to see the PPS's, 60MHz Pentium PC's
> will be going for $1,500 and less.
And IBM will merely try to get people to pay twice as much for
incompatible hardware. Somehow, I think a little more is planned.
>
> * Speed.
> IBM future 604 PPC chip is going to be about as fast as
> Intel's P6 chip coming out next year. The first P6 (a 133MHz
> chip) is supposed to have an Integer SPEC of 200 or higher and
> the initial 604 (a 100MHz chip) has an Integer SPEC of 160.
Your chip-mania is lunacy... Let's try a new lens...
Q2,'95, PPC 604 in machines, chip costs around $400 at 100MHz.
Q4,'95 P6 in machines, chip costs around $1100 at 133MHz.
PPC 604 matches P6 performance (or betters it) with
133MHz and 150MHz versions. 100MHz version is $250.
Developers routinely recompiled Win32 apps for PPC.
PPC 620 shipping in quantity. Initial price, $999.
AMD K5 variants and Cyrix M1 variants begin really annoying
Intel by matching all P5 performance points with lower
prices. The ensuing price war begins chopping away something
from Intel's gargantuan profits.
Q1,'96 P6 machines now available in quantity. Few willing to pay
the high price. Really fast P5s keep those not looking for
change quite happy. Some, looking for price/performance,
begin thinking about PPC machines.
Q2,'96 PPC620 machines ship. Faster 604s, P5s, P6s, abound. Nothing
can touch 620 in the PC marketplace. Machines expensive.
PPC604 chip price now at about $150. P6 at about $750.
Q4,'96 Word of the PowerPC 800 series just swept Comdex (IBM and Moto
did a big show on the new series). Systems should begin appearing
in 12-18 months.
Intel cloners becoming really annoying. IBM, making money selling
PPC systems, also has figured out what Intel already knows: you
can make more money selling whole logic boards to PC cloners rather
than just chips. Using Cyrix technology, they are cutting deeper into
the P5-class x86 business.
Intel/HP briefing ignored. Who cares about a chip coming out in two
years?
Gateway and Dell merge but keep identities separate.
HP again considers buying Apple.
Q2,'97 Intel fights back with much cheaper P6s and much faster ones.
First PPC 800 series silicon is becoming available.
Q4,'97 PPC running 800 SPECint.
Q2,'98 Intel/HP first silicon using VLIW technology. Compatible with
existing x86 binaries. Intel encouraging ISVs to write to the
"native mode" of the new chips, though.
That was fun....
> * Windows 4.0.
> Microsoft Chicago (Windows 4.0) and Chicago's apps. will not run
> on the IBM PPS computer - not even in emulation, at least not for
> some time. Microsoft, and many computer analysts, predict that
> Chicago will ship around 50 million copies within the first year
> of its' release (the most pessimistic figure I've read put Chicago
> at over 12 million copies in the first year.)
You don't need to run Chicago, just Win32 apps. IBM is working really hard
on this. Don't count them out. Also, Chicago is really a half-step for
Microsoft. The eventual goal is to move the whole Chicago-base to Cairo
(or some other NT follow-on). Then the MS OS will be platform independent.
>
> * Intel/HP alience.
See above. My vapor is bigger than your vapor.
> * Competition and commitment.
> There are many reports coming out from different magazines
> (example: Wall Street Journal) claiming that IBM is starting
> to down-play the importance of their PPC line and that IBM is
> re-emphasizing its' Intel x86 line (PS/2s, Value Points,
> Ambras... .)
The Journal doesn't grasp the computer industry at all.
> Me think, IBM is re-thinking their strategy
> about butt-heading Microsoft (the largest software company in
> the world), Intel (the biggest processor maker in the world),
> and now HP (the 2nd biggest computer company in the world).
> IBM is in business to make money for their stock holders.
> Competing against the other giants on the block for something
> this big, isn't probably the simplest way to get IBM out of the
> red. I think IBM is probably having second doubts. Is IBM in a
> shape good enough to compete against Intel/MS/HP. After all,
> Intel/MS/HP are fat pigs (so to speak). They are the ones that
> have been making the billions in profit for the last few years.
The IBM PC Company loses money. IBM has nothing to lose butt-heading
Microsoft, Intel and HP. If they change the paradigm in personal
computing, they'll win back revenue. If they don't, they'll be no worse
off then they are now. You are simply mistaken about the strategy.
> I do know that there will be followers for the IBM PPS's, but I don't
> think there will be a significant number of them. Microsoft/Intel/HP have
> the momentum and it will take a lot more to change this.
One thing history teaches us is that no matter how dominant the strong
seem today, tomorrow they will be either (a) less strong or (b) very weak.
I can't see how now but somehow Intel and Microsoft will not be ruling the
roost like they do now in a decade. HP, contrary to your opinion, is not a
particularly important player in any of this: they don't have a PC OS,
they are like No. 6 in PC market share behind "giants" like Gateway 2000,
and they don't sell microprocessors to much of anyone. HP dominates the
laser-printer business, to be sure, but this is another matter.
>
> A positive thing: competition is [usually] good. This is not all
> going to waste for us, the computer users. All of this is forcing Intel
> and Microsoft (and Apple, HP, DEC among others) to better their products
> and lower their prices fast. PCes will have Chicago (with true pre-emptive
> multitasking and threads - something the Mac will have to do without for
> at least 2 more years), Plug and Play (PnP), be Multi-Media ready, and
> have enough power to do many new great things (telephony, video...) soon.
IBM's PPCs will have preemption, threads, telephony, video, etc. (as soon
as they ship). Macs will have threads, telephony, video, etc. (as soon as
7.5 ships). I don't think lack of preemption will kill, or even severely
wound, Apple's efforts to keep -- and perhaps increase -- its market
share. Also, that PnP stuff and multimedia will still be better on Macs
(because they've always been plug and play and because QuickTime is really
going to win the race over Video for Windows -- call Bell Atlantic if you
disagree).
>
> ... What comes to mind right now, IBM's new CEO was
> caught saying, "the last thing IBM needs right now is vision." I
> think he proved to me that this is the case. IBM either has no
> vision or too much, but not just right. These usually don't work.
Gerstner backed off that vision comment and it was taken out of context.
Read the papers from this year for a change.
>
> In conclusion, I don't think the current market share of about ~80%
> Intel, ~10% Apple, and ~10% others will change much in the near future.
See above. The future never becomes what it seems in the present. Just 3
years ago, Windows wasn't really that important and the 486 wasn't either.
Even industry gurus, like Richard Schaeffer (sp?), were stunned by the
changes as in instant retrospective -- check Computer Letter from the end
of 1992 (I think that's the year)
> for at least few years to come. Intel/Microsoft/HP are more than ready
> to fight anybody for what they perceive as theirs. They've done it
> successfully many times before and probably will do it again.
Again, you place HP in some location they are not. They are making money,
but not because they control a key piece of the PC pie like Intel and MS.
> * He also claimed that Intel (with the help of a small drill),
> cut the wire connecting the fpu to the rest of the 486 chip to
> come up with the 486sx. I'll spare you the details, but if this
> was the case then why don't IBM show us one of those 486sx chips
> that they are talking about - if they exist that is. (anyway, the
> sx chips aren't supposed to have an fpu.)
The original 486SX chips were indeed 486DX chips with the FPU disabled.
Over time, this did chain. True, no drills were involved.
Mark
ghost
: OS/2 is availble for Intel already and will be available for PPC. NT is
: available for Intel and will be available for PPC. This is two sides of
: the same coin. Also, you know nothing about PCI apparently. It was
: invented by Intel and is now an open standard. You can build PCI
: buses/chipsets without paying Intel a nickel.
the first beta just shipped out, and the final release is expected late
Q4/94 or Q1/95. IBM's PPC machines are due out in October (or so I hope).
That means the OS that bundles with these machines will definetely be NT.
However IBM's reasoning that people who buy their machines that early are
going to be OS/2 enthusiasts as well and will switch over as soon as OS/2
comes out is pretty sound. But the fact remains that NT will be shipping
with IBM PPC machines, and if IBM is wrong in their assumption...
: Your 100 million WinMarks is trash stats. PC Magazine tells us that all
: those overblown numbers are just that. Apparently, a more accurate figure
: for high-end video cards and high-end machines is 40 million WinMarks (or
: some such).
For those who don't know, these 100 or other extremely high WinMarks are
tested with WinBench 3.0 which has long since been outdated. Version 4.0
is the version to use, as PC Mag keeps reminding us, and the top cards
score from 38-42 on this test. Had to add that to your statement, Mark.
--
gh...@dorsai.dorsai.org
Hussein Kanji
--
Ú--- ÄÄÄ ÄÄ ú ú Ä-Ä¿
| ÄÄðGh™äØðÄÄ ú Life's a big joke ú Internet: gh...@dorsai.dorsai.org ú
: aka. ù so don't worry ù :
ú PSyCHoTiC NeURoSiS ú about it so much. ú [718]GET.REAL |
R S Rodgers
> Yes, but OS/2 for the PPC isn't due out for a while. From what I've heard
> the first beta just shipped out,
Where did you hear that? (Serious question.)
