| Monday
March 19, 2001
Jerry,
I've been reading your articles online for quite
some time -- it's nice to have some of the issues placed in a
"user" point of view as you seem to attempt. In your recently
posted article concerning the AMD motherboards and processors vs. Intel's
line, I must say I disagree with you quite strongly.
While I do run a dual processor motherboard for
Intel processors as my primary system at home, it is utilizing a VIA
chipset as Intel refuses to release anything of theirs that you can use
good old SDRAM with (at least in a dual setup for the faster P3s). RAMBUS
memory is finally becoming more of a feasible option, but in my opinion
there's still no compelling reason to use it over traditional SDRAM, and
even less of an argument when considering DDR SDRAM. In that respect, I
agree with your opinion to stay away from the Pentium 4.
I work at your typical mom &; pop computer shop,
and our sales have gone at least close to 50-50 with respect to AMD vs
Intel systems. The fact of the matter is that VIA is producing some very
stable chipsets now and many of the big motherboard manufacturers are
making some very interesting, feature-filled motherboards based around
those chipsets. Most boards -do- have thermal monitoring of some sort and
typically offer many more features than you would normally find on an
Intel motherboard. In addition, considering Intel's change of socket
design with the P4 (and oddly enough I hear they will change it very soon
AGAIN for another revision of the P4), AMD has committed itself very
strongly to pushing their Socket A platform as far as possible. In that
respect, considering AMD based systems on VIA chipsets that are just as
stable, in most cases faster, and in almost every case cheaper than a
comparable Intel system, I can't help but recommend AMD systems over Intel
time after time to my customers. As I stated I still am an
"Intel-man" with my home system, but if something catastrophic
were to happen and it needed replacing, I would spend my money on AMD.
On a side note, it's interesting that I of all
people would be responding in this manner. For quite some time I've been
strongly opposed to AMD products. I've seen far too many K6s and K6-2/3s
go bad for no particular reason and cause general system instability. For
me to be recommending them over Intel really does mean something IMO.
Thanks for you time. For your info, some of the motherboards we use are
the ASUS A7V133, the MSI K7Pro 2-A, and the ABIT K7T.
Jeff Messer jefmes@hotmail.com
It may just be innate conservatism
on my part. I can only go by my own experiences and the reports of people
I rely on. I've had a lot of systems here and I have had fewer problems
with Intel systems on Intel motherboards. That may be peculiar to me, but
it seems to be the case with most of my correspondents.
It's also true that I haven't built
any dual processor systems recently; the dual that I use is a Compaq Dual
SP750 which my workhorse system and has been for over a year. All the
systems I have built in the last year have been single processors of
nearly every kind, and all have used either Kingston or Crucial SDRAM. You
may recall that I gave RAMBUS a dubious achievement award for a fractional
gain in performance at an enormous cost increase.
My colleagues at Nikkei BYTE have
been impressed with Athlon and gave the Athlon system the Best In Show
award at the Tokyo PC EXPO.
My conclusion remains: for single
processor systems, Good Enough is an Intel D815EEAL motherboard with
on-board AGP video, really good sound, and a good LAN card; with the
capability of adding a different AGP video card if that's what you want.
Any processor you are likely to be able to buy for that board will be good
enough; I like 750 P III but you aren't likely to notice the difference in
faster or slower chips (within reason) unless you're doing high end video
processing in which case you already know what you are going to need.
I am grateful to AMD for bringing
the prices down for all of us, and I'll continue to build and use Athlon
systems, and we'll see; but for now the D815EEAL board is the main reason
I prefer P III for "standard" systems.
