Searching \ for 'just how fast is a PIC?' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: www.piclist.com/techref/microchip/devices.htm?key=pic
Search entire site for: 'just how fast is a PIC?'.

Truncated match.
PICList Thread
'just how fast is a PIC?'
1999\09\16@014816 by tec

flavicon
face
How fast is a PIC???  My understanding is that the PIC's clock speed is
very under rated.  Has anybody had success at overclocking a PIC (
>20Mhz on 16CXX or >10Mhz on 16F84)?  If so, what frequency range,
operating temperature, exact part (including die shrink).  Is there any
charts or test results available (even if they are not blessed)?

Todd.

1999\09\16@015243 by Dave VanHorn

flavicon
face
> How fast is a PIC???  My understanding is that the PIC's clock speed is
> very under rated.  Has anybody had success at overclocking a PIC (
> >20Mhz on 16CXX or >10Mhz on 16F84)?  If so, what frequency range,
> operating temperature, exact part (including die shrink).  Is there any
> charts or test results available (even if they are not blessed)?
>
> Todd.


You realize of course that this kind of thing is only useful in hobby
projects.
Overclocking a micro in a commercial design, is a career-ending move.
Or it will be, the first time you get a batch of chips that dosen't go as
far beyond the specs as you'd like.

1999\09\16@023022 by tec

flavicon
face
Dave VanHorn wrote:

> > How fast is a PIC???  My understanding is that the PIC's clock speed is
> > very under rated.  Has anybody had success at overclocking a PIC (
> > >20Mhz on 16CXX or >10Mhz on 16F84)?  If so, what frequency range,
> > operating temperature, exact part (including die shrink).  Is there any
> > charts or test results available (even if they are not blessed)?
> >
> > Todd.
>
> You realize of course that this kind of thing is only useful in hobby
> projects.
> Overclocking a micro in a commercial design, is a career-ending move.
> Or it will be, the first time you get a batch of chips that dosen't go as
> far beyond the specs as you'd like.

That depends on the application and the customer.  Is the customer aware of
the risks and are they willing to take them?  For example, a 16C74-04 PIC
will easily run at 20Mhz based apon my experience.  It is the same die for a
-04 and a -20.  My customer is aware and has accepted the risk of buying -04
parts and running them at max published die speed and it is not my
responsiblity as I specified the -20.  In the past 5 years they have never
had a problem.  I would question whether the testing is different or if it is
just the price.  The cost for the distributors to inventory all these
'different' flavors of the same part (die) has got to be more that the cost
of the fallout between a -04 and a -20.  Why not label all the parts at max
published die speed, cost reduce the products by inventory reduction, and
gain a marketing advantage over the competitors?  (This allocation would be a
good time to let the inventory on slow parts sell before switching to all
high speed parts.)

What I would consider overclocking a PIC is anything beyond the max published
die speed. In the case of the 16C74A this would be anything over 20Mhz and in
the case of the 16F84A it would be anything over 10Mhz.  So to clarify, any
clock rate over these speeds I would consider overclocking.  Futhermore,  I
fully agree with you that  to design a circuit using clocks greater than the
max published clock speed would be career suicide, but my curiousity still
asks: How fast can a PIC go beyond the specs.? (not that I'd ever run it at
that speed in a commercial design)

Todd

1999\09\16@032013 by Dag Bakken

flavicon
face
Your question here was how fast a PIC can go beyond spec.  Well, you
run -04 at 20MHz.  That's 5 times beyond spec.  Considering this, PICs
are quite fast.  Even if -04 and -20 are the same die, perhaps a -04
only passes 98% of the tests at 20MHz.  I don't know if that may be a
peripheral that would show periodical flaws or whatever.
Either way, you will see more and more parts failing as speed
increases beyond spec.  There has been made software designed to run
on a 16C54 at 70MHz.  Only one of four parts could do it, and they
were quite hot.

-DS
Speed Kills! Use Windows 95.