> and the final release is expected late
> Q4/94 or Q1/95. IBM's PPC machines are due out in October (or so I hope).
> That means the OS that bundles with these machines will definetely be NT.
IBM is sure to offer at least AIX and NT, but Solaris may be available
as well.
> However IBM's reasoning that people who buy their machines that early are
> going to be OS/2 enthusiasts as well and will switch over as soon as OS/2
> comes out is pretty sound.
Although I agree (and have stated as much) that the early adopters
will be mostly willing to switch OSes, the above isn't a good
assumption. Why would OS/2 enthusiasts buy an expensive PPC machine
running NT when they can buy a cheaper, faster Intel machine and run
OS/2 (which compounds the speed by being faster for many things than
NT)? Doesn't seem like sound reasoning to me.
[...]
> For those who don't know, these 100 or other extremely high WinMarks are
> tested with WinBench 3.0 which has long since been outdated. Version 4.0
> is the version to use, as PC Mag keeps reminding us, and the top cards
> score from 38-42 on this test. Had to add that to your statement, Mark.
I've personally tested the Diamond Stealth 64 4MB at more than 42 on a
dx4-100. Since the original poster had a Viper (2/3rds the speed of
the S64 at best, 1/2 average) and a Pentium-90, 100 is a little hard
to believe, but 70-75 wouldn't be, and 80+ for an s64 would be
believable.
ghost
: > the first beta just shipped out,
: Where did you hear that? (Serious question.)
There were a couple of posts on comp.os.os2.advocacy a few days ago. And
someone mentioned that he was beta testing a copy either on this
conference or os2.advocacy.
: IBM is sure to offer at least AIX and NT, but Solaris may be available
: as well.
Well, AIX goes without saying. I seriously doubt Solaris will be able to
put their project together by October. Maybe by the time OS/2 comes out.
: Although I agree (and have stated as much) that the early adopters
: will be mostly willing to switch OSes, the above isn't a good
: assumption. Why would OS/2 enthusiasts buy an expensive PPC machine
: running NT when they can buy a cheaper, faster Intel machine and run
: OS/2 (which compounds the speed by being faster for many things than
: NT)? Doesn't seem like sound reasoning to me.
Probably for the same reason I would, the future. Workplace should be out
a couple of months after the release of the PPC machine and should run
faster than Os/2 on Intel Pentiums. Plus (another rumor here from posts
mentioned above) it should run DOS/Windows at 486/66 speeds, and hopefully
Mac System 7 software as well. And $3000 which I hear will be the price
of these PPC machines isn't as bad as it seems. A little on the expensive
side, but not overly so.
: I've personally tested the Diamond Stealth 64 4MB at more than 42 on a
: dx4-100. Since the original poster had a Viper (2/3rds the speed of
: the S64 at best, 1/2 average) and a Pentium-90, 100 is a little hard
: to believe, but 70-75 wouldn't be, and 80+ for an s64 would be
: believable.
Right the Stealth 64 is a good buy based on the S3 chip, (perfect for
OS/2) as is the ATI mach64. Both score around 40 WinMarks on 4.0
--
Ú--- ÄÄÄ ÄÄ ú ú Ä-Ä¿
| Gh™st ú Life's a big joke ú Internet: gh...@dorsai.dorsai.org ú
: Hussein Kanji ù so don't worry. ù [718]GET.REAL :
Ã-ÄÄ ú ú ÄÄ ÄÄÄ --Ä´
ghost
: received people don't seem to realize are automobiles and not computers --
: I made a joke about beta testing OS/2 for the PowerPC in a Mustang (or was
: it a Winstar?). It should've been obvious to everyone reading that is was a
: joke. The thread was about cars, for goodness sake.
No that wasn't it, Mark. I saw that post, and believe me that wasn't the
post that said that beta Workplace's were being released. I got your joke.
Tony Hamilton
> the ball on PPC. But I had to take this opportunity to vent some frustration.
>
> Issa El-Hazin (is...@cwis.unomaha.edu) wrote:
>
> > The IBM PPS's will have no, or next to no, native applications
> > when they come out around Q4/94 or Q1/95. (OS/2, didn't make it
> Pentiums don't have much native sw either (code has to be optimized to take
> advantage of Pentium speed). Of course, x86 code will run faster on a Pentium
> than comparable emulated code on a PPC.
Just to kill this one before it gets started again - ALL x86 CODE IS NATIVE
ON THE PENTIUM PROCESSOR, and NO x86 CODE MUST BE OPTIMIZED TO TAKE ADVANTAGE OF
PENTIUM PERFORMANCE. The x86 instruction set is the Pentium processor's
_native_ instruction set, and this processor, with _extremely rare_ exceptions,
runs all code faster than all 486 processors at equivalent clock rates.
This, in contrast to PowerPC chips, for which 680x0 code is _not_ native,
must be emulated, and in many (most?) cases, runs SLOWER than 68040 chips at
equivalent clock rates (and even lower clock rates).
The fact that x86 _can_ be even further optimized for the Pentium processor
is completely irrelevant. It is not required to achieve significant performance
increases.
Sheesh.
--
TONY HAMILTON ==================================================
Intel Corporation "Have compiler - will travel"
Tony_H_...@ccm.fm.intel.com -or- tham...@pcocd2.intel.com
========================(I speak only for myself ... not Intel)=
Amancio Hasty Jr
>In article <bbutlerC...@netcom.com>, bbu...@netcom.com (Bryan Butler) writes:
>> I agree with a lot of what you have to say, and I agree IBM is dropping
>> the ball on PPC. But I had to take this opportunity to vent some frustration.
>>
>> Issa El-Hazin (is...@cwis.unomaha.edu) wrote:
>>
>> > The IBM PPS's will have no, or next to no, native applications
>> > when they come out around Q4/94 or Q1/95. (OS/2, didn't make it
>> Pentiums don't have much native sw either (code has to be optimized to take
>> advantage of Pentium speed). Of course, x86 code will run faster on a Pentium
>> than comparable emulated code on a PPC.
>
>Just to kill this one before it gets started again - ALL x86 CODE IS NATIVE
>ON THE PENTIUM PROCESSOR, and NO x86 CODE MUST BE OPTIMIZED TO TAKE ADVANTAGE OF
>PENTIUM PERFORMANCE. The x86 instruction set is the Pentium processor's
Hmmm... I am on your side however I finding out that intensive floating
point apps such as povray must be compiled with pentium optimizations
in order to take advantage of the Pentium and yes to perform faster
than a 486DX266.
Further inquiries on the raytracer, povray, to comp.graphics.raytracing.
Amancio
--
FREE unix, gcc, tcp/ip, X, open-look, netaudio, tcl/tk, MIME, midi,sound
at freebsd.cdrom.com:/pub/FreeBSD
Amancio Hasty, Consultant |
Home: (415) 495-3046 |
e-mail ha...@netcom.com | ftp-site depository of all my work:
| sunvis.rtpnc.epa.gov:/pub/386bsd/X
Salvatore Denaro
>: > Yes, but OS/2 for the PPC isn't due out for a while. From what I've heard
>: > the first beta just shipped out,
I heard that one of the main reasons the PPS are not out yet is that
the OS/2 port is *Way* behind schedual...
>
>: IBM is sure to offer at least AIX and NT, but Solaris may be available
>: as well.
>
>Well, AIX goes without saying. I seriously doubt Solaris will be able to
>put their project together by October. Maybe by the time OS/2 comes out.
I saw Solaris running on a PPC prototype at PC Expo, Sun says that it
will be ready "Six months" after the PPS is ready.
I am looking forward to a cheap, fast Sun workstation...>
--
s...@panix.com Yes, I use PGP.
Salvatore Denaro
Live fast, Die young, Hack C++ My heart is broke/but I have some glue
Sex, Drugs and Cryptography. Help me inhale/and mend it with you
Bryan Butler
> In article <bbutlerC...@netcom.com>,
> bbu...@netcom.com (Bryan Butler) wrote:
> > > The IBM PPS's will have no, or next to no, native applications
> > > when they come out around Q4/94 or Q1/95. (OS/2, didn't make it
> > Pentiums don't have much native sw either (code has to be optimized to take
> > advantage of Pentium speed).
> No, it doesn't. To get the _maximum_ benefit, it has to be compiled
> with Pentium optimization, but only a fool or someone who has never
> used a Pentium thinks that the Pentium offers nothing but the
> clockspeed advantage over a 486 when running non-optimized software.
> Try running Windows 3.1 (*386* code) on a 486dx4-100 and on a P60, and
> the latter will be noticibly faster -- at a little more than half the
> clockspeed.
Yes, I knew this, I just wasn't precise enough. Sorry.
