On Wed, Dec 16, 1998 at 03:50:07PM -0500, Chris Mauritz wrote:
>
> I'm sure with enough hacking, one can get a pair of 386's running in SMP
> mode. Does that mean that they need to bother everyone on the list about
> it? How many people do you think want (let's forget about "can" for a
> minute) to bother physically hacking their processor boards to do this?
> I suspect scant few.
I don't know how hard this would be, nor how many people
have the equipment, time and expertise. But If someone
manages to do this, I think that I'd like to know about
it. In a hack value measure, this would be a precious gem.
> Again, when you consider the amount of time to get this type of system
> running, it makes little (if any) economic sense.
By this measure, I believe that it is arguable whether
the development of Linux made economic sense before its
recent surge in popularity. How much effort went into
the development of what is now Debian 2.x or RedHat 5.x?
How many years did it take? How many developers? For what
proportion of that development time was it clear that there
was a net economic benefit to the effort? Is it *yet* clear
that the economic benefit of this development effort solidly
outweighs the time, expense and expert mindshare that this
effort consumed? How do you measure this, and why on Earth
did the first Linux developers strike out on this path? What
else could have been produced with those resouces? [A rhetorical
question if ever there was one...]
I can think of a few reasons to allow this sort of discussion
on this list:
1. Most readers find it interesting. (Note that I say that
this would be a reason to allow it, not that I know that
this criteria is met.)
2. When people like Dave tell us of these exploits, it
is a big advantange to the real SMP developers because
they then know who's trouble reports to ignore.
3. Conversely, if people know that they would be
flamed for confessing to overclocking, you may have
an increase in trouble reports from people who
will not admit that they are doing this. At least,
I think that this is at least as likely an outcome
as a decrease in the number of people overclocking
because they might be flamed for it.
I can also think of a few reasons not to allow it:
1. Discussion in which Linux plays only an incedental
role seems not to be in the charter of the list.
What Dave has done could as well have been done
with Solaris X86, Netware or, God forbid, NT as
the OS. The OS is irrelevent to his results.
2. His results may be irreproducable. The success of
such an effort will likely be highly variable,
depending on the individual product samples as
well as local environmental conditions such as
power stability. Thus these reports are anecdotal,
and may not have much scientific value.
3. Some who do contribute valuable expertise to this
list may pay less attention, or take their expertise
elsewhere.
Personally, having used the Internet for over a dozen years
and Usenet from when it came in on uucp over 1200 baud
modems, I firmly believe that more discussion is generally
better. Any decent mail reader today has wonderfully
sophisticated sorting and filtering capabilities. I'm
on about 20 mailing lists and generally get two to three
hundred messages per day. If I got bent out of shape every
time a discussion wandered off topic, I'd be a pretty
stressed out fellow. I generally don't even go to much
effort to filter spam because it is generally so easy to
spot it and ignore it. My advice would be: Chill out, note
the subject line and hit the delete button if it doesn't
interest you. All this stress over an errant thread is a
tremendous waste of emotional energy.
This having been said, I will also add that I think that
people who overclock today are the people who would have
been hot-rodding stock cars had they lived in the 1960s.
The similarities between the two activities are several,
right down to the fact that the manufacturers would get
*very* testy if a hot-rodder ever had the gall to bring
a modified car in for warranty service.
Now if you recall, much as the car makers detested
the hot rodders, they also learned a great deal from
them, and some of what they learned eventually found
its way into factory products (Holly carburators come
to mind). As for overclocking, one can note that IBM's
G5 Mainframes are shipping "overclocked"...
------------------------------------------------------------
(My apologies to Computerwire for pasting this in... I'll
mention http://www.computerwire.com/ as an execellent
source of information on the computer industry in the way
of a quid pro quo...)
