where's the multiplier?

my interest wasn't really which were locked and which were unlocked.


I found the ans to where the mult is - all quotes are googlable on
usenet or the web.

--
"The multiplier is "inside" the CPU. "
--
"The multiplier is in the processor and is input to a phase locked loop

which multiplies up the FSB clock for the internal processor clock. "
--

and this gerat one from
http://www.computerdoctor.co.uk/FAQs/processors.html

explaining the multiplier and then second paragraph, contrasting with
the fsb

Computer Doctor: There are two factors that control CPU speed, the
speed of the motherboard bus and the internal multiplier in the CPU.
Motherboards may have multiplier settings, but these settings merely
"tell" the CPU what to do, and the CPU is free to ignore those settings
if so designed. Most Intel processors come with the multiplier locked,
so that if you do change the multiplier setting on the motherboard, the
CPU will simply ignore it.

Starting with the Pentium II Intel designed it such that the CPU
"tells" the motherboard what the bus speed (and voltage) should be via
two pins called BSEL0 and BSEL1. Motherboards can however include
override capabilities for these settings, which will ignore what the
CPU says and set the bus speed (or voltage in some cases) to whatever
you want. Unfortunately, Intel's own boards do not allow this type of
"hot rod" control, and you are left stuck with the default settings
dictated by the CPU. ..

---------------

I assume that the memory multiplier is internal to ram , and so the
mobo tels the ram what multiplier to use. Thus far, - it seems to me -
memory is unlocked.
 >> Stay informed about: where's the multiplier? 

q_q_anonymous.DeleteThis@yahoo.co.uk wrote:

 > i see that new CPUs have the multiplier within. But I recall that old
 > board had a jumper for setting the multiplier. Am I then right in
 > assuming that the old CPUs that went in those old boards had no
 > multiplier of their own?
 >
 > where do they get the multipiler number from like 10.5 or 9. A signal
 > the board? or within the cpu itself.
 >
 > when i read abou the cpu being locked, it always says "it doesn't
 > respond to a multiplier change in the BIOS or board" . But it seems to
 > me that the CPU doesn't just not respond. It multiplies the FSB by some
 > fixed number. The old CPUs didn't do that, did they?
 >
 > the multiplier seems to have moved from the board(in the old days) to
 > the cpu(nowadays)?!
 >

The multiplier is, and always was, 'on the chip'.

A low frequency clock is provided to the CPU which internally multiplies it
up to the needed core frequency. Now, from a practical matter of maximizing
the usable CPUs from the large manufacturing die it's convenient to have
more than one multiplier on the die so the cpus which can run faster can be
clocked faster but the slower ones used as well, albeit at a slower speed
(and for a lower price).

And the easiest way to do that is to have some select lines that pick which
of the available multipliers to use. And it's also easiest to simply
provide pins for them and put jumpers on the motherboard with instructions
on which ones to set where for the particular processor purchased.

This, btw, is how the 1.5x setting on a Pentium Classic becomes a 3.5x on
the Pentium MMX and why Cyrix multipliers often didn't match what the Intel
ones were. What the motherboard 'multiplier' lines select is determined by
the processor. By themselves they have no meaning.


As the manufacturing process improves more and more of the total processors
will run at the faster speeds, but there's still a market for the cheaper,
slower rated, ones. And if enough people will buy the slower ones it's
worth it to simply take some of the faster ones and mark them with the
slower rating. There is also the matter that a slow one may run 'fast'
under some conditions although not all and so don't meet specification.
Like, maybe it works as long as the room temperature is under 80F. Or it'll
work if the power supply runs a tad bit high but not if it's on the low
side of the acceptable range. And, lastly, some of the slow ones may still
be simply slow ones.


Now enter people trying to make a fast buck at the buyer's expense. Buy
slow processors then remark and sell them as the faster ones. They may or
may not work. Of course, the flaky ones get blamed on the processor
manufacturer.