Thursday, September 16, 1999, 8:29:21 AM, you wrote:

TC> had a problem.  I would question whether the testing is different
TC> or if it is just the price.  The cost for the distributors to
TC> inventory all these 'different' flavors of the same part (die) has
TC> got to be more that the cost of the fallout between a -04 and a
TC> -20.  Why not label all the parts at max published die speed, cost
TC> reduce the products by inventory reduction, and

TC> fully agree with you that  to design a circuit using clocks
TC> greater than the max published clock speed would be career
TC> suicide, but my curiousity still asks: How fast can a PIC go
TC> beyond the specs.? (not that I'd ever run it at that speed in a
TC> commercial design)

TC> Todd

1999\09\16@065626 by Rob Bakker

flavicon
picon face
Heh heh, well this topic made me smile....

- while building my first ever Pic (PIC16F84-04) project some months ago, I
was hunting around in the scrap pile looking for a crystal to clock it with.
I found lots of 'canned' crystal oscillators on some old 486 video cards -
saw one marked 12000kHz and in my eagerness, thought 'Cool, 1.2Megs, so
it'll run a little slow, but should be ok to make a LED flash'

well, a couple of months later on I looked closer at that oscillator and was
quite amazed!  its still going, doing the knight rider thing with port b and
some LEDs...

that's 3 times overclocked! I wouldn't recommend this to any one who has a
commercial venture in mind however...

too dang eager I guess

RoB>)

> How fast is a PIC???  My understanding is that the PIC's clock speed is
> very under rated.  Has anybody had success at overclocking a PIC (
> >20Mhz on 16CXX or >10Mhz on 16F84)?  If so, what frequency range,
> operating temperature, exact part (including die shrink).  Is there any
> charts or test results available (even if they are not blessed)?
>
> Todd.
>

1999\09\16@090935 by tec

flavicon
face
Rob Bakker wrote:

> Heh heh, well this topic made me smile....
>
> - while building my first ever Pic (PIC16F84-04) project some months ago, I
> was hunting around in the scrap pile looking for a crystal to clock it with.
> I found lots of 'canned' crystal oscillators on some old 486 video cards -
> saw one marked 12000kHz and in my eagerness, thought 'Cool, 1.2Megs, so
> it'll run a little slow, but should be ok to make a LED flash'
>
> well, a couple of months later on I looked closer at that oscillator and was
> quite amazed!  its still going, doing the knight rider thing with port b and
> some LEDs...
>
> that's 3 times overclocked! I wouldn't recommend this to any one who has a
> commercial venture in mind however...
>

I wouldn't consider that 3 times overclocked.  Even though the part is marked as
4MHz  that same silicon is rated at 10MHz just not tested at 10MHz (at least
that's what we're told).  Therefore, you're 12MHz clock is only over clocking
the part by 20%.  I've run 11.0592MHz crystals on the 16F84 with no problems
while waiting for 10MHz crystals to show up (no I wouldn't recommend it for
production).

I still can't imagine running separate tests on the die.  Why not just test at
10MhHz in the case of the 16F84?  Wouldn't it be cheaper to have one test on the
silicon?

{Quote hidden}

1999\09\16@094233 by bowman

flavicon
face
Todd Conard wrote:
>
> had a problem.  I would question whether the testing is different or if it is
> just the price.  The cost for the distributors to inventory all these
> 'different' flavors of the same part (die) has got to be more that the cost
> of the fallout between a -04 and a -20.  Why not label all the parts at max
> published die speed, cost reduce the products by inventory reduction, and
> gain a marketing advantage over the competitors?


I don't know the  particulars of MicroChips manufacturing, but in many
cases, lower spec'ed parts are the fallout from testing. For instance,
in the old composition resistor lines, one  didn't try to make 5% and
10% resistors; you just measured them and marked them accordingly. It's
possible a batch of chips that fall out of the 20 mhz tests become 4 mhz
parts.
--
Bear Technology  Making Montana safe for Grizzlies

http://people.montana.com/~bowman/

1999\09\16@113205 by Wagner Lipnharski

picon face
When I bought my actual car, it came with expensive Goodyear Eagle GT
tires, cost more than $160 each, and I actually needed to exchange 3 of
them due nails and screws at the street (they were building up houses in
that street... arggggg).  Those tires aren't needed in Orlando, they
offer a nice stability to drive (with 30psi they look like to be almost
flat), great performances in sharp curves at 55mph, but... what else?
They stand speeds up to 150mph I think, but I can't drive at 90 or
100mph as I would like with those tires, except if I can afford those
nasty speed tickets...