Dr. Pournelle,
Just a note on your article concerning the lack of
heat and voltage irregularity sensor support on Athlon motherboards. I
have one of the first Athlon boards produced - the FIC SD11, with a VIA
82c686A frontside chipset. This chipset does indeed support such sensors,
as the output from the "sensors" command (part of the lmsensors
package used to talk to i2c devices) on my Linux machine will attest:
debecker@oberon:~$ sensors
eeprom-i2c-0-51
Adapter: SMBus vt82c596 adapter at 0400
Algorithm: Non-I2C SMBus adapter
Memory type: SDRAM DIMM SPD
SDRAM Size (MB): 128
via686a-isa-6000
Adapter: ISA adapter
Algorithm: ISA algorithm
CPU core: +1.61 V +2.5V: +0.22 V I/O: +3.24 V +5V: +3.11 V +12V: +7.55 V
CPU Fan: 6490 RPM (min = 3000 RPM, div = 2) P/S Fan: 0 RPM (min = 3000
RPM, div = 2) SYS Temp: +24.2 C (limit = +60 C, hysteresis = +50 C) CPU
Temp: +29.0 C (limit = +60 C, hysteresis = +50 C) SBr Temp: +24.5 C (limit
= +60 C, hysteresis = +50 C)
No idea why WinTach wouldn't work on an Athlon, i've
had no problems running anything on mine (but then I haven't tried WinTach
and have no idea what it might be trying to do -- for all I know it's
trying to do a CPU ID and failing, and then refusing to run further).
Best regards,
Phil DeBecker
I have no idea why Wintach didn't
work on the Athon system earlier; it does now. On the other hand it gives
such ridiculously low numbers that I have no intention of reporting them.
There is something very wrong here. Note that we have an AMD-built Athlon
with nVIDIA geForce 2 board and it screams at Crimson Skies and such like;
and I have no complaints at all about the Athlon system. (Well, one minor
one: the CD-RW drive was put in as the top drive and the DVD as the lower;
but that's not any fault of the machine or the chips, merely a preference
of mine minor enough that I haven't taken the trouble to adjust -- I won't
say fix or correct.)
I'm not a big testing lab. I draw on
some experts for advice and sanity checking, and I am not getting into
religious wars. I do have more reports of heat and power problems with AMD
than with Intel, but I do not have high quality data -- and it may be that
there are more people trying to overclock AMD systems (because they can)
and that may contribute to the Athlon's sensitivity to power stability and
cooling requirements.
Thanks for the data. And one more
solid commentary:
I'm sure that you're going to get flamed about your
comments about Athlons as many people are passionate about their CPUs. On
my last upgrade last year I went with an P3-550E and overclocked to 733 as
at that time your comments about Athlon MBs were correct. However, over
the last year Intels chipset blunders (ie 810 and 820, the 815 is actually
pretty good just overpriced) gave over 50% of the Intel chipset market to
Via. The Via 694 northbridge chip in my current motherboard is fully equal
to the 815 and came out more than six months before the 815. The 686
southbridge chip which can handle both Via's Intel and Athlon northbridge
chips has voltage and temp monitors built in. Via currently has more than
90% of the Athlon chipset market and the current KT133 MBs and the new
KT133a MBs are equal or better than any Intel chipset board. Athlons are
much cheaper than the equivalent Intel CPUs and the MBs are roughly equal
in quality, features and price. This makes Athlons the pick for most
people and is indeed the current source of Intels troubles in Europe where
AMD currently has more than 50% of the CPU market.
I would mainly agree with your comments about the
Duron situation as the Via KM133 motherboards with built in Savage video
is not in wide distribution yet. However, this will shortly change as the
KM133 is cheaper than the 815, has more features and has much better 3d
video. Since the Duron price/performance advantage over the Celeron is
even bigger than the Athlon-P3 margin, Durons should quickly take over the
low end of the market.
========== Bob Whitworth HST LAN Support Goddard Space Flight Center
rwhitworth@hst.nasa.gov
When I get a KM133 I will be more
than pleased to build a system around it and beat the holy Hell out of it;
as is my bent. And I certainly never recommended the 810. You may recall I
even warned that they had cost-reduced the 810 to a worrisome point; an
observation that was not popular at Intel. Later 810's seem to have
been better built than the first ones I saw but I was never all that happy
with them. The 815, though, is Good Enough, and the sound quality is
splendid.
