🚀 go-pugleaf

From: mikerios@ix.netcom.com (Michael Rios )
Subject: Re: Cheap speed boost
To:  smiteq@iccu6.ipswichcity.qld.gov.au (Enter your name here ???)
You wrote: 
>
>mikerios@ix.netcom.com (Michael Rios ) wrote:

>
>>Most modern boards allow you to set the DMA clock independently
>OK... TELL ME PLEASE HOW YOU DO THIS WITHOUT SPECIAL SOFTWARE.

If your board can do it, it will be in the Advanced Chipset setup. 
Occasionally it 
will be in the Advanced CMOS setup.  You get to either in the same way
that you would 
get to the standard CMOS setup, usually by pressing a particular key
when you first 
turn the system on.

>>When you put a heat-sink/fan on them, they run *way* cooler than
>>factory spec, even if overclocked.
>I AGREE BUT THE ORIGINAL ARTICLE ATTEMPTED TO PLAY DOWN ANY COSTS
>OTHER THAN FOR THE CRYSTAL... FANS ARE NOT FREE WITH CRYSTALS.

We sell them for $7.  How close to free do you want?

BTW, if your board uses crystals, it is *not* very modern.  Virtually
all modern 
boards use a programmable frequency generator with clock speeds that
are jumper 
selectable.

>>Not even close.  Chips marked with a lower speed in most cases were
>>never tested at higher speeds.  
>SO TELL US PLEASE WHERE YOU GOT THIS INFORMATION... MINE CAME FROM
>INTEL IN 1989 AND FROM AMD IN 1993.

I'm not sure what information you want.  By 1993, virtually all the SX
CPU's that 
Intel made, regardless of marking would run at least at 33, and all but
a few would 
run at 40 with a fan/heatsink.

All you have to do, tho, is try it, and then run tests for a day or so.
 If it doesn't 
creash during that time, you have "recertified" it at a higher level
than Intel ever 
does.
>
>>  Of course, we knew that; we'd been running
>>25's and 33's at 40 MHz for years!  In all that time, we had three
>>-25's and one -33 that wouldn't clock up to 40 stably.  And we've
never
>>had one fail in the field.  (we're talking hundreds, maybe thousands
of
>>units).
>WHO IS WE? 

Computer Care, Inc., Arlington, VA, USA

>IN QUEENSLAND THERE WAS A COMPANY CALLED DATAQUIP WHO WERE FORCED OUT
>OF BUSINESS IN 1995 BY THE CONSUMER AFFAIRS DEPARTMENT AND MUCH TV
>PUBLICITY FOR SUBSTITUTING 25 AND 50 MHz CPU's FOR 33 AND 66 MHz ONES.

Clearly a case of government regulatory & media excess; it probably
should have been 
handled with something like a consent decree.  The customers, almost
certainly, were 
not harmed in any way, unless Dataquip wasn't running them at the
higher speed, or 
unless they were absolutely incompetent about testing & recertifying
them.  It 
probably depends on how they worded their ads & literature.  
  
In our case, we always make it clear that we are using an overclocked
chip.  Since we 
are selling the CPU at a *faster* speed than Intel rates *any* chip
for, it is obvious 
that we are not *substituting* a cheaper chip.   Intel never made a
"486sx40".


>HAVE YOU BEEN DOING THIS TOO?

Of course.  We *brag* about it, because it is in the best interests of
our customers.  
We can offer them computers that are engineered for the best bang for
the buck, 
without market considerations skewing the engineering.  We do almost no
advertising; 
we depend on word-of-mouth.  The local non-profit consumer magazine
just gave us their 
top rating.


>> This is to say all chips have a manufactured 
>>>: "speed target".
>
>>Nope.  They make them, & then see how many pass at the speed they
need
>>to fill existing orders.
>CAN YOU SUBSTANCIATE THIS STATMENT?

Sure.  Call & ask Intel, or any other chip mfr.  This is common
industry practice for 
*all* speed-rated devices, including RAM, EPROM, CPU's, etc.
If you think about it, no other approach makes economic sense.  Why
bother testing 
chips at speeds higher than necessary to fill orders?  It would
decrease your yield, 
however slightly, and be more hassle.  Many purchasing agents are not
very technical, 
and if they order -25's, they'll reject -33's.

>>this would apply to a very small # of chips.  The vast majority are
>>never *tested* at a speed higher than the mark.
>
>>Think of the speed mark as a warranty, not as a technical spec.  If
it
>>doesn't pass that speed they'll replace it.
>
>
>> and those 33s with a faulty FPU might perform as a 25Mhz SX.
>
>>Not anymore.  Witness the slaughter of the early Pentiums, which have
>>been ground to dust.  Why didn't they just mark them as "SX", and
sell
>>them at half price?
>
>BECAUSE PENTIUM BOARD MAKERS NEVER DESIGNED IN A "TURBO" SWITCH OR
>FACILITIES FOR DISABLING THE FPU... THAT'S WHY AND THIS INFORMATION
>CAME FROM INTEL TOO!

A "turbo switch" would be irrelevant to the SX issue.  There was no
need to disable 
the FPU, because the floating point error was so trivial, that it would
never have 
caused an error for non-scientific uses.  Remember, floating point is
only active if a 
program *specifically* calls it; word processors, databases,
communications, even most 
graphics programs don't use it (CAD, yes; non-CAD graphics, no).  Most
spreadsheets 
won't use it, either,  because if they are money-denominated, they use
fixed point 
calcs.  And simple, non-recursive calcs, even in floating point, won't
show an error 
until the 9th decimal place or beyond.  Most of us round at a lower
precision than 
that.  So they just had to mark them "Non-scientific use only" or some
such; I'd use 
one happily, if it saved me $100 or so.  This again was a marketing
decision, plain & 
simple; it was an embarassment, & they wanted to get it out of the
public mind as fast 
as possible.  Literally, destroying the evidence.