So, why not change for cheap $50 tires?  They are 100% enough to run at
max of 55mph in Orlando, flat topology, but what about safety? what
about rain? would they work nice with ABS?  Of course I could use them
without much worries, isn't 90% of cars around using them?  Well, In 5
years I was forced to a dozen of situations where very hard break were
needed, as well very quick lane change in a split of a second, even
under rain or wet asphalt... would a regular $50 tires respond in such
efficiency in those situations? or my car would be involved in another
"car accident" as I can see a couple daily around?

Of course those cheap tires are also over-rated, they stand 70mph? but
in real they go up to 100mph, isn't? but just because this extra margin
of safety would you drive at 99mph during 10 uninterrupted hours with
them? Of course not.

When selecting a resistor you always double its power size, right?  An
electrolytic capacitor is bought at least with 50% higher voltage than
the application, right?  A power transistor is always selected with 100%
or more standing current and power dissipation. If you do this for
components why sacrifice and suck juice from the main component? just
because a couple of dollars?

By the same reason I would not expose my reputation, board and
components expenses, just to save a couple of dollars and use a 4MHz
chip running at 10MHz, not even if the customer ask me to do that.  It
still my responsibility as a professional to "not only advise" the
customer, but to make him understand *why he can not* and *why he will
not* do that. By the same reason he can find around some "professional"
that can do that for him, he also can find a "cheap" $200 brain surgeon,
but ... will he use it?

Remember that *Quality* can be translated to "Perceived Quality".  The
honorable cheap surgeon can operate by few dollars, but the instruments
used still needing care and it cost money.  The operation can be a
success, but for how long?  

You can save more than a couple of dollars using another processor
manufacturer.  Just guess who the customer will blame when its 2 dollars
savings result in a inoperative device, even that he knows about the
risks... and a nice 20MHz Atmel device could make it work for 20 years,
costing $4 less than the stressed and overclocked part.

just my 2¸ cents.

Wagner.

1999\09\16@122351 by Harold M Hallikainen

picon face
On Thu, 16 Sep 1999 09:07:56 -0400 Todd Conard <spam_OUTtecTakeThisOuTspamSCANTOUCH.COM>
writes:

>
>I still can't imagine running separate tests on the die.  Why not just
>test at
>10MhHz in the case of the 16F84?  Wouldn't it be cheaper to have one
>test on the
>silicon?
>

       I suspect that the speed suffix is determined in a die sort of
parts that are SUPPOSED to be identical.  For example, I suspect all
16c74b chips are tested at 20 MHz (possibly higher to give some sort of
safety margin).  Those that pass are marked -20.  Those that fail are
tested at 10 MHz.  Those that pass this test are marked -10.  Those that
fail this test are tested at 4 MHz.  Those that pass this test are marked
-4.  The rest are discarded (I think the L series for really low
frequency and power may be a different die).
       By doing this die sort, they get an increased yield over just
testing everything at 20 MHz and throwing away those that don't make it.
Eventually we may see their process improve to the point where they all
pass the 20 MHz test.  At that point, we'd see the price on the -20 part
drop to that of the -4 part.  They would continue to make -4 parts, using
parts that pass the 20 MHz test to fill orders.  That way we
manufacturers don't have to change our bill of materials; we can just
keep buying 16c74b-4 .
       I suspect this sort of die sorting is very common in the
semiconductor industry.  I recall years ago designing a circuit with the
then new NE5534 op amp.  The A version had lower noise and higher price.
I'm sure this was just a die sort during test.