There are plenty of publications and
writers who specialize in detailed examinations of all these things. I
have always been in the "good enough" business. I will confess
that saving a few dollars, particularly at the expense of time or
uncertainty, has never been high on my priority list.
Thanks for the information, and I'll
look into sources on new Via motherboards.
And for it to be put as bluntly as
possible:
I enjoy your stories about struggles with systems and getting things to
work; I can definitely relate. But you are really off base in your
comments about Athlon vs. Pentium III and Duron vs. Celeron. The
price/performance differences are quite remarkable, in fact. Please check
Tom's Hardware Guide for detailed facts. In my own experience, I just
assembled for home use an Asus A7V133 with 1Gig Athlon, ATA100,
SoundBlaster Live Value, Geforce 2 GTS, 256MB PC 133, Wireless Ethernet,
Antec case, Win2K, DVD, TrippLite UPS with power management, CD-RW, etc.
The system is rock solid. The equivalent Pentium III would have been at
least $200 more. At work, I have an Intel 815-based system with a 866MHz
Pentium III and I am pleased with it also. My experience indicates:
a. At the same price point, Athlon is dramatically faster than Pentium
III. The same difference is true about Duron vs. Celeron.
b. For the same performance, Athlon is much less expensive than the
Pentium III. Same for Duron vs. Celeron
c. At the same clock speed, Athlon performs better than Pentium III and
costs much less. Same for Duron vs. Celeron.
c. For the good boards, such as Asus, VIA systems are rock solid.
d. Intel processors convert to mobile processors for laptops better.
I have been buying Intel, AMD, etc. systems for 20 years. You seem to
have a thing for Intel, which is reasonable given all the goods things
they have done over the years. But they have recently dropped the ball.
You seem to realize that about the Pentium IV; good systems with 800
memory are ridiculously expensive for the performance. The same is true,
though not as dramatically, for the other processors.
Bottom line: If you are spending your own money, then there is no
reason now to buy an Intel desktop. AMD rules (for now).
And I would say it depends on what
you are doing with the system. The cost of a video board is not trivial.
The Athlon systems available to me at the time I wrote that did not
include video, sound, and LAN on the motherboard; adding those makes a
significant cost addition. If you are setting out to build a really hot
games machine -- no bad reason to use an Athlon -- then the video board
will be the same for either machine, and one advantage of the Intel 815 is
washed out. I implied that but perhaps I should have said it more
explicitly.
These matters change like dreams,
and it may be that in a few weeks I'll have a different view. I'm
particularly interested in seeing the new Athlon motherboards. We'll see.
But for now, if it was up to me, I'd stay with Intel.
For a correction from a frequent
correspondent that may be significant:
Jerry,
I have what I believe is a correction to your latest
Byte column.
In it you state "Athlon changed all that, and
it's no longer possible to compare Intel and AMD chips directly."
I'm pretty sure _Intel_ changed all that - when they
shifted everything over (temporarily) to the slot-based CPUs in an attempt
to squeeze AMD and Cyrix out of the market. Once they had broken that
hardware compatibility, AMD had to come up with _something_ to succeed the
Socket7 format.
Is my take on this correct?
Sincerely,
Calvin Dodge
And I expect you are right, as
usual.
And finally a comment by Bob
Thompson on much of the above:
Actually, there's nothing wrong with the 810/810E
chipset at all, and the only problem with the 820 was when Intel tried to
support SDRAM via the Memory Translator Hub. That was a disaster, but used
with RDRAM as intended, the 820 is a solid chipset. VIA chipsets, in my
opinion, and that of many other observers, are and always have been a good
step behind Intel's in terms of stability, robustness, and performance.
The KT133/133A/266 are pretty decent chipsets, but they still aren't up to
Intel standards. Whether or not the chipset has support for
thermal/voltage sensing has nothing to do with anything. The problem is
that the AMD processors themselves have no thermal diode, so unless the
motherboard maker puts a thermal diode inside the socket itself, there's
nothing to sense. In the absence of a motherboard thermal diode, an AMD
processor can burn to a crisp without the chipset knowing anything is
wrong.
