Motherboard Forums


Reply
Thread Tools Display Modes

Would this be an overclocked system?

 
 
David Farber
Guest
Posts: n/a
 
      10-10-2007, 10:40 PM
Athlon XP3000+ M CPU in an ASRock K7VT4A+. 512Mb PC2700 ram.

The default settings according to AMD is fsb=133MHz and a multiplier of 16
(2128MHz).

Would there be an advantage gained if I change the fsb to 166MHz and the
multiplier to 13 (2158MHz)? I believe this motherboard allows you to change
the multiplier settings. Could the increased fsb cause problems with the
other devices on the motherboard or does the motherboard automatically set
the correct dividers for the other components?

Thanks for your reply.
--
David Farber
L.A., CA


 
Reply With Quote
 
 
 
 
Paul
Guest
Posts: n/a
 
      10-11-2007, 06:47 AM
David Farber wrote:
> Athlon XP3000+ M CPU in an ASRock K7VT4A+. 512Mb PC2700 ram.
>
> The default settings according to AMD is fsb=133MHz and a multiplier of 16
> (2128MHz).
>
> Would there be an advantage gained if I change the fsb to 166MHz and the
> multiplier to 13 (2158MHz)? I believe this motherboard allows you to change
> the multiplier settings. Could the increased fsb cause problems with the
> other devices on the motherboard or does the motherboard automatically set
> the correct dividers for the other components?
>
> Thanks for your reply.


Executive summary:

Yes. There is an advantage. More FSB speed allows better memory transfer speeds.

There is no "lock" setting or discrete frequency setting ( AGP/PCI 66.66/33.33MHZ )
in your BIOS.

There are four frequencies of interest.

1) Clock feeding CPU.
2) Clock feeding memory (usually a ratio of simple integers, times CPU clock signal)
3) Clock feeding PCI cards
4) Clock feeding AGP card.

3 & 4 tend to be related to one another (AGP = 2 * PCI), so I won't treat them
separately, and assume they are related by a factor of 2. If PCI is 37.5 MHz
(the max useful value with no danger), then AGP would be 75MHz. The standards
define AGP and PCI as 66.66MHz and 33.33MHz as nominal values. Values above 75MHz
for AGP, cause problems for cards more modern than ATI 9800Pro. Older cards,
can take clocks higher than that.

In the clock generator chip, for older technologies (where the BIOS doesn't show
a separate setting for AGP and PCI clocks), the PCI clock is derived from the
CPU clock. If CPU is 133Mhz, they use a divider of 4 inside the clock generator
chip. If CPU is 166MHz, they use a divider of 5. The "special" frequencies
of 133MHz and 166MHz, also happen to be the values selected by the processor maker.

Intermediate frequencies may be selected manually in the BIOS. For example, bumping
up the 133Mhz clock to 150MHz, gives 150/4 = 37.5MHz for PCI. Therefore, for maximum
safety on an older motherboard, you avoid frequencies between 150 and 165MHz, and the
next useful value is 166MHz (if it works). It means that while the BIOS may offer
a fine tuning option, there are still "zones" of frequencies to avoid. Corruption
of your disk drive may result if you violate the rule. (Which is why, if testing
for initial stability, I like to boot with a Knoppix CD in the CDROM drive. You
can't hurt a CD. Knoppix is a 700MB .ISO download from knopper.net and is a Linux
operating system.)

The memory clock settings in the BIOS, are not absolute frequencies as such. If you
select 133MHz for memory, and the CPU is 133MHz as well, the design probably has a
1:1 ratio between the two clock signals. If you then bump the CPU clock manually
to 150MHz, that should result in the memory overclocking by the same amount. A
memory at 150MHz, is running at DDR300. When the CPU clock is using one of the magic
values of 133MHz or 166MHz, it is also possible that the memory values will
again read correctly. (So, in a sense, "sane" ratios exist at the magic clock
values. Oddball values, which you can still adjust around, exist for intermediate
settings. You can observe how this works, by using a copy of CPUZ from cpuid.com
and make tiny changes, then observe the reported results in CPUZ.)

So, by all means, try the CPU jumper of 166MHz. Your RAM is PC2700 or DDR333, and it
is a perfect fit for a DDR333 setting in the BIOS. The best transfer efficiency, is
often when the CPU and memory on these systems, is at the 1:1 ratio. So if using
FSB266, then DDR266 might be good. (To do better on the memory, you might have to
go to DDR400 setting, to beat the latency disadvantage, with the clock rate advantage.
Add latency happens, when the memory bus samples need to be resynchonized to the
FSB interface.)

If using FSB333, then DDR333 might be good. Check that the multiplier setting you need,
works properly while you are still at FSB266 or lower. If you are sure the multiplier
works properly, then switch up to FSB333 and try it. If the setting doesn't work,
try FSB266 jumper and DDR266 BIOS settings, then use the manual BIOS control to go
from a CPU input clock of 133MHz, to 150MHz. 150MHz should be the next lowest safe
clock choice.