Harold


Harold Hallikainen
.....haroldKILLspamspam@spam@hallikainen.com
Hallikainen & Friends, Inc.
See the FCC Rules at http://hallikainen.com/FccRules and comments filed
in LPFM proceeding at http://hallikainen.com/lpfm

___________________________________________________________________
Get the Internet just the way you want it.
Free software, free e-mail, and free Internet access for a month!
Try Juno Web: dl.http://www.juno.com/dynoget/tagj.

1999\09\16@130745 by Dave VanHorn

flavicon
face
> You can save more than a couple of dollars using another processor
> manufacturer.  Just guess who the customer will blame when its 2 dollars
> savings result in a inoperative device, even that he knows about the
> risks... and a nice 20MHz Atmel device could make it work for 20 years,
> costing $4 less than the stressed and overclocked part.
>
> just my 2= cents.
>
> Wagner.


Sounds like you've been there too :)

I've always wondered how you assure that in the overclocked state, your part
will always do everything you need it to do, under all circumstances.  I'd
hate to tell BofA that we lost a few million in credit card charges because
well, you see, we were running these chips a little faster than spec, and
when this certain pattern is in the ram......  I think it would be a bad
scene.

My one experience with overclocking a single chipper was taking over a
design from another engineer. I discovered that he had a 12 MHz zilog part
(1 MIPS) clocked to 12.288, and some were failing to run.  In the end, I
totally re-wrote the software, and was able to run their slowest grade (8
MHz) at 3.579 MHz and do the same job.

Believe it or not, I uncovered a bug that caused the code to fail when I
re-wrote one module and speeded it up. The bug required the code I was
working on not to complete too quickly, or the remainder of the incoming
data would be missed!

I run both my PCs overclocked mildly, but it's only a few keyboard ops to
de-clock them, should it ever become necessary.

1999\09\16@211731 by D. Schouten

picon face
> You can save more than a couple of dollars using another processor
> manufacturer.  Just guess who the customer will blame when its 2
dollars
> savings result in a inoperative device, even that he knows about the
> risks... and a nice 20MHz Atmel device could make it work for 20
years,
> costing $4 less than the stressed and overclocked part.

I my case after a development of 8 months it turned out that to keep
everything in sync we had to increase the clock speed of our 16C73A/B
from 20MHz to 20.48MHz. It had to do with an earlier made decision for
a certain number of A/D samples (5k samples per second) and the amount
of time between these samples to proces the data with lot's of
multiply and
divide operations.
At that time we didn't had the time to rewrite the entire code to make
it less demanding in order to lower the clock speed a bit. We weren't
aware of the Atmel CPUs either, however if we did, we still had the
problem
of rewriting the entire code including the (slight?) architecture
change.

After some extensive tests we had decided to put it into production.
About 1000 16C73 powered units later, with different fab date 73's and
the change from 73A to 73B, we still haven't seen any problems.
Perhaps the small overclocking of 480kHz isn't big enough yet to
create
potential problems.

BTW, how difficult is it to port 16C73 code to an equivalent Atmel
part. Is there a slight compatibility between the two?

Daniel...

1999\09\16@212723 by Stuart O'Reilly

flavicon
face
Re: just how fast is a PIC?

100% fast <g>

1999\09\17@040916 by Morgan Olsson

picon face
Hej D. Schouten. Tack fšr ditt meddelande 23:14 1999-09-16 +0200 enligt nedan:

-snip-

>Perhaps the small overclocking of 480kHz isn't big enough yet to
>create
>potential problems.

The specified maximum clock is guaranteed over the entire operating voltage and temperature range (simultaneously)

If we run the device at 5V+/-5%, good decoupled, and at room temperature, we probably have very wide margin, just how wide I have not seen any spec.

Anybody?

/Morgan


Morgans Reglerteknik, HŠllekŒs, 277 35 KIVIK, SWEDEN
  tel +46(0)414-446620, fax -70331,   mrtspamKILLspaminame.com

More... (looser matching)
- Last day of these posts
- In 1999 , 2000 only
- Today
- New search...