The KM133 boards will indeed provide a challenge to
Intel at the low-end, once those boards become more readily available.
Right now, Celerons sell for more than Durons simply because Intel can
afford to price them higher, knowing that there're no integrated
motherboards for the Duron. Once those become readily available, you can
expect to see pricing parity between Intel and AMD at the low end.
The chances of Duron "taking over the low end
of the market" are pretty slim. Corporations still buy Intel, even if
they have to pay a premium, and the mass market consumer grade PCs are
unlikely to shift to AMD if Intel can provide a similar solution at the
same price point. Relative performance actually counts for very little.
People buy MHz numbers, not performance, and if Intel has a 900 MHz
processor that's much slower than the AMD 900 MHz processor, it won't make
any difference to buyers. They'll regard them as equal because they're
both 900 MHz. That's why the Pentium 4 is such a deadly threat to AMD. It
may well be that a 1.4 or 1.5 GHz Pentium 4 is in fact slower than a 1.2
GHz Athlon, but buyers won't see it that way. They'll see only the slower
number for the AMD part and assume the Intel is better because it has a
higher number. And the Athlon is about out of room to grow in terms of
MHz, while the P4 is just getting started. I'd bet that AMD executives are
wondering now if they can re-introduce the old PR rating scale.
-- Robert Bruce Thompson thompson@ttgnet.com
http://www.ttgnet.com/rbt/thisweek.html
Note that my reservations on the 810
were with the original boards delivered to me which seemed to have been
cost-reduced to less than optimum reliability; later ones worked fine and
are the basis of the machine named "Seattle" here as well as
others. The video and sound are not as good as on the 815, but they ain't
bad.
And as Bob notes, Intel has lots of
room for maneuver. And I think this may be enough on this subject until I
hear something new. Perhaps at WINHEC
Subject continued below.
One interesting aspect of recent technology related
to paternity testing. In many places (30 some states was the most recent
number I saw) a young women giving birth at a hospital is once more
legally required to provide "father's" name. She may not decline
to do so under law. The named "young man" is then sent an
"invitation" to appear for paternity testing within 30 days.
Again, compliance is non-optional. Should the named person prove not to be
the father, the young woman is informed that she has apparently made a
mistake that must be corrected. At least a couple of jurisdictions are
holding repeated "mistakes" (in non-extenuating circumstances) a
criminal offense.
Why? It seems with the new welfare economy of the
'90's there's increased interest in the System not continuing to pick up
the full tab. The change from traditional insurance to HMOs is another
factor (who will pay long term?). Of course, there are also the growing
databases of DNA samples useful in locating all kinds of
"criminals" and no statute of limitations to appeal to . . . .
With April 15 fast approaching I must admit to a certain sympathy. Just
not sure I'm going to like the world we'll be living in . . .
Brian Belleville
I am not all that sure I will like
it either. On the other hand, making babies for the State to raise is a
serious matter. Perhaps it would be better for the State to bug out and
leave it all to charities; or perhaps the State could raise unwanted
children in creches and turn them into Janissaries; or --
What a brave new world.
And now for something completely
different:
Dr. Pournelle,
Shooting yourself in the foot with various
programming languages. Examples:
1. C++ You accidentally create a dozen instances of
yourself and shoot them all in the foot. Providing emergency medical
assistance is impossible since you can't tell which are bitwise copies and
which are just pointing at others and saying, "That's me, over
there." 2. Assembler A) You try to shoot yourself in the foot, onoly
to discover you must first invent the gun, the bullet, the trigger, and
your foot. B) You crash the OS and overwrite the root disk. The system
administrator arrives and shoots you in the foot. After a moment of
contemplation, the administrator shoots himself in the foot and then hops
around the room rabidly shooting at everyone in sight.
http://www.noncorporeal.com/people/pathfinder/shoot_yourself_in_the_foot.html
Don McArthur http://www.mcarthurweb.com
These stories have been around a
while and I love them all. This seems to be a link to the most complete
collection. Thanks!