My point, which I didn't state too well, is that the lack of native apps is
not necessarily a barrier to sales. If I buy a Pentium machine without
Pentium optimized applications, there is some wasted performance. Hopefully,
I will eventually get optimized apps to take advantage of that performance
in the future. So, buying the hardware now is investing in future performance.
--
-------
Bryan Butler
bbu...@netcom.com
Matt Kennel
: I saw Solaris running on a PPC prototype at PC Expo, Sun says that it
: will be ready "Six months" after the PPS is ready.
: I am looking forward to a cheap, fast Sun workstation...>
Folks ought to remember that by this time it's likely that Solaris will
include OpenStep, which is the upper layers of the 'next' version of
NextStep. So you might be looking forward to a cheap fast Sun
that works like a Next and is still "standard Unix" if you need it to be.
sounds nice to me.
: Salvatore Denaro
--
-Matt Kennel m...@inls1.ucsd.edu
-Institute for Nonlinear Science, University of California, San Diego
-*** AD: Archive for nonlinear dynamics papers & programs: FTP to
-*** lyapunov.ucsd.edu, username "anonymous".
Dennis O'Connor
bbu...@netcom.com (Bryan Butler) writes:
] My point, which I didn't state too well, is that the lack of native apps is
] not necessarily a barrier to sales.
Using the Intel486(tm) and Pentium(tm) processors is a very poor choice
for an exampleif this is what you want to show. The two processors
execute the same native instruction sets, with only a few system-level
rarely-used instructions added to the Pentium processor. "486" apps
_are_ native apps for the Pentium processor : just not at the highest
possible level of optimization.
Different processors with the same native instruction set can still
have different optimizations : if (in a made up example) one processor
does shifts in 1 cycles and adds in 2 (silly, but let it go) and the
other does shifts in 2 cycles and adds in 1, then when I want to
multiply a C unsigned integer by two (C doesn't trap on overflow, BTW)
I'll use "shift val 1 bit left" on one processor, and "add val to itself"
on the other.
] If I buy a Pentium machine without Pentium optimized applications,
] there is some wasted performance.
If you buy _any_ machine with _any_ processor, and (for example)
put less than say a 1 Gigabyte Level-2 0ns Flash Cache and
a 1ms Average Latency Hard Drive, and a Speed of Light 256-bit
Video Processor, you won't get 100% of the CPU's performance.
Heck, you're probably losing more performance to "paranoid" error checking
(in well-coded apps) and poor algforityym choices (in less-well-coded apps).
--
Dennis O'Connor doco...@sedona.intel.com
Intel i960(R) Microprocessor Division Solely responsible for what I do.
Nevin Liber
Tony Hamilton <tham...@fnugget.intel.com> wrote:
>This, in contrast to PowerPC chips, for which 680x0 code is _not_ native,
>must be emulated, and in many (most?) cases, runs SLOWER than 68040 chips at
>equivalent clock rates (and even lower clock rates).
Yea, but who cares? I don't know anyone running a 601 at a speed as
slow as 42MHz, which is the speed I am running my Q800 at. As some
Intel VP (I can't remember who) said in MacWeek, its the total system
that counts; you can't just look at the numbers.
>The fact that x86 _can_ be even further optimized for the Pentium processor
>is completely irrelevant. It is not required to achieve significant performance
>increases.
In other words, even recompiling an app with Pentium-specific
optimizations (or whatever Dennis O'Connor's really long phrase about
instruction times describing this is, since he doesn't like the term
"native"), the performance will not significantly increase.
Just playing devil's advocate,
--
Nevin ":-)" Liber ne...@cs.arizona.edu (602) 293-2799
+++ (520) after 3/95
summer office: (602) 621-8112
Only 11 more shopping days 'til my birthday (August 12th)!!
Patrick Doyle
>
>Hi all,
>
> About a week ago, I went to an IBM show flanting their future PPC's
>and I was moved enough to sit down and write this. Moved not out of being
>impressed, but rather due to the lack of it. Here's my $0.02 worth of what
>I got out of the show.
>
I too went to an IBM show this past week. It was the most
disappointing two hours I've spent in a long time. I spent the time
listening to an IBM sales rep. tell us all how great the new Power
Personal systems are going to be. He had no hardware to show us, but
instead, showed slides with the details of the three systems that are
going to be released. He admitted several times that the slides were
inaccurate and that one feature or another had changed already.
I asked him about compatibility with existing hardware and
software and his response was that each of the various OS's that will
run on the PPS (OS/2, Windows NT, Solaris, AIX, ... I forget the other
one or two) will be capable of running DOS or Windows in emulation,
but native device drivers will be required for devices. Fair
enough..., but I wonder if this strategy is going to work. Consider
two scenarios...
1) I'm a home user. I've spent $3000 over the years on my PC and now
have it configured just about the way I want it. I have a 14.4 kbps
modem, a Sound Blaster, a CD-ROM, and ~350 MB of disk space on two
disks. Of course, since I bought the CPU first, it is only a 33 MHz
'386. I would like to upgrade, but I don't want to throw away too
much of my existing equipment. I could buy a PPS, but they will all
be shipping with built in audio and CD-ROM, so I have to throw out my
SB and CD-ROM (or try to sell, which is just as annoying). I don't
know if the PPS uses SCSI or IDE for its disk technology, but it
better be IDE, or I'll have to throw out the disk as well. I
understand that using these periperals only under emulation will be
slow, but I can accept that (after all, I'm used to a 33 MHz '386 :-)
but I expect, if the PPS revolution catches on, native device drivers
and other disk controllers will be made available. If it doesn't
catch on, well at least I'm no worse off than I was before.
2) I'm a business (engineering) user. I use ISA based cards for
emulators, IEEE-488 controllers, embedded processors, etc... I would
like to use the PPC since I believe it represents a better technology
than the x86 architecture, but right now, I will still have to buy x86
architecture in order to continue my product development. Since I can
but a 90 MHz Pentium for about the same price as a PPS, which do you
think I'm going to buy?
I tried explaining this to the sale's rep., but he just took
down the name of one of the emulator's that I use and promised to
check to see if TI would be providing a native port of the product.
When I tried to press him, he said "If all you want to do is to run
existing applications, you're better off buying a Pentium. If you
want to run applications of the future, you need a PowerPC". My
problem is, I need to do both.
I believe that the PowerPC technology will push Intel to
increase their floating point performance and I wouldn't be too
surprised if, sometime in the near future, Intel developes a new
architecture that is capable of running x86 legacy code, but can
switch modes into a more RISC-like mode in order to provide an upgrade
path. (Hmm.. remember the revolutionary 80386 processor that had a
bit in the flags register that enabled it to run as a "virtual 8086"?
I wonder how long it will be until there's a flag which will enable a
task to run as a "virtual 80[3,4,5]86"?) The truth is, the Pentium is
the second most backwards compatible system in the world (the
telephone is the first IMHO), and there are compelling reasons for
maintaining that compatibility, while providing an upgrade path from
1970's design criteria to 1990's.
The 80386 represented a significant advance over its
predecessors in terms of capabilities (32 bit registers, useful MMU,
etc..). Now we have software that can only run on a '386 or higher
processor. I expect to see a similar shift in the future. I
personally don't know if it will be to a PowerPC architecture or to an
(improved) Intel architecture, but I don't expect it to happen without
backards compatibilty.
Note to IBM: Any more news on the 615 chip?
R S Rodgers
ro...@forsythe.stanford.edu (Mark Rogowsky) wrote:
> In article <gfjEkO9R...@wam.umd.edu>, rsro...@wam.umd.edu (R S
> Rodgers) wrote:
> > I've personally tested the Diamond Stealth 64 4MB at more than 42 on a
> > dx4-100. Since the original poster had a Viper (2/3rds the speed of
> > the S64 at best, 1/2 average) and a Pentium-90, 100 is a little hard
> > to believe, but 70-75 wouldn't be, and 80+ for an s64 would be
> > believable.
>
> according to PC Magazine?
I have no idea.
R S Rodgers
[me; why would OS/2 users buy a slower, more expensive PPC to run a
larger OS with more overhead instead of buying x86 and running the
much slimmer OS/2?]
> Probably for the same reason I would, the future. Workplace should be out
> a couple of months after the release of the PPC machine and should run
> faster than Os/2 on Intel Pentiums.
What makes you think so? Not only is OS/2-PPC larger and more complex
(show me a fast Mach3.0-based system) than OS/2 x86, it also has to
run all the current x86 software under emulation.
> Plus (another rumor here from posts
> mentioned above) it should run DOS/Windows at 486/66 speeds,
Right. And SoftWindows is as fast as a 486sx-25. Even if they do
make dx2-66 speeds, that's about 1/3rd as fast as a comparably priced
P90 -- not even counting the additional overhead, in terms of CPU,
disk and memory -- that OS/2-PPC is very likely to incur. OS/2-PPC
isn't the slim, gung-ho OS/2 that you find on x86, it's PM atop Mach 3.0.