Counter move: Fix the select lines so they cannot be changed. They're
accessible during manufacture and for testing to qualify the processor
speed rating, so you can still use from slow to fast, and then 'burned in',
like a programmable memory, after the rating is selected.

Functionally it's the same. It's just that the external select lines are
intentionally broken to prevent tampering.<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: where's the multiplier? 

Try some more research.

Phil Weldon

<q_q_anonymous.DeleteThis@yahoo.co.uk> wrote in message
news:1118430624.291756.129880@g14g2000cwa.googlegroups.com...
 > my interest wasn't really which were locked and which were unlocked.
 >
 >
 > I found the ans to where the mult is - all quotes are googlable on
 > usenet or the web.
 >
 > --
 > "The multiplier is "inside" the CPU. "
 > --
 > "The multiplier is in the processor and is input to a phase locked loop
 >
 > which multiplies up the FSB clock for the internal processor clock. "
 > --
 >
 > and this gerat one from
<font color=purple> > <a style='text-decoration: underline;' href="http://www.computerdoctor.co.uk/FAQs/processors.html</font" target="_blank">http://www.computerdoctor.co.uk/FAQs/processors.html</font</a>>
 >
 > explaining the multiplier and then second paragraph, contrasting with
 > the fsb
 >
 > Computer Doctor: There are two factors that control CPU speed, the
 > speed of the motherboard bus and the internal multiplier in the CPU.
 > Motherboards may have multiplier settings, but these settings merely
 > "tell" the CPU what to do, and the CPU is free to ignore those settings
 > if so designed. Most Intel processors come with the multiplier locked,
 > so that if you do change the multiplier setting on the motherboard, the
 > CPU will simply ignore it.
 >
 > Starting with the Pentium II Intel designed it such that the CPU
 > "tells" the motherboard what the bus speed (and voltage) should be via
 > two pins called BSEL0 and BSEL1. Motherboards can however include
 > override capabilities for these settings, which will ignore what the
 > CPU says and set the bus speed (or voltage in some cases) to whatever
 > you want. Unfortunately, Intel's own boards do not allow this type of
 > "hot rod" control, and you are left stuck with the default settings
 > dictated by the CPU. ..
 >
 > ---------------
 >
 > I assume that the memory multiplier is internal to ram , and so the
 > mobo tels the ram what multiplier to use. Thus far, - it seems to me -
 > memory is unlocked.
 ><!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: where's the multiplier? 

q_q_anonymous DeleteThis @yahoo.co.uk wrote:
[...]
 > "The multiplier is in the processor and is input to a phase locked
 > loop
 >
 > which multiplies up the FSB clock for the internal processor clock. "

This one is the correct one, but possibly can be interpreted wrong. What it
means is the following:

FSB >--- PLL ------->------+---> CPU CLK
^ |
| |
+-- Divider <----+

A PLL can be thought of a a magic thing that adjusts its output so that the
two inputs are the same. The divider is loaded with the "multiplier". So if
you have a 200MHz FSB and a 10x "multiplier" then the PLL will put out 2GHz:
the divider takes the 2GHz, divides it by 10 to give an output of 200MHz,
which matches the other input to the PLL. The divider usually can be
configured for a range of values. This is all inside the CPU.

[...]
 > I assume that the memory multiplier is internal to ram , and so the
 > mobo tels the ram what multiplier to use. Thus far, - it seems to me -
 > memory is unlocked.

Err, memory multiplier? I think you may be slightly confused here ...

--
Michael Brown
<a style='text-decoration: underline;' href="http://www.emboss.co.nz" target="_blank">www.emboss.co.nz</a> : OOS/RSI software and more
Add michael@ to emboss.co.nz ---+--- My inbox is always open<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: where's the multiplier? 

  >> I assume that the memory multiplier is internal to ram , and so the
  >> mobo tels the ram what multiplier to use. Thus far, - it seems to me -
  >> memory is unlocked.

 >Err, memory multiplier? I think you may be slightly confused here ...