Dear Jerry,
I hope you didn't put much effort into the font change, since on my
system, the difference is invisible between the two fonts (until you get
to 24pts with font explorer). To be honest I had to go check make sure I
had Giorgia installed, I've never noticed it before.
Besides your natural desire to tinker (which is why I read you, of
course), I for one can say I'd rather some little extra anecdote from your
vast experience than this kind of subtle graphics detailing. Keep up the
good work, I'll keep coming back anyway for the content. Best wishes,
James SIddall jr jsidaall@tiscalinet.it
Dear Jerry,
After taking a closer look at the HTML code in your current VIEW and MAIL
pages, I didn't even see a trace of any default font definition.
There aren't any CSS Cascading Style
Sheets or links that I can find, and neither Times New Roman nor Georgia
are cited as font specifications, so I think you may have set the default
for YOUR browser, not the HTML document. There are several pieces which
are defined using Arial or Traubuchet, which appear correctly in the
selected font. Your change either was only on your local machine, or
FrontPage didn't make it stick to the pages you published.
Best wishes,
James Siddall jr jsiddall@tiscalinet.it
In fact I don't know how to set it
so that Georgia becomes the default on other people's systems, I guess.
But I set the Front Page editor to have Georgia as the default font. I
then had to do that for my Internet Explorer as well. I think it looks a
little better, but there's not a lot of difference I suspect. As to time,
it took about 30 seconds.
Hi Jerry,
Perhaps you or one of your other readers could recommend to me a source
of high quality analog vga cables?
I have an electronic switch box that connects the keyboard, monitor and
mouse to my 4 computers sitting under my desk. Unfortunately, the vga
cables that I have are of low quality and significantly degrades the
quality of the video output above 800x600.
I would like to replace these as I prefer to use 1600x1200 resolution.
Any suggestions?
Thanks in advance.
- Paul pdwalker@quagmyre.com
Fry's sells Belkin cable sets and
they work fine for me. Cybex makes systems including cables that are
legendarily good.
Jerry:
Interesting article. When we are dumping our Pentium
III (retired to be test machines) and upgrade them machines to machines
with Athlon. You are promoting Pentium machines.
Frankly speaking, we think even Pentium IV can not
compare with Athlon in the near future.
Chang-Hsin Chang
P.S. One of my program takes 47 minutes to be
compiled on Penium III 400, but it only takes 17 minutes compiled on
Athlon 700 Mhz machine.
I think you make my point for me. 400/700 x 47 =
27, which presumably you would get by going from a 400 Intel to a 700
Intel. You say you get 17 in a highly CPU intensive task. That sounds like
a LOT better, and 10 minutes is 10 minutes, but not many people are going
to have 47 minute compilations to begin with.
400/1000 x 47 = 18, about what you get. I can
believe that for that task, a 700 Athlon is about equivalent to a 1000
Intel although that is better than I expected, and is about the best I
have ever heard.
Good luck with your changeover. You will probably
be happy with it, but do pay attention to power supply quality and
cooling. But you knew that.
As I have said, I can only relate my own
experiences.
Jim Allchin wasn't just commenting on the GPL out of
a theoretical or legal concern; he was commenting based on specific
brushes Microsoft has had with open-source software.
Microsoft took some heat when Windows 2000 came out
for its broken implementation of the Kerberos authentication spec. They
added Windows-specific extensions to Kerberos that conflicted with some
planned ones the core Kerberos maintainers were working on, damaging
interoperability and--in a very real way--hijacking the specification.
Impolite, but not uncommon in the open-source world.
Where this gets interesting is how they used
licensing to their advantage. Microsoft didn't provide documentation for
their changes to Kerberos. Unix administrators were left to scratch their
heads and wonder why their systems couldn't talk to Win2K systems
supposedly using Kerberos. Microsoft was put under pressure by security
organizations to publish documentation for what they did so that WIn2K's
security could be properly assessed.