> and hopefully
> Mac System 7 software as well.
Can you show me a recent quote from an IBM spokesman that makes this
claim? I haven't seen one for almost a year. The Byte article on
WPOS made it clear that IBM plans to have Window/DOS support and a PM
personality out at ship, no AIX, no Mac, no Win32.
> And $3000 which I hear will be the price
> of these PPC machines isn't as bad as it seems. A little on the expensive
> side, but not overly so.
It's more than a 90MHz Pentium. Why would an OS/2 user buy dx2-66
performance for 90MHz Pentium prices -- especially if they'd end up
running NT for anywhere from 6 to 8 months?
Kenneth Fair
[stuff deleted]
> A positive thing: competition is [usually] good. This is not all
> going to waste for us, the computer users. All of this is forcing Intel
> and Microsoft (and Apple, HP, DEC among others) to better their products
> and lower their prices fast. PCes will have Chicago (with true pre-emptive
> multitasking and threads - something the Mac will have to do without for
> at least 2 more years), Plug and Play (PnP), be Multi-Media ready, and
> have enough power to do many new great things (telephony, video...) soon.
I must point out that threading is already possible for the Macintosh with
the Threads Manager Extension which can be obtained from ftp.apple.com.
I don't use it, so I don't know how robust it is, but it is available now.
Furthermore, I think it could be argued that Apple is ahead of
Intel/Microsoft when it comes to multimedia. "Multimedia" as a buzzword
really became prevalent with HyperCard years ago. Not to mention the
QuickTime standard which has been out for quite some time now. I'm
certainly not saying that Apple has had a monopoly on the innovations,
but I think that they have been thinking for a longer period of time
about this sort of integration.
--
KENNETH FAIR - Rice University # ken...@owlnet.rice.edu
Power Macintosh!
Nathan Boyd
Mark Rogowsky <ro...@forsythe.stanford.edu> wrote:
>In article <issa.775452041@cwis>, is...@cwis.unomaha.edu (Issa El-Hazin) wrote:
>
>OS/2 is availble for Intel already and will be available for PPC. NT is
>available for Intel and will be available for PPC. This is two sides of
>the same coin. Also, you know nothing about PCI apparently. It was
>invented by Intel and is now an open standard. You can build PCI
>buses/chipsets without paying Intel a nickel.
OS/2 and NT are certainly not "the same coin":
* NT is now a mature (or at least maturing) product, while WOS
isn't even in Beta
* NT/PPC is simply another port of NT, which is already on Intel,
Alpha, MIPS, etc.
* IBM is going to bundle NT with its PPS machines until WOS becomes
available; how can this not hurt OS/2's future?
>> * Video sub-system.
>> What is the thing with IBM making a big deal out of 60-100
>> win-marks that their future machines will achieve?
>>
>> By the time the IBM PPS's make
>> it to market, 128bit video cards will be a common place on the
>> Intel platform. Another strike for IBM.
>
>Your 100 million WinMarks is trash stats. PC Magazine tells us that all
>those overblown numbers are just that. Apparently, a more accurate figure
>for high-end video cards and high-end machines is 40 million WinMarks (or
>some such).
The issue here is probably which version of WinBench is being used. 3.11
is responsible for those 100M+ scores, while the best cards generally get
30-40M WinMarks with 4.0.
As for the original poster's point about IBM striking out; this is
conceivably a mistake for IBM to build the video subsystem onto the MB.
Apple is moving away from this with its next generation of PMacs, and I
think that most buyers prefer the choice of plugging in a PCI video card.
>> * Price.
>
>And IBM will merely try to get people to pay twice as much for
>incompatible hardware. Somehow, I think a little more is planned.
IBM has never been known to offer good price/performance. What they
do provide is superb reliability and support, however.
>> * Speed.
>
>Your chip-mania is lunacy... Let's try a new lens...
Unfortunately, you've been accepting all of the advertising propaganda
that Motorola and Apple have been force-feeding you in MacWorld.
>Q2,'95, PPC 604 in machines, chip costs around $400 at 100MHz.
>
>Q4,'95 P6 in machines, chip costs around $1100 at 133MHz.
>
> PPC 604 matches P6 performance (or betters it) with
> 133MHz and 150MHz versions. 100MHz version is $250.
> Developers routinely recompiled Win32 apps for PPC.
On what basis do you claim that a 604 will outperform, or even match, a
P6?
The P6 machines, Intel has said, will deliver 250-300 SpecInt, while I
would expect the 100 Mhz 604 (who knows when the higher clock rate 604's
will be available) will produce SpecInt's in the range of 150.
> PPC 620 shipping in quantity. Initial price, $999.
>
>Q1,'96 P6 machines now available in quantity. Few willing to pay
> the high price. Really fast P5s keep those not looking for
> change quite happy. Some, looking for price/performance,
> begin thinking about PPC machines.
Given Intel's current pricing scheme and how it has undercut Pentium
prices, your claim that few will be willing to pay for a P6 sounds rather
unrealistic.
>Q2,'96 PPC620 machines ship. Faster 604s, P5s, P6s, abound. Nothing
> can touch 620 in the PC marketplace. Machines expensive.
> PPC604 chip price now at about $150. P6 at about $750.
The 620 will likely offer performance on-par with the P6. I don't think
that there is a snowball's chance in hell that is will be out of the P6's
league as you suggest.
>Q4,'96 Word of the PowerPC 800 series just swept Comdex (IBM and Moto
> did a big show on the new series). Systems should begin appearing
> in 12-18 months.
Where are all those anti-Intel folks complaining about Intel pre-announcing
its products and plans?
And what happened to the 700 series?
> Intel/HP briefing ignored. Who cares about a chip coming out in two
> years?
Then why do you mention the 800 series?
> HP again considers buying Apple.
This was a joke, right? Given the HP/Intel P7 alliance, I'd say this one
is way out there in left field.
>Q2,'97 Intel fights back with much cheaper P6s and much faster ones.
From here on out all you or I can do is make baseless speculations.
>IBM's PPCs will have preemption, threads, telephony, video, etc. (as soon
>as they ship).
Thanks to NT.
>Macs will have threads, telephony, video, etc. (as soon as
>7.5 ships). I don't think lack of preemption will kill, or even severely
>wound, Apple's efforts to keep -- and perhaps increase -- its market
>share.
The Mac OS is probably at once both the PMac's greatest advantage and its
greatest hindrance. It is an advantage when it comes to a consistent UI that
fully takes advantage of the graphics hardware and supports great Plug and
Play; while it is a disdvantage when it comes to pre-emption, mission-critical
tasks, networking, security, stability, file system, etc.
> Also, that PnP stuff and multimedia will still be better on Macs
>(because they've always been plug and play and because QuickTime is really
>going to win the race over Video for Windows -- call Bell Atlantic if you
>disagree).
QuickTime and Video for Windows are very comparable in their 32bit
incarnations. However, I don't see either of them "winning" so much as I
see the MPEG-2 standard becoming the dominant medium for animation.
>The original 486SX chips were indeed 486DX chips with the FPU disabled.
>Over time, this did chain. True, no drills were involved.
Hasn't Intel repeatedly denied this?
--
Nathan D. T. Boyd Sequent Computer Systems, Inc.
bo...@sequent.com Beaverton, OR 97009
(503) 578-4506
Mark Rogowsky
wdo...@hilbert.coe.northeastern.edu (Patrick Doyle) wrote:
> 1) I'm a home user. I've spent $3000 over the years on my PC and now
> have it configured just about the way I want it. I have a 14.4 kbps
> modem, a Sound Blaster, a CD-ROM, and ~350 MB of disk space on two
> disks. Of course, since I bought the CPU first, it is only a 33 MHz
> '386. I would like to upgrade, but I don't want to throw away too
> much of my existing equipment. I could buy a PPS, but they will all
> be shipping with built in audio and CD-ROM, so I have to throw out my
> SB and CD-ROM (or try to sell, which is just as annoying). I don't
> know if the PPS uses SCSI or IDE for its disk technology, but it
> better be IDE, or I'll have to throw out the disk as well. I
> understand that using these periperals only under emulation will be
> slow, but I can accept that (after all, I'm used to a 33 MHz '386 :-)
> but I expect, if the PPS revolution catches on, native device drivers
> and other disk controllers will be made available. If it doesn't
> catch on, well at least I'm no worse off than I was before.
You have a major fallacy here which is common on the net and utterly
irrelevant in the real world: New machines come with big disk drives. Most
Pentium machines have 540MB drives. Is this holding back sales to users
with old, pathetically slow drives on their system who hope to bring those
drives along (and then do what, sell the old 386 as a boat anchor)? No,
it's totally irrelevant. Not on the radar screen of PC cloners nor of
PowerPC makers.