To see some of how the clock generator chip works, this is a sample datasheet.
This clock generator is a couple generations older than the one on your motherboard.
You can see how the PCI clock is affected by the main clock choice here. The jumpers
on your motherboard, would go to the equivalent of the FS3,FS2,FS1,FS0 signals,
and helps to select a "magic" value for system startup.

http://web.archive.org/web/200410141...ics9250-08.pdf

There are many clock generators made by that company. To see how many, have
a look at this list.

http://web.archive.org/web/200410141...w.icst.com/pdf

And yes, these are all aspects of overclocking.

HTH,
Paul
 
Reply With Quote
 
 
 
 
David Farber
Guest
Posts: n/a
 
      10-11-2007, 05:39 PM

"Paul" <(E-Mail Removed)> wrote in message news:fekgj7$9qu$(E-Mail Removed)...
> David Farber wrote:
> > Athlon XP3000+ M CPU in an ASRock K7VT4A+. 512Mb PC2700 ram.
> >
> > The default settings according to AMD is fsb=133MHz and a multiplier of

16
> > (2128MHz).
> >
> > Would there be an advantage gained if I change the fsb to 166MHz and the
> > multiplier to 13 (2158MHz)? I believe this motherboard allows you to

change
> > the multiplier settings. Could the increased fsb cause problems with the
> > other devices on the motherboard or does the motherboard automatically

set
> > the correct dividers for the other components?
> >
> > Thanks for your reply.

>
> Executive summary:
>
> Yes. There is an advantage. More FSB speed allows better memory

transfer speeds.
>
> There is no "lock" setting or discrete frequency setting ( AGP/PCI

66.66/33.33MHZ )
> in your BIOS.
>
> There are four frequencies of interest.
>
> 1) Clock feeding CPU.
> 2) Clock feeding memory (usually a ratio of simple integers, times CPU

clock signal)
> 3) Clock feeding PCI cards
> 4) Clock feeding AGP card.
>
> 3 & 4 tend to be related to one another (AGP = 2 * PCI), so I won't treat

them
> separately, and assume they are related by a factor of 2. If PCI is 37.5

MHz
> (the max useful value with no danger), then AGP would be 75MHz. The

standards
> define AGP and PCI as 66.66MHz and 33.33MHz as nominal values. Values

above 75MHz
> for AGP, cause problems for cards more modern than ATI 9800Pro. Older

cards,
> can take clocks higher than that.
>
> In the clock generator chip, for older technologies (where the BIOS

doesn't show
> a separate setting for AGP and PCI clocks), the PCI clock is derived from

the
> CPU clock. If CPU is 133Mhz, they use a divider of 4 inside the clock

generator
> chip. If CPU is 166MHz, they use a divider of 5. The "special" frequencies
> of 133MHz and 166MHz, also happen to be the values selected by the

processor maker.
>
> Intermediate frequencies may be selected manually in the BIOS. For

example, bumping
> up the 133Mhz clock to 150MHz, gives 150/4 = 37.5MHz for PCI. Therefore,

for maximum
> safety on an older motherboard, you avoid frequencies between 150 and

165MHz, and the
> next useful value is 166MHz (if it works). It means that while the BIOS

may offer
> a fine tuning option, there are still "zones" of frequencies to avoid.

Corruption
> of your disk drive may result if you violate the rule. (Which is why, if

testing
> for initial stability, I like to boot with a Knoppix CD in the CDROM

drive. You
> can't hurt a CD. Knoppix is a 700MB .ISO download from knopper.net and is

a Linux
> operating system.)
>
> The memory clock settings in the BIOS, are not absolute frequencies as

such. If you
> select 133MHz for memory, and the CPU is 133MHz as well, the design

probably has a
> 1:1 ratio between the two clock signals. If you then bump the CPU clock

manually
> to 150MHz, that should result in the memory overclocking by the same

amount. A
> memory at 150MHz, is running at DDR300. When the CPU clock is using one of

the magic
> values of 133MHz or 166MHz, it is also possible that the memory values

will
> again read correctly. (So, in a sense, "sane" ratios exist at the magic

clock
> values. Oddball values, which you can still adjust around, exist for

intermediate
> settings. You can observe how this works, by using a copy of CPUZ from

cpuid.com
> and make tiny changes, then observe the reported results in CPUZ.)
>
> So, by all means, try the CPU jumper of 166MHz. Your RAM is PC2700 or

DDR333, and it
> is a perfect fit for a DDR333 setting in the BIOS. The best transfer

efficiency, is
> often when the CPU and memory on these systems, is at the 1:1 ratio. So if

using
> FSB266, then DDR266 might be good. (To do better on the memory, you might

have to
> go to DDR400 setting, to beat the latency disadvantage, with the clock

rate advantage.
> Add latency happens, when the memory bus samples need to be resynchonized

to the
> FSB interface.)
>
> If using FSB333, then DDR333 might be good. Check that the multiplier

setting you need,
> works properly while you are still at FSB266 or lower. If you are sure the

multiplier
> works properly, then switch up to FSB333 and try it. If the setting

doesn't work,
> try FSB266 jumper and DDR266 BIOS settings, then use the manual BIOS

control to go
> from a CPU input clock of 133MHz, to 150MHz. 150MHz should be the next

lowest safe
> clock choice.
>
> To see some of how the clock generator chip works, this is a sample

datasheet.
> This clock generator is a couple generations older than the one on your

motherboard.
> You can see how the PCI clock is affected by the main clock choice here.