Microsoft complied by providing skeletal
documentation for what they'd done, but did so in an interesting way. They
distributed it as an unlocked, un-watermarked Word document--and distributed
it under a clickwrap NDA. They made the docs unusually easy to obtain and
distribute, with contractual penalties if anyone used the spec to make an
interoperable system. By handing out an explanation of the algorithm and
data fields they used and claiming them as protected intellectual
property, they poisoned the well; any interoperable system would be
difficult to prove "clean", since it wasn't code but rather the
method that Microsoft was handing out to all comers.
After an even bigger uproar--from security bodies
and governments--Microsoft retreated and relaxed the terms and conditions
on use of the spec. But they were in the middle of an antitrust trial that
was going badly for them, so a characterization of what happened as a
"misunderstanding" comes across as disingenuous.
The point is--and make no mistake, this is where Allchin
is coming from--that this episode was only possible because Kerberos code
is released under a BSD-style license. MIT's code is released freely as
open source, but there is no GPL-style provision requiring that modified
versions or derivative works be open source. In other words, anyone can
grab the code, modify it, and distribute it as a closed, commercial
product if they wish.
Normally, this is largely harmless. The splintering
of BSD Unix, only now being rectified mostly because the commercial BSDs
are fading and the open-source ones are gaining and their developers are
working together on merging things back together, came out of this. But
for the most part, the BSD license doesn't lead to interoperability
problems because most companies don't want interoperability problems
themselves.
Unless you're Microsoft. With their market share and
degree of customer lock-in, they can distribute intentionally
"broken" versions of tools and still hold on to their customer
base because the cost of switching out is so high. And doing so only
further raises that cost, so it's to their advantage to do this.
If Kerberos were GPL'ed software, this wouldn't have
happened. Microsoft would have had to release their changes, without
use-restrictions, and would have had to release the source code for use by
all comers. Which may still have meant two conflicting "flavors"
of Kerberos, but it wouldn't have prevented anyone else from taking
Microsoft's changes and using them. If Microsoft wanted to shy away from
GPL code, well, their other option was reverse-engineering, just like he
rest of the world does when faed with Microsoft's own technologies. But
the DMCA, which Microsoft fought for, may put an end to that.
All of this means that Microsoft has a deep and
urgent interest in seeing that what open source software is out there s
licensed in such a way that it can be grabbed and pulled behind the
corporate curtain son that it can be turned into something proprietary and
exclusive. Imagine if, say, Berkeley, California made hundreds of bicycles
free for use by residents on a wide-open honor system and a local company
took the bicycles and brought them to its office campus, where only its
employees could get to them. Under the terms the city laid out, no theft
has taken place, but the company in question isn't playing well with
others.
You start to see the wisdom in the GPL, especially
when it's applied to low-level "plumbing" software. See also
flareups of controversy regarding Microsoft's DNS server (based on BIND),
its FTP client, and other Unix-world technologies they've rolled into
Windows NT and Windows 2000 over the past few years. It's the same story.
And one more thing: your remark about Ada wasn't all
that well informed, or maybe I'm not getting your point. When the GPL
covers a program, it only covers the code for the program itself and
things that incorporate the program's own code, not things created through
use of the program. There are quite a few compilers out there under the
GPL, not the least of them the ubiquitous GCC compiler and its accompanying
Fortran, Pascal, C++ and Ada(!) compilers. Programs built with it aren't
in any way bound under the GPL unless they contain part of the compiler itself.
Some tehnologies--some fairly hefty ones--are
already developed by the US government under the GPL. Take the Beowulf
clustering technology for Linux. It was developed by NASA, and since it's
made up largely of kernel modifications, it had to be released under the
GPL. A few thousand supercomputers later, the world hasn't come to an end
and NASA hasn't moved to ban Linux.
You probably wouldn't want the US military to
develop missile guidance software under the GPL. But why would it? And why
wouldn't they use GPL'ed tools to do it if they wanted to? Not everything
should be written under the GPL. But plenty of it can, with benign results
to all but a few boards of directors.