Mark Rogowsky
ro...@forsythe.stanford.edu <-- Currently reflects to the QuickMail box
ro...@cindy.stanford.edu <-- The QuickMail/SMTP box
Mark Rogowsky
Rodgers) wrote:
> In article <rogo-310...@tip-mp4-ncs-10.stanford.edu>,
> ro...@forsythe.stanford.edu (Mark Rogowsky) wrote:
> > In article <gfjEkO9R...@wam.umd.edu>, rsro...@wam.umd.edu (R S
> > Rodgers) wrote:
> > > I've personally tested the Diamond Stealth 64 4MB at more than 42 on a
> > > dx4-100. Since the original poster had a Viper (2/3rds the speed of
> > > the S64 at best, 1/2 average) and a Pentium-90, 100 is a little hard
> > > to believe, but 70-75 wouldn't be, and 80+ for an s64 would be
> > > believable.
> >
> > Why has no system ever attained anything near 80 with the latest WinBench
> > according to PC Magazine?
>
> I have no idea.
I'll take that as an "I'm using the older WinBench which gives me 80 and is
widely regarded as meaningless by the folks who write WinBench."
Yiupun Michael Kwong
Nathan Boyd <bo...@sequent.com> wrote:
>In article <rogo-29079...@cindy.stanford.edu>,
>Mark Rogowsky <ro...@forsythe.stanford.edu> wrote:
>>In article <issa.775452041@cwis>, is...@cwis.unomaha.edu (Issa El-Hazin) wrote:
>The P6 machines, Intel has said, will deliver 250-300 SpecInt, while I
>would expect the 100 Mhz 604 (who knows when the higher clock rate 604's
>will be available) will produce SpecInt's in the range of 150.
Are you sure you got your figures correct. From what I heard, a 100
MHz P6 (or was it 150 MHz?) will deliver a SpecInt of around 200.
Faster than a similarly clocked 604, but not *that* much faster.
>>Q2,'96 PPC620 machines ship. Faster 604s, P5s, P6s, abound. Nothing
>> can touch 620 in the PC marketplace. Machines expensive.
>> PPC604 chip price now at about $150. P6 at about $750.
>
>The 620 will likely offer performance on-par with the P6. I don't think
I suspect 620 will perform better than P6. 620 is supposed to be 2x
plus faster than the 604. I personally doubt P6 is going to be twice
as fast as 604.
>that there is a snowball's chance in hell that is will be out of the P6's
>league as you suggest.
>
>>The original 486SX chips were indeed 486DX chips with the FPU disabled.
>>Over time, this did chain. True, no drills were involved.
>
>Hasn't Intel repeatedly denied this?
From what I've heard, the original 486SX are the exact same chip as
486DX, only the FPU doesn't work. I don't know whether Intel
deliberately disabled the FPU or not. A more likely scenario is,
486SX were really 486DX that did not pass the quality control: the FPU
subsystem doesn't work. These chips, which cannot be sold as 486DXs,
aret then sold as 486SXs at a lower price. That's my speculation
only, though.
I heard Intel has since changed the 486SX design. The 486SX chips we
can get these days are not the same ones as 486DXs.
>
>--
>
>Nathan D. T. Boyd Sequent Computer Systems, Inc.
>bo...@sequent.com Beaverton, OR 97009
>(503) 578-4506
Michael Y. Kwong
Electrical Engineering
Stanford University
Timothy Jehl~
> From what I've heard, the original 486SX are the exact same chip as
> 486DX, only the FPU doesn't work. I don't know whether Intel
> deliberately disabled the FPU or not. A more likely scenario is,
> 486SX were really 486DX that did not pass the quality control: the FPU
> subsystem doesn't work. These chips, which cannot be sold as 486DXs,
> aret then sold as 486SXs at a lower price. That's my speculation
> only, though.
>
is worth. Nothing. Not only would Intel not do this, but, to the best of
my knowledge, none of the major (reputable) uprocessor vendors would do this.
There has been, in the recent past on this news group, an extended discussion
of why this is not a reasonable mode of operation.
> Michael Y. Kwong
> Electrical Engineering
> Stanford University
TJ Jehl
--
Tony Hamilton
> In article <31gi0n$h...@ornews.intel.com>,
> Tony Hamilton <tham...@fnugget.intel.com> wrote:
> >The fact that x86 _can_ be even further optimized for the Pentium processor
> >is completely irrelevant. It is not required to achieve significant performance
> >increases.
>
> In other words, even recompiling an app with Pentium-specific
> optimizations (or whatever Dennis O'Connor's really long phrase about
> instruction times describing this is, since he doesn't like the term
> "native"), the performance will not significantly increase.
You need to take a course in logic. "In other words" should be changed to
read "Based in no part on the words which were spoken...". I made no claims,
positive or negative, about what recompilation provides. I _did_ state
that recompiling with Pentium optimizations wasn't necessary for significant
performance increases - implying that unoptimized code alone will see
signficant boosts just by running on a Pentium vs. a 486. There is nothing
in that statement from which one could logically deduce anything about what
optimization _does_ provide.
Rodolfo Paiz
>OS/2 and NT are certainly not "the same coin":
> * NT is now a mature (or at least maturing) product, while WOS
> isn't even in Beta
> * NT/PPC is simply another port of NT, which is already on Intel,
> Alpha, MIPS, etc.
> * IBM is going to bundle NT with its PPS machines until WOS becomes
> available; how can this not hurt OS/2's future?
Now wait a second here... the rest of your post makes sense, and is
reasonably fair. But your first statement here is selective reality
at best.
Just for the sake of balance in the argument, I note that OS/2 for x86
was at least a maturing product while NT was, for a long time, still
vaporware.
-----
****************************************************************************
* Rodolfo J. Paiz * *
* Black Diamond Technology, Inc. * "You have to start a circular argument *
* rp...@husc.harvard.edu * SOMEWHERE..." --Prof. Westervelt *
* Tel # (617) 493-2109 * Harvard University *
* Fax # (617) 493-2109 * *
****************************************************************************
--
-----
****************************************************************************
* Rodolfo J. Paiz * *
* Black Diamond Technology, Inc. * "You have to start a circular argument *
* rp...@husc.harvard.edu * SOMEWHERE..." --Prof. Westervelt *
* Tel # (617) 493-7733 * Harvard University *
* Fax # (617) 493-7731 * *
****************************************************************************
Adam Nash
>MHz P6 (or was it 150 MHz?) will deliver a SpecInt of around 200.
>Faster than a similarly clocked 604, but not *that* much faster.
>
Actually, the Intel media now says 133Mhz P6 gets ~200 specInt (peak)
Noting a 100MHz 604 gets 160, straight scaling would put a 604 at
133MHz at ~213 specInt, which beats the P6. I would also expect the
PowerPC to scale better at higher speeds due to its RISC design.
For example, look at the P5. The 66MHz P5 has higher specInt than the
66MHz PPC 601, BUT the 100MHz 601 is clearly a notch above the 100MHz P54C
So I would expect a 200MHz 604 to be ~15-20% faster than the 200MHz P6.
But, this is speculation based on the P5s scaling. The P6's cache may
help it some. Nonetheless, it seems clear to me that the 604 beats
the P6 at the same speed, but just a little, and who knows what speeds
who is going to ship when?
>
>I suspect 620 will perform better than P6. 620 is supposed to be 2x
>plus faster than the 604. I personally doubt P6 is going to be twice
>as fast as 604.
>
I already answered the P6/604 question. The PPC620 may eventually double the
604, but not initially. I hear they plan to ramp the P6 up to 400MHz.
I hear they also plan to ramp the PPC620 that high. Eventually for both.
However, at the same speed, the 620 should be 30-40% faster than the P6,
AND it should scale better, see above.
Interestingly enough, the P6 and the PPC620 seem to be on the same time scale,
a 4Q94 announcement, and a 3Q95 shipping goal. However, the 604 is the answer
to faster P5s, and it CLEARLY wins that fight, as even a 150MHz P5 will at
BEST have a specInt of 150, but probably ~140, which is noticeably slower than
a 100MHz 604.
But, as always, we have to look at a machine/machine level. Currently,
I think it's even. There is some inconvenience for a couple months until
MS Office comes out (and realize guys that that is ALL that's noticeably missing)
Then again, there is inconvenience running Windoze 3.1 since Chicago is
delayed. Then again, The 6100 is pretty cheap. Then again, it only runs System
7.1 (7.5) by Apple. Then Again...
I think Apple whethered this year well. Next year will be the BATTLE.
Faster P5s running Chicago versus PPC 604s running System 7.5, perhaps Copland
towards the fall as the P6s see the light of day...
I'm not sure I'm ready to assume that Intel/Microsoft will outsell
Apple/PowerPC (since their may be mac clones) 8:1.
6:1 I can believe, but with Copland on 604s...4:1 seems more likely.
AND that, my friends is what 20% marketshare is all about...
-Adam
PS Please email me if you have some new/different info...