------------------------------------------------------------
: ------------------------------------------------------------
: Computergram International
: Number 3462 - 07/29/98
: Section: Top Stories
:
: IBM BOOSTS G5 PERFORMANCE USING CHIP-COOLING
:
: IBM Corp is using Kryotech Inc's chip-cooling technology
: and other enhancements to wring additional performance from
: the Symphony CMOS processors which drive its forthcoming
: S/390 Turbo G5 mainframes. News that a high-end ten-way G5
: server will perform 1,040 MIPS versus the 900 MIPS IBM had
: originally planned is sure to light a fire under Big Blue's
: chief mainframe rival Hitachi Ltd. A high-end uniprocessor
: G5 Turbo engine now performs at 150 MIPS - up from 125
: MIPS - the same as Hitachi's current high-end bi-CMOS
: Skyline engine which does some 975 MIPS as a ten-way. The
: standard G5 CPU, formerly set to deliver 115 MIPS, now
: performs 125 MIPS. The improvement has been achieved in
: part by raising the 32-bit chips' clock speed from 465MHz
: to 500MHz, and by other techniques which IBM wasn't very
: specific about, such as more data paths to main memory and
: increased voltage. One way to increase clock speed is to
: operate the chip at a lower temperature. IBM eventually
: admitted it is using NCR Corp spin-out Kryotech's
: refrigeration technique as well as technology from
: computer air conditioning specialist Liebert Corp. West
: Columbia, South Carolina-based Kryotech - which supercools
: Alpha, AMD, UltraSparc, Pentium Pro and other chips -
: is manufacturing the cooling compressor units which are
: used in G5 while IBM's own engineering team designed the
: control unit. Three Model 9672 G5s configured with the 150
: MIPS Turbo engine use eight, nine and ten processors and
: are due to ship in September. The other 12 models use a 125
: MIPS engine, previously rated at 115 MIPS. They are due to
: ship in the third week in August. IBM says it will have a
: dozen installations by the time the servers are generally
: available; first sites went live over July 4 weekend. A
: G5 (Symphony) multichip module (MCM), the core of every
: new 9672 server, includes twenty-nine separate chips on
: a 127mm-square multi-chip module: 12 G5 CISC CPUs - each
: with 256Kb cache - two system controllers; eight 1Mb L2
: cache chips; four memory bus adapters; one clock chip;
: and two cryptographic processors. Ten of the G5 chips can
: be used as CPUs in an SMP single image system; the two
: others in the MCM can be either used as hot backup in the
: event of a chip failure or as 3990-style disk controller
: (IBM ported the 3990 controller program to the S/390 CMOS
: chip family during the G4 generation). G5 uses one fifth
: the electricity and one tenth the glue chips required in
: IBM's H5's bipolar thermal conduction module. Upgrading
: to the G5 CPU and memory boards from G4 takes around four
: hours, it claims. IBM said it would not provide list prices
: for the systems even though the 1956 Consent Decree, which
: IBM signed to put a stop to an antitrust lawsuit brought
: against it by the US Department of Justice, is still in
: effect for the AS/400 and S/390 lines and requires IBM to
: release list prices for all of the equipment and software
: it sells under these lines (see separate story). It told us
: to consult an analyst. Expect prices to be around $6,000
: per MIPS for complete G5 systems, down from $7,000 per
: MIPS for G4 servers, according to analysts. As reported,
: IBM's second quarter mainframe business was down 25%
: on MIPS deliveries that were flat as the customers await
: G5. IBM's Wall Street watchers say it already has some
: 140 customers booked to swap out Hitachi or Amdahl Corp
: mainframes with G5. IBM's mainframe hardware business
: is estimated to be worth some $5bn- $18bn once software,
: services and maintenance are included. Analysts caution
: users to beware of costly new software licenses for G5
: (see separate story).
:
: � ComputerWire Inc, 1997/1998.
: ------------------------------------------------------------
--Bob
--
============================================================
Bob Drzyzgula It's not a problem
[EMAIL PROTECTED] until something bad happens
============================================================
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