SDRAM - Is Synchronous with a capital S, so it has a clock. And all
clocked devices in a computer derive their clock from the FSB. And if
they're not running at the speed of the FSB, then they must be using a
multiplier (or dividor). I recall running FSB=100, and SDRAM at
133MHz. then some electronics (a memory multiplier and/or dividor )
was making it run at 4/3 (as in four thirds) times the FSB. The
question, was then whether this thing (which I was calling a ''memory
multiplier'), is inside the memory - just as the CPU multiplier is
inside the CPU. I assumed it was inside, and that the only significant
difference, was that memory clocks are not locked, - they listen to
the speed suggested by the mobo.<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: where's the multiplier? 

thanks for that info. I pity the BIOS programmer that includes a drop
down menu where the user 'selects a speed' !
Most speeds probably aren't supported (unless processors have 30
multipliers going from 0 to 15 in 0.5 increments!),. The drop down
menu is misleading - and a misleading GUI is dangerous. And that's just
from the user's perspective.
I imagine that the only reason they don't cause people problems is that
CPUs nowadays are multiplier locked!
 >> Stay informed about: where's the multiplier? 

q_q_anonymous.TakeThisOut@yahoo.co.uk wrote:

   >>>I assume that the memory multiplier is internal to ram , and so the
   >>>mobo tels the ram what multiplier to use. Thus far, - it seems to me -
   >>>memory is unlocked.
 >
 >
  >>Err, memory multiplier? I think you may be slightly confused here ...
 >
 >
 > SDRAM - Is Synchronous with a capital S, so it has a clock.

True. The system provided memory bus clock.

 > And all
 > clocked devices in a computer derive their clock from the FSB. And if
 > they're not running at the speed of the FSB, then they must be using a
 > multiplier (or dividor).

True, depending on who 'they' are.

 > I recall running FSB=100, and SDRAM at
 > 133MHz. then some electronics (a memory multiplier and/or dividor )
 > was making it run at 4/3 (as in four thirds) times the FSB.

Yes.

 > The
 > question, was then whether this thing (which I was calling a ''memory
 > multiplier'), is inside the memory - just as the CPU multiplier is
 > inside the CPU.

No

 > I assumed it was inside,

No.

 > and that the only significant
 > difference, was that memory clocks are not locked, - they listen to
 > the speed suggested by the mobo.


Originally, memory was run at the same speed as the FSB, using the FSB
clock, (which is a lot simpler for the chipset as everything is running in
step with each other) but modern chipsets provide a separate memory clock
that can be different than the FSB.

The memory has no idea of it nor any multiplier. It just runs at the speed
of the memory bus clock.<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: where's the multiplier? 

Since you already seem quite convinced that you know the answers, why ask
the questions?

All memory currently used in computers have a clock signal to drive memory;
DRAM, DDRAM, DDRAM2, SDRAM, RAM,SRAM, ROM, EROM, EEROM, RDRAM, ferrite core
non-volatile memory, ...

You really should try the Intel website. It might clear up some points that
we can't seem to explain. After all, the three who have posted in reply to
your original question have only posted 10,000 or so messages here.

Phil Weldon






<q_q_anonymous.TakeThisOut@yahoo.co.uk> wrote in message
news:1118528442.016051.47160@f14g2000cwb.googlegroups.com...
   >>> I assume that the memory multiplier is internal to ram , and so the
   >>> mobo tels the ram what multiplier to use. Thus far, - it seems to me -
   >>> memory is unlocked.
 >
  >>Err, memory multiplier? I think you may be slightly confused here ...
 >
 > SDRAM - Is Synchronous with a capital S, so it has a clock. And all
 > clocked devices in a computer derive their clock from the FSB. And if
 > they're not running at the speed of the FSB, then they must be using a
 > multiplier (or dividor). I recall running FSB=100, and SDRAM at
 > 133MHz. then some electronics (a memory multiplier and/or dividor )
 > was making it run at 4/3 (as in four thirds) times the FSB. The
 > question, was then whether this thing (which I was calling a ''memory
 > multiplier'), is inside the memory - just as the CPU multiplier is
 > inside the CPU. I assumed it was inside, and that the only significant
 > difference, was that memory clocks are not locked, - they listen to
 > the speed suggested by the mobo.
 ><!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: where's the multiplier? 