Dennis O'Connor~
adam...@Xenon.Stanford.EDU (Adam Nash) writes:
] Actually, the Intel media now says 133Mhz P6 gets ~200 specInt (peak)
] Noting a 100MHz 604 gets 160, straight scaling would put a 604 at
] 133MHz at ~213 specInt, which beats the P6. I would also expect the
] PowerPC to scale better at higher speeds due to its RISC design.
A reminder : SPEC numbers do _not_ scale with frequency, unless
every component of the system also scales it's speed. Now, if you
can get DRAM, hard-drives, busses et cetera to do that, you'll be
famous and rich, because no else has managed it yet.
Clint Olsen
Adam Nash <adam...@Xenon.Stanford.EDU> wrote:
>>Are you sure you got your figures correct. From what I heard, a 100
>>MHz P6 (or was it 150 MHz?) will deliver a SpecInt of around 200.
>>Faster than a similarly clocked 604, but not *that* much faster.
>>
>
>Actually, the Intel media now says 133Mhz P6 gets ~200 specInt (peak)
>Noting a 100MHz 604 gets 160, straight scaling would put a 604 at
>133MHz at ~213 specInt, which beats the P6. I would also expect the
>PowerPC to scale better at higher speeds due to its RISC design.
"from what I heard"
"I would expect"
"straight scaling would put"
I don't know why you people spend so much time trying to perform
scaling on stuff that isn't here yet. Don't you have something
better to do with your time than to play into this?
Comparisons at this level are garbage.
-Clint
Nevin Liber
Tony Hamilton <tham...@fnugget.intel.com> wrote:
>I _did_ state
>that recompiling with Pentium optimizations wasn't necessary for significant
>performance increases - implying that unoptimized code alone will see
>signficant boosts just by running on a Pentium vs. a 486. There is nothing
>in that statement from which one could logically deduce anything about what
>optimization _does_ provide.
<sigh> Maybe I've just been unlucky this week, but it sure seems like
you Intel folk would rather discuss the phrasing of a posting than to
actually discuss the issues. (I don't intend this as a flame; the
quality of the technical posts from the Intel people is really good;
I'd just like to see a bit more of them over discussings about
phrasing. I may be partially to blame for that; if so, I do
apologise.)
Hopefully this is clear enough: what kind of performance increases can
someone expect if they are compiling an app with 80486 optimisations
turned on vs. Pentium optimisations turned on, with both versions of
the app running on the same Pentium-based system?
Please don't get caught up on the fact that I didn't define the rest of
the system; I'd just like to know some ballpark percentages. If you
need to, go ahead and define the rest of the system and use that as the
context for answering the question.
--
Nevin ":-)" Liber ne...@cs.arizona.edu (602) 293-2799
+++ (520) after 3/95
summer office: (602) 621-8112
Only 8 more shopping days 'til my birthday (August 12th)!!
Adam Nash
>>Noting a 100MHz 604 gets 160, straight scaling would put a 604 at
>>133MHz at ~213 specInt, which beats the P6. I would also expect the
>>PowerPC to scale better at higher speeds due to its RISC design.
>
>"from what I heard"
>"I would expect"
>"straight scaling would put"
>
>I don't know why you people spend so much time trying to perform
>scaling on stuff that isn't here yet. Don't you have something
>better to do with your time than to play into this?
>
>Comparisons at this level are garbage.
Actually, they are not. I'd actually say they are very close to the truth.
The RISC design of the PowerPC does favor scaling better than the P5 or
probably the P6 architecture. That's why in SpecInt, a 66MHz P5 outperforms
a 66MHz 601, , but the 100MHz situation is reversed.
I'm glad you are so much better than us Clint,
but in reality this type of discussion, on the scalability of the x86
design and it's future incarnations may be arguably what this newsgroup
is for.
Don't you have an emulator to write or something?
-Adam
103t_e...@west.cscwc.pima.edu
I don't think that you can generalize about scalability from the example of the
601 vs the P5 and the 601+ vs the P54C.
The P5 uses a full-speed data bus and 512k of cache while the 601 uses a 1/2
speed data bus and no L2 cache.
OTOH, the P54C uses a 2/3 speed data bus with 512K of cache while the 601+ uses
a 2/3 speed data bus and 1 meg of L2 cache.
Difficult to say how well each scales compared to the other, I think.
Lawson
Adam Nash
Dennis O'Connor~ <doco...@sedona.intel.com> wrote:
>
>adam...@Xenon.Stanford.EDU (Adam Nash) writes:
>] Actually, the Intel media now says 133Mhz P6 gets ~200 specInt (peak)
>] Noting a 100MHz 604 gets 160, straight scaling would put a 604 at
>] 133MHz at ~213 specInt, which beats the P6. I would also expect the
>] PowerPC to scale better at higher speeds due to its RISC design.
>
>A reminder : SPEC numbers do _not_ scale with frequency, unless
>every component of the system also scales it's speed. Now, if you
>can get DRAM, hard-drives, busses et cetera to do that, you'll be
>famous and rich, because no else has managed it yet.
This only has limited truth here. Spec numbers historically have scaled,
with degradation based on how their architecture deals with the
higher speed. Note the Alpha specs, for a scalable architecture at
higher speeds, where the specMarks reasonably do scale.
Anyway, the point was that from preliminary info, the 604 and P6 look
equivalent in performance since at 160 specInt at 100MHz, getting a
25% boost in spec from 33% boost in speed is easy, especially since all
these specMarks are on 66MHz buses, so we're only talking clock-doubling
My other point was that not only to specMarks degrade while scaling with
speed, but due to the scalability of the RISC design, the PowerPC
performance degrades LESS with speed than the x86, at least you can
see that with the Pentium
That's why the 66MHz P5 specInt > 66MHz 601 specInt, but the
100MHz 601 specInt > 100MHz P54C specInt
-Adam
PS I'm sure you have info on the subject, since you are on the i960
design team. Doesn't the i960 scale with speed better than CISCy
competitors?
Patrick Chase
|> My other point was that not only to specMarks degrade while scaling with
|> speed, but due to the scalability of the RISC design, the PowerPC
|> performance degrades LESS with speed than the x86, at least you can
|> see that with the Pentium
|>
|> That's why the 66MHz P5 specInt > 66MHz 601 specInt, but the
|> 100MHz 601 specInt > 100MHz P54C specInt
No, no, no. The reason the 100MHz 601's quoted SPECint results are better
than the P54C's is because Motorola's testing methodology changed between
the 601-66 and the 601-100. The quoted SPEC results for the 601/66 were
obtained with a 33 MHz system bus and no L2 chache. The results for the
601/100 were obtained with a faster system bus and a 1 MB L2 cache. Both
Intel processors, on the other hand, were run with a 66 MHz system bus and
a 512KB cache.
When people say "RISC processors are more scalable", they mean that in their
opinion it's easier to increase the clock speed and/or improve the micro-
architecture to get better performance on a RISC than on a CISC. They do
not mean that increasing the clock speed by a given amount on a RISC will
result in a larger performance boost than increasing the clock by the same
amount on a CISC chip which had equivalent performance to start out with.
Although in specific cases this might be true, it's a complex function of
memory access patterns and memory architecture, as well as several other
inputs. One could even argue (albeit probably ineffectively) that the
increased code density of a CISC makes it more memory-friendly, and hence
more likely to scale well if clock speed is increased in the absence of
memory-subsystem improvements than a RISC. Only an idiot or a technical
incompetent would be stupid enough to generalize either way, though.
P.S. - I'm still not convinced that clear-cut classifications like "RISC" and
"CISC" can be applied to real-world processors anyway, so that would
seem to make the whole argument rather moot, wouldn't it?
|>
|> -Adam
|>
-----------------------------------------------------------------------------
Patrick Chase Not speaking for Hewlett-Packard...
H-P San Diego
Paul Rubin
Adam Nash <adam...@Xenon.Stanford.EDU> wrote:
>Actually, the Intel media now says 133Mhz P6 gets ~200 specInt (peak)
>Noting a 100MHz 604 gets 160, straight scaling would put a 604 at
>133MHz at ~213 specInt, which beats the P6. I would also expect the
>PowerPC to scale better at higher speeds due to its RISC design.
What does "peak specInt" mean???? I though the whole idea of
SPEC was to get rid of bogus measurements of "peak MIPS" (number
of nop's per second the cpu can execute, especially on superscalar
and other parallel architectures).
Henry J. Cobb
>A reminder : SPEC numbers do _not_ scale with frequency, unless
>every component of the system also scales it's speed. Now, if you
>can get DRAM, hard-drives, busses et cetera to do that, you'll be
>famous and rich, because no else has managed it yet.
>
>Dennis O'Connor doco...@sedona.intel.com
>Intel i960(R) Microprocessor Division Solely responsible for what I do.
A reminder : SPEC numbers do _not_ scale with Hard Drive or system
bus speeds on any decent machine. If your system has at least 2 MB of
cache the DRAM speed doesn't matter either.
--
Henry J. Cobb hc...@fly2.berkeley.edu
All items Copyright (c) 1994, by their respective authors, permission is
granted to redistribute as long as proper credit is given.