The jumpers
> on your motherboard, would go to the equivalent of the FS3,FS2,FS1,FS0

signals,
> and helps to select a "magic" value for system startup.
>
> http://web.archive.org/web/200410141...ics9250-08.pdf
>
> There are many clock generators made by that company. To see how many,

have
> a look at this list.
>
> http://web.archive.org/web/200410141...w.icst.com/pdf
>
> And yes, these are all aspects of overclocking.
>
> HTH,
> Paul


I believe your executive summary really connected the dots for me. Let me
make sure I understand this though. As long as I pick one of the preset (is
that what you define as magic?) frequencies, the "frequencies of interest"
take care of themselves. The tricky part is what happens when you bump the
fsb speed a bit and all the other frequencies get bumped as well which is
why you need to know what frequencies to avoid.

I changed the fsb to 166, the multiplier to 13, and so far, the system is
stable.

Thanks for your reply.
--
David Farber
L.A., CA


 
Reply With Quote
 
Paul
Guest
Posts: n/a
 
      10-11-2007, 11:56 PM
David Farber wrote:

> I believe your executive summary really connected the dots for me. Let me
> make sure I understand this though. As long as I pick one of the preset (is
> that what you define as magic?) frequencies, the "frequencies of interest"
> take care of themselves. The tricky part is what happens when you bump the
> fsb speed a bit and all the other frequencies get bumped as well which is
> why you need to know what frequencies to avoid.
>
> I changed the fsb to 166, the multiplier to 13, and so far, the system is
> stable.
>
> Thanks for your reply.


The set of frequencies defined by the processor manufacturer, all seem to be
a multiple of the PCI standard frequency. 133MHz, 166Mhz, and 200Mhz are
examples. They may have chosen those frequencies, so they could use a
simple divider to make PCI and AGP frequencies. They have other techniques
now, for creating PCI and AGP, which allows the PCI and AGP to be "locked".
So on more modern boards, you can dial the CPU clock in 1MHz increments,
without having to worry about PCI and AGP. The frequencies are no longer
"magic".

But previous generations were not quite as flexible, and the PCI/AGP/memory
clocks floated along with the CPU clock. And that is why there are certain
CPU clock settings, which might not make the other bits of hardware happy.

One reason for corrupting a disk, is the PCI clock is used to derive a
clock for the IDE ribbon cable to the disk drive. If the IDE clock goes
too far out of spec, that corrupts data transfer to and from the drive.
Even SATA and PCI Express interfaces have issues like this, and there are
some SATA motherboards, where overclocking will corrupt SATA. Even more
recent clocking schemes, have fixed those kinds of issues, so that
SATA and PCI Express are locked (i.e. independent) as well.

If you changed to the CPU input to 166MHz (=FSB266), and set the multiplier
as you liked, and CPUZ verifies that all the resulting frequencies are
correct, you have nothing to worry about. Test with Orthos or Prime95 "torture
test", and verify that the machine is really stable. Just booting into
Windows is not a stability test. I put the Orthos link first here, simply
because the interface is a little easier to understand. Orthos will run more
than one thread for testing, and so Orthos can be used on dual core processors.

http://sp2004.fre3.com/beta/beta2.htm (Orthos tester)
http://www.mersenne.org/freesoft.htm (Prime95)

Prime95 was invented to find Prime Numbers, a mathematics thing. The torture
test option in the Prime95 program, was simply put there to verify that a
computer about to work on finding prime numbers, was working properly. No
sense verifying a prime number, with a computer that makes mistakes. Purely
as an accident, overclockers discovered the testing option, and the
Prime95 program became more popular because of that feature, than it
did in furthering the study of mathematics.

Paul
 
Reply With Quote
 
 
 
Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
Prescott Overclocked to 4.2GHz!!!!! Wayne Youngman Abit 18 02-16-2004 09:05 PM
over overclocked eng Asus 7 11-15-2003 04:28 AM
My overclocked P4 rig SilverBack Abit 3 11-07-2003 06:13 PM
NF7-m most overclocked SL Abit 1 10-10-2003 12:30 PM
Overclocked NF7-S v2 Keyboard & Mouse problems Doughnut Abit 5 09-09-2003 08:35 PM


All times are GMT. The time now is 10:23 AM.


Welcome!
Welcome to Motherboard