-- Steve Koppelman steve@haless.com
Thank you. Well said, and well reasoned. Roland
Dobbins more or less agrees. I have my own view as to why Allchin said
what he did, and perhaps I didn't say all of it in the article, but the
reality is that I think no harm was done and perhaps some good was: a
trial balloon was launched, it was shot down, and perhaps we are all
better off allowing it to pass forgotten. Microsoft certainly got The Word
following Allchin's remarks and that is the important thing.
As to the ADA example you certainly missed the
point I intended to make, and thus clearly the example was not well
chosen. You are welcome to add your own.
There is a tension between the commercial and the
free parts of this community, and there ought to be. If we are going to
take advantage of the new hardware capabilities we need companies that can
afford to invest in plumbing. We also need the insightful people who do
things because they want to do them. Sometimes those coincide, often they
don't.
My own view is that we won't make real progress
until we have languages that can be learned incidentally as part of an
education, so that physicists and chemists and social scientists and
operations researchers can write programs as they write models. When I was
at Aerospace Corp. I could model nuclear exchanges, and while I knew some
FORTRAN I didn't know enough to do it very well; fortunately the company
had programmers who could work with me to get them running, and I could
fight my nuclear wars (bloodlessly) several times a week with different
components of the US Strategic Offensive Forces, and look into the
relative merits of spending resources to increase reliability vs.
survivability vs. simply adding to the force. It's that kind of casual use
of computers I have in mind when I say "real progress" and
"new languages."
And that may well come from GPL. In any event
thank you for the informative letter. For more on
this see below.
Hi Gerry,
Sorry to be contentious but it wasn't AMD or its Athlon that changed
all that. Intel put an end to CPU interchangeability when it broke away
from the Socket 7 design on which it held no patent to a new (Slot-1)
which it could patent. Intel did this in order to kill AMD but was
spectacularly unsuccessful. The move to Slot 1 ended all chance of
Intel/AMD compatibility for the foreseeable future. And now the
introduction of the P4 ends backward compatibility with the P3....way to
go Intel
Thanks for putting out your column, I always enjoy it.
Please call me if you need more info.
Regards, AH
Alan Hart (Technical Sales)
I expect you are right; what I meant and should
have said was "the advent of" changed all that. I wasn't really
trying to assign praise or blame.
As to the backward compatibility of P4 and P III,
we will see what success that experiment has. As I have said, unless you
know you do there is no reason to go to P 4 at all. (That is if you need
to you know it and know why; if you have no idea about why you ought to
you have no reason to.)
As I've said many times, for what most of us do,
you can't tell the difference between a high speed Celeron and a 1.5 GHz
Athlon or a P4; they all work fast enough to get most of our work done. If
you have 50 minute compilations then you have a need to look into the
fastest hardware; but for most of us the best improvement you can make is
to add memory, get rid of Windows ME for either W 2000 Pro or W 98 SE, and
only then think about a faster processor.
Former BYTE editorial
colleague Russell Kay says
Joseph Pfeiffer wrote: I read your last column on
Celeron processours and heatsinks. I'm wondering about the question: Are
there possibilities to have completely quiet computers? Without noisy
fans, maybe with a watercooling mechanism? Have you ever heard of this?
There are quiet machines; see my article,
"Selecting PC Components," http://www.computerworld.com/cwi/story/0,1199,NAV47_STO54941,00.html
This was originally going to be all about quiet PCs,
but I had to change the spin on it some, so the stuff about noise is
mostly buried in the middle. Still, the Silent PC case is quite special.
Russell Kay Sr. Reviews Editor, Computerworld 500
Old Connecticut Path, Framingham, MA 01701 508-820-8175 fax: 508-875-8931
main: 508-879-0700 russell_kay@computerworld.com
Thanks. In the old days we'd have got together
before I replied to that letter. On the other hand in the old days neither
the letter nor the replies would be in print yet...
And of course I have many letters pointing out
that the new Macs don't have fans. Neither did the 128K Mac which needed
one (and you ended up putting an external fan on top of it, at least if
you lived in Los Angeles). I probably need to get back to Macs, but David
Em and my son Alex do that for the high end graphics world, so that's one
area I don't have to cover; and for good or ill the variety of software
and peripherals for Macs is just not large compared to the Wintel world --
which is itself too large for me to cover or even stay completely informed
about.