q_q_anonymous.RemoveThis@yahoo.co.uk wrote:
   >>> I assume that the memory multiplier is internal to ram , and so the
   >>> mobo tels the ram what multiplier to use. Thus far, - it seems to
   >>> me - memory is unlocked.
 >
  >> Err, memory multiplier? I think you may be slightly confused here ...
 >
 > SDRAM - Is Synchronous with a capital S, so it has a clock. And all
 > clocked devices in a computer derive their clock from the FSB. And if
 > they're not running at the speed of the FSB, then they must be using a
 > multiplier (or dividor). I recall running FSB=100, and SDRAM at
 > 133MHz. then some electronics (a memory multiplier and/or dividor )
 > was making it run at 4/3 (as in four thirds) times the FSB. The
 > question, was then whether this thing (which I was calling a ''memory
 > multiplier'), is inside the memory - just as the CPU multiplier is
 > inside the CPU. I assumed it was inside, and that the only significant
 > difference, was that memory clocks are not locked, - they listen to
 > the speed suggested by the mobo.

Well, it possibly gets a little curly RAM is clocked "externally" in the
sense that they operate exactly at the speed they are fed with by the
northbridge. This is why the ability to run ram async to the FSB is chipset
dependent, not RAM dependent. The situation in non-K8 system is:

Other stuff (southbridge, AGP, etc)
^ ^ ^
| | |
| | |
| | |
CPU <--- FSB ---> Northbridge <---- System clock
^
|
| RAM bus
|
V
RAM

The CPU is fed the system clock for the FSB, but the northbridge doesn't
always feed this same clock to the RAM. It can apply a 4/3 multiplier or
other variations to it beforehand. However, you could say that some RAM does
have an internal multiplier. DDR sort of has a 2x multiplier, and QDR sort
of has a 4x multiplier. These are fundamental to the RAM chips themselves
(ie: what make DDR DDR) so obviously cannot be changed.

The K8 situation is slightly different, though still has the same
fundamental idea. Rather than repeating myself, I'll refer you to
<a style='text-decoration: underline;' href="http://www.emboss.co.nz/amdmults/k8mults.html" target="_blank">http://www.emboss.co.nz/amdmults/k8mults.html</a>
(readme file from my A64 clock viewer tool)

--
Michael Brown
<a style='text-decoration: underline;' href="http://www.emboss.co.nz" target="_blank">www.emboss.co.nz</a> : OOS/RSI software and more
Add michael@ to emboss.co.nz ---+--- My inbox is always open<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: where's the multiplier? 

q_q_anonymous DeleteThis @yahoo.co.uk wrote:
 > thanks for that info. I pity the BIOS programmer that includes a drop
 > down menu where the user 'selects a speed' !
 > Most speeds probably aren't supported (unless processors have 30
 > multipliers going from 0 to 15 in 0.5 increments!),. The drop down
 > menu is misleading - and a misleading GUI is dangerous. And that's
 > just from the user's perspective.
 > I imagine that the only reason they don't cause people problems is
 > that CPUs nowadays are multiplier locked!

Works perfectly for my AMD XP1800+ T'bred B CPU on an nForce2 Ultra 400
chipset mobo. I just select 10.5 x, put the FSB at 200MHz and up the vcore
to 1.775 and she flies. 2.1GHz instead of the 1.53 it was sold as, still
using the AMD-supplied heatsink and fan.

CBId gives it a XP2750+ rating but in reality it's faster than that due to
the 200MHz FSB. It crunches SETI WUs in ~107% of the time an XP3200+ takes.
(On the same mobo).

Confused yet?
--
~misfit~<!-- ~MESSAGE_AFTER~ -->
 >> Stay informed about: where's the multiplier?