103t_e...@west.cscwc.pima.edu
> In article <rogo-29079...@cindy.stanford.edu>,
> Mark Rogowsky <ro...@forsythe.stanford.edu> wrote:
[much snipt]
>
>> Also, that PnP stuff and multimedia will still be better on Macs
>>(because they've always been plug and play and because QuickTime is really
>>going to win the race over Video for Windows -- call Bell Atlantic if you
>>disagree).
>
> QuickTime and Video for Windows are very comparable in their 32bit
> incarnations. However, I don't see either of them "winning" so much as I
> see the MPEG-2 standard becoming the dominant medium for animation.
>
Hmmm... I'm not so sure about this. Care to list the features of QuickTime 2.0
vs Video for WIndows? I think that you will be *very*, *VERY* surprised as to
what QT 2.0 (or even 1.6.x) does...
Lawson
Nathan Boyd
Yiupun Michael Kwong <yiu...@leland.Stanford.EDU> wrote:
>In article <1994Aug3.1...@sequent.com>,
>Nathan Boyd <bo...@sequent.com> wrote:
>>
>>would expect the 100 Mhz 604 (who knows when the higher clock rate 604's
>>will be available) will produce SpecInt's in the range of 150.
>
>MHz P6 (or was it 150 MHz?) will deliver a SpecInt of around 200.
>Faster than a similarly clocked 604, but not *that* much faster.
I think that your 200 SpecInt number is exaggerated; from all reports that
I have read the 604 will be approximately 50% faster than the 601 at the
same clock rate (as opposed to the approximately 100% increase your figure
implies).
J. Heather Patrick
Bryan Butler <bbu...@netcom.com> wrote:
>Don't know about the drill or the exact technique used, but supposedly the
>SX and DX dies were the same when they first came out, but the SX had the
>FPU disabled somehow (the FPU was there, you just couldn't use it). THis is
>hearsay, and it might not be true anymore, given the volume of SX and DX
>chips.
>
>I heard about a line of minis that DG sold back in the 70s. The high end
>machine was identical to the low end machine. If you bought the low end \
>machine and wanted an upgrade, they would sell you a tape with new microcode
>that deleted extra wait states to make it run faster. Again, hearsay.
Well, I believe IBM was the firm that had the following rumour attached
to it..
Back in the old days of mainframes, some company sold two fairly similar
mainframes. Completely binary-compatile with each other, except one model
was faster. And you could "upgrade" the slower model to a faster one.
The rumor was that the "upgrade" consisted of a technician coming in, and
cutting (or adding?) a single wire. For which the company of course got
paid $$$$$
Terje Mathisen
This wasn't IBM, it was the old Unisys (Univac?) which had a series of
machines called U1100. The wire that was cut would, while still installed,
tie one address line to ground, thereby limiting max memory.
After the expensive "snip", you still had to pay for the memory cards.
-Terje
Pat Caudill
>>mainframes. Completely binary-compatile with each other, except one model
>>was faster. And you could "upgrade" the slower model to a faster one.
>>
>>The rumor was that the "upgrade" consisted of a technician coming in, and
>>cutting (or adding?) a single wire. For which the company of course got
>>paid $$$$$
Humm, There were several companys that this has been attributed to. The NCR
Century 100 could be converted to a 200 by removal of a wire in the clock
curcuit. The IBM 360/67 could go to a 360/67-II at a hefty rental rate
increase with the flipping of an internal switch (and the book that told
where to find it). Many IBM card readers could be made to read much faster
(up to 12 times) by moving the belt to a different pulley or changing the
gears on the drive shaft. At one shop I worked in the IBM rep swapped the
big and little pulley to see what would happen. It read very, very fast but
the cards missed the output hopper and splattered all over the wall. It was
a great hose effect.
Harry H Conover
: It is worse than that. The machine was artificially slowed down not
: out of simple greed, but because the GOVERNMENT REQUIRED IT (part of
: the antitrust deal, price/performance ratio was not allowed to exceed
: a certain amount, and they wanted to be able to offered a low priced
: machine). I believe that at least one model of the 1130 was done this
: way for this reason.
Intentional crippling of machine performance still continues, at least
as recently as last September (Sept. 1993).
When hp introduced its Model 712 workstations, the 7200LC chips SPECfp
was throttled back. Presumably, to protect the Model 715/75 from lower
priced Model 712 competition.
After significant pressure from both industry press and its customer
base, hp recanted and now ships full-performance 712 systems.
Hey, you can't blame a guy for trying! ;-)
Harry C.,
103t_e...@west.cscwc.pima.edu
I don't know about IBM, but the Burroughs B4700 was actually a B3700 with a
faster clock-chip and a "4" pasted over the "3" on the console...
Lawson (B35/37/4700 computer operator in USAF, 1978-1983)
Eric Liber
Ok one more ... but NOT HEARSAY ... I saw this happen ... I was there..
Many years ago I worked in an office where we decided to use Wang word
processing equipment on a Wang NET. The server had a massive 4MB (yes 4
four...) Hard drive. Of course we soon ran out of storage and wang
offered an upgrade to a larger 8MB drive for about $4000. We bought the
upgrade and when the technician showed up to "install" the 8MB drive he
opened the server box and removed one jumper ... instant 8MB! Fortunately
the boss saw this with me and he called Wang and had them remove their
equipment the next day... boy was he mad.
Eric Liber
Consulting Analyst
Westinghouse Information Systems
Westinghouse Electric Corp.
"Dopo il fatto il consiglio non vale" - an old Italian Proverb
Paul Rubin
J. Heather Patrick <jpat...@crl.com> wrote:
>Well, I believe IBM was the firm that had the following rumour attached
>to it..
>Back in the old days of mainframes, some company sold two fairly similar
>mainframes. Completely binary-compatile with each other, except one model
>was faster. And you could "upgrade" the slower model to a faster one.
>
>The rumor was that the "upgrade" consisted of a technician coming in, and
>cutting (or adding?) a single wire. For which the company of course got
>paid $$$$$
It is worse than that. The machine was artificially slowed down not
103t_e...@west.cscwc.pima.edu
> Adam Nash <adam...@Xenon.Stanford.EDU> wrote:
>>Dennis O'Connor~ <doco...@sedona.intel.com> wrote:
>>
>>100MHz 601 specInt > 100MHz P54C specInt
>
> by more than just the CPU; for example, the bus speed and cache design/size
> also play important roles.
>
> Without knowing more about the testing methodologies used for all four
> SpecInt numbers (66 Mhz for both CPU's, 100 Mhz for both CPU's) it is highly
> unlikely that Apple's significantly lower 66 Mhz SpecInt would suddenly
> become higher at 100 Mhz.
66mhz P5:
66hz data bus, 512K L2 cache, 66mhz CPU
66mhz 601:
33mhz data bus. no L2 cache, 66mhz CPU
100mhz P54C:
66.7mhz data bus, 512K L2 cache, 100mhz CPU
100mhz 601+:
66.7mhz data bus, 1 meg L2 cache, 100mhz CPU
++++++++++++++++++++++++++++++++++++++++++++++++++
There are obvious reasons why the 601+ goes "scales better" then the P54C...
>
> PS - Don't take all of IBM's/Apple's/Motorola's hype about the PPC architecture
> too seriously. Yeah, it is an excellent architecture, but it isn't the
> panacea they would have you believe.
Ditto for hype about Px architecture...
Lawson
Nathan Boyd
>Dennis O'Connor~ <doco...@sedona.intel.com> wrote:
>
>That's why the 66MHz P5 specInt > 66MHz 601 specInt, but the
>100MHz 601 specInt > 100MHz P54C specInt
This may be true, but your reasoning is invalid. SpecInt scores are affected
by more than just the CPU; for example, the bus speed and cache design/size
also play important roles.
Without knowing more about the testing methodologies used for all four
SpecInt numbers (66 Mhz for both CPU's, 100 Mhz for both CPU's) it is highly
unlikely that Apple's significantly lower 66 Mhz SpecInt would suddenly
become higher at 100 Mhz.
PS - Don't take all of IBM's/Apple's/Motorola's hype about the PPC architecture
too seriously. Yeah, it is an excellent architecture, but it isn't the
panacea they would have you believe.
--
103t_e...@west.cscwc.pima.edu
> In article <1994Aug1...@west.cscwc.pima.edu>, 103t_e...@west.cscwc.pima.edu writes:
> |> In article <1994Aug11.1...@sequent.com>, bo...@sequent.com (Nathan Boyd) writes:
>
>
>
> |> There are obvious reasons why the 601+ goes "scales better" then the P54C...
>
>
> Have heard this claim from time to time, but perhaps missed the
> reason why it is true (if it is indeed true).
>
> <lotsa more deleted>
>
I'm sorry, I was being sarcastic. THat is why I put "scales better" in quotes.
The 66mhz and 80mhz MPC601's were using no L2 cache and running on a 1/2 speed
data bus, while the 66mhz Pentiums HAVE L2 cache and were running on a
full-speed data bus.