Dear
Dr. Pournelle,
Sorry to hear you're getting flames over the AMD vs.
Intel debaters. Some folks seem to have entirely too much time and energy
on their hands...imagine what they could achieve if they channeled it into
productive avenues!
About Fergie:
Doesn't Win2K pre-allocate the swap file? WinNT
does. That would keep it contiguous, if it was created at installation
time and not subsequently messed about with. Of course, even then it can
grow, and the newly allocated space can be fragmented to heck and back.
I don't know if Diskeeper has changed, but as of a
couple of years ago it only defragged files, not the filesystem as a
whole. So, even when it can get each file nicely contiguous, there'd be
empty blocks between files such that new files would instantly be
scattered about the disk. SpinRite used to defragment both files and the
filesystem, but I don't think it runs on NTFS.
Another idea would be to access the C: drive from
another box, to see if you could access files normally. WinNT always had
"administrative" shares for each fixed disk (e.g. C: was shared
as "C$", a hidden share). If you haven't reformatted, you might
want to give that a whirl, just to see if it's the GUI part that's hosed
or the lower-level parts of the OS.
Just tossing my US$0.02 into the Intel motherboard
debate, I haven't tried the 815 series, but I have a rackful of 810-based
systems, which are configured as Linux servers for web, database, Java app
server, and one WinNT/SQL Server. The Linux boxes have Celeron 466s and
128 MB RAM, the WinNT box has a PIII-800 and 256 MB of 133MHz RAM. All run
fine, and I couldn't be happier.
It will take the next round of MSFT bloatware to get
people to upgrade again, the stuff that's out now is plenty fast for the
majority of work.
How's Sasha doing? Hope he's well.
Gordon.
-- It doesn't get any easier, you just go faster. --
Greg LeMond
I never pay much attention to email that has
clearly been generated by people who didn't read what I said. A few did
and responded rationally and I have posted some of that; some have good
points. Apparently there are some boards out now that I had least heard
were just coming out. But the last time AMD made any contact with me about
boards and systems none of that was mentioned.
As I said in the column, I am not concerned about
high end games; if you want the screaming latest you don't read me to
begin with. Nor do I much care about dhrystones whatever those are; I do
care about reliability and confidence and "good enough" in the
speed, video, and sound departments, and I am willing to pay a bit extra
to have those.
I have mail telling me that Intel is a has been,
doomed, sell all your stock (of course I don't own any) and then berating
me for being childish and inaccurate. I have a folder for such mail...
Sasha is fine. We walked down to the district
today. He thinks we should go up the hill but the vet says not until the
end of the week.
And your final point is the one I was trying to
make: it hardly matters what systems most of us use now. They are all fast
enough, and what we want to to stop thinking about them. And for better or
worse I had enough problems with VIA chipsets a couple of years ago that I
hesitate to recommend them now. Perhaps that leopard has changed his
spots. Some do. And I will be glad to look at the new stuff and get my
scope out, and go on reporting what I find. Which is all I can do.
Dear Jerry:
As you say, "this is getting silly."
Arguing over which is a better (and what does that mean? Faster? More
reliable? Less costly?) processor, Intel or AMD, reminds me too much of
the old Chevy versus Ford debates that raged endlessly in my youth. Those
owners of yore, most of them, had very good reasons for their deeply held
belief, but by their passionate allegiance to one brand or the other the
simple fact that whichever car was in the garage got them to work for the
most part safely and reliably, let them take the family for a Sunday drive
in comfort and security, and expanded both their horizons and their
freedom got lost.
Competition is a good thing. Beyond that, six of one
and half a dozen of the other; may it be ever so (and in Redmond, too).
All the best,
Tim Loeb
Amen. Amen. Amen. And Good Enough remains Good
Enough. Why people who have not read what I said take the trouble to send
me hate mail denouncing things I didn't say is beyond me, but what the
heck.
TOP
|
CLICK HERE for
a Special Request. 