The 100mhz MPC601+ SPEC and the 100mhz P54C SPEC both use L2 cache (more for
the 601+ in fact) and both are running on 2/3 speed data buses.
OBVIOUSLY, the 601+'s are going to show more of a speedup then the P54C's,
assuming that the design of the chips otherwise remained the same.
Lawson
Hank Oredson
<lotsa deleted>
|> There are obvious reasons why the 601+ goes "scales better" then the P54C...
Could you explain why this is so?
Have heard this claim from time to time, but perhaps missed the
reason why it is true (if it is indeed true).
<lotsa more deleted>
... Hank
--
Hank Oredson @ Mentor Graphics Library Operations
Internet : hank_o...@mentorg.com "Parts 'R Us!"
Amateur Radio: W0...@W0RLI.OR.USA.NOAM
Bob Dalgleish
>In article <bbutlerC...@netcom.com>,
>Bryan Butler <bbu...@netcom.com> wrote:
>>SX and DX dies were the same when they first came out, but the SX had the
>>FPU disabled somehow (the FPU was there, you just couldn't use it). THis is
>>hearsay, and it might not be true anymore, given the volume of SX and DX
>>chips.
More than likely the disabling occurred after the chips were tested.
Given the expense of wafer production, and the desire to increase
yields, some Intel bright boy (not disparaging, I think this was good
economic sense) decided to test FPU versus non-FPU functionality. This
increased production yields, and helped drive down the cost of all of
the chips -- think about it: a wafer costs $X to make and the nominal
yield is 15%. That makes the manufactured cost of each chip to be
$X*15%*chips/wafer. Change the selection process to select chips with
nonfunctioning FPUs, which might have double the nominal yield. This
drops the cost of each chip to $X*45%*chips/wafer, or about one-third of
the original cost.
>Well, I believe IBM was the firm that had the following rumour attached
>to it..
>Back in the old days of mainframes, some company sold two fairly similar
>mainframes. Completely binary-compatile with each other, except one model
>was faster. And you could "upgrade" the slower model to a faster one.
IBM didn't sell mainframes for the most part (other than under terms
negotiated for them under a consent decree); they leased them, and they
leased them by the CPU cycle. The Model 20 and Model 25 machines of the
System 360 line were the same hardware, but there was a switch on the
back that read (apocryphally) "Fast/Slow". There was also a meter on
the computer that counted how many CPU cycles were consumed and you were
billed based on that. It did not pay you to turn the switch, unless you
had an emergency. Since this was in the late '60's, it was not uncommon
for these computers to run only one shift a day. Seldom would you run
two shifts, since that would double the number of CPU cycles that you
"consumed". To put this in perspective the difference in speed between
the 20 and 25 were on the order of 20 microsecond cycles and 30
microsecond cycles (easy for me to be wrong on these). Since it took
between four and eight cycles to load a register from memory -- you do
the math.
--
-- * * * CFV: net.short.signatures * * *--
Bob Dalgleish zaphod!bo...@tribune.usask.ca CompuServe: 70521,2011
Jeffrey Reilly
>In article <1994Aug11.1...@sequent.com>, bo...@sequent.com (Nathan Boyd) writes:
>> Adam Nash <adam...@Xenon.Stanford.EDU> wrote:
>>
>> Without knowing more about the testing methodologies used for all four
>> SpecInt numbers (66 Mhz for both CPU's, 100 Mhz for both CPU's) it is highly
>> unlikely that Apple's significantly lower 66 Mhz SpecInt would suddenly
>> become higher at 100 Mhz.
>
>66mhz P5:
>
>66hz data bus, 512K L2 cache, 66mhz CPU
>
Hmmmmm... a minor point of correction....
data from various issues of the SPEC Newsletter (June 1993, Sep 1993 and June
1994) show the 66MHz Pentium Processor results were run with a 256KB L2 cache
and NOT a 512KB L2 Cache...
Jeff
Associate Editor, SPEC Newsletter
Jeff Reilly | "There is something fascinating about
Intel Corporation | science. One gets such wholesale returns
jwre...@mipos2.intel.com | of conjecture out of such a trifling
(408) 765 - 5909 | investment of fact" - M. Twain
Bruce Hoult
> <lotsa deleted>
>
> |> There are obvious reasons why the 601+ goes "scales better" then the P54C...
>
> Could you explain why this is so?
>
> Have heard this claim from time to time, but perhaps missed the
> reason why it is true (if it is indeed true).
Well, if you'd cared to *read* the article from which you "<lotsa deleted>"
then you would know.
Simply, the benchmark for the 100 MHz 601+ was done on a system with a big
external cache, while the benchmark for the 66 MHz 601 was done on a system
with no external cache at all.
That's why the 100 MHz version had more SPEC/MHz. It doesn't scale (with MHz)
better at all. That's why the original poster put the phrase in quotes.
-- Bruce
103t_e...@west.cscwc.pima.edu
You forgot the 2-fold speedup of the data-bus. That accounts for some of it
also, I'd guess...
Lawson
Reinhard Kirchner
>
> This wasn't IBM, it was the old Unisys (Univac?) which had a series of
> machines called U1100. The wire that was cut would, while still installed,
> tie one address line to ground, thereby limiting max memory.
>
> After the expensive "snip", you still had to pay for the memory cards.
to build only one machine in higher quantity and to introduce some
artificial slowdowns than to design and build separate machines.
From a Siemens man I heard of a case where they had to downgrade a machine
by installing 'slowdowns' for several 10-thousand DM so the customer
could pay less rent for this machine!
Reinhard Kirchner
Univ. of Kaiserslautern, Germany
kirc...@informatik.uni-kl.de
Yohannon
Patrick) wrote:
Folks, I am here to CONFIRM that the idea of deliberately crippling
hardware for marketing reasons is not only true, I've participated in the
"design" of such a card.
I worked for supermac for a few years, and it was common for engineering
to come up with new products *based on the same card design*. 100%
identical. It started with the Spectrum/8 - ColorCard/8 for the SE/30.
The Spectrum/8 supported more monitors and cost $1000.00 dollars more.
The colorcard only supported the 13" Apple display. The difference was a
ROM.
After the Thunder/24 came out with accelerated 24 bit graphics, someone in
marketing decided we needed mid level/low end cards to "fill in" the
product line. Essentially I took a thunder/24, walked into engineering,
and they put on a slower clock. The only change they made to the ROM was
to give the card a new name...essentially a saavy end-user could buy the
slower card, add a 40Mhz osc, and save (at the time) something like 1800
bucks.
Other cards didn't even change the oscillator...engineers just inserted
wait states in the ROM code. For $50.00 a user could buy a ROM "Upgrade"
for the faster product, and find themself with a much faster card.
Guess what? The higher end cards sold as well as the lower end, in some
cases better. That's how a lot of hardware co.'s make money.
The equivalent in some people's eyes would be if a car dealer crippled two
cylinders on a V8 engine and sold it as a cheaper V6. ;-)
~Yohannon~
Robert C. Lokerson
> In article <32a3jk$p...@crl3.crl.com>, jpat...@crl.com (J. Heather
> Patrick) wrote:
>
> >
> > Well, I believe IBM was the firm that had the following rumour attached
> > to it..
> > Back in the old days of mainframes, some company sold two fairly similar
> > mainframes. Completely binary-compatile with each other, except one model
> > was faster. And you could "upgrade" the slower model to a faster one.
> >
> > The rumor was that the "upgrade" consisted of a technician coming in, and
> > cutting (or adding?) a single wire. For which the company of course got
> > paid $$$$$
>
I clearly recall an article describing the upgrade of some System 360 IBM
system which only required the (lease) of an upgrade ROM.
R.C. LOKERSON (BOB) MAIL ADDRESS: VOICE : 407.729.2257
HARRIS MDSO/MIS/DCC/ATG HARRIS E.S.S. SMTP : RLOK...@HARRIS.COM
BLDG 15/RM 838 P.O. BOX 37 CCMAIL: RLOKER01
PALM BAY, FLORIDA MELBOURNE FL FCC : WA3PKX
32902 HOME : 407.725.8024
(I WROTE IT, NOT MR. HARRIS)
"Mache'tes dont kill people, people kill people." NRA (Natl Rwandan Army)
Donald A. Johnson
with FPU's that were disabled?
--
Don Johnson
joh...@vpnet.chi.il.us
Mark A. Lee - 4MD
NO!!!! This has been mentioned before, and though a lot of people refuse to believe
it, the FPU was actually removed for the 486SX (experiments may have been done
with in Intel, but the outside world did not see them). Pop the lids off of a DX/25 and
an SX/25. The die will not look the same
---------------------------------------------------------------
Mark A. Lee - MD4
Intel Corporation
Folsom, CA
ml...@pcocd2.intel.com
The opinions expressed are my own and not those of Intel Corp.
---------------------------------------------------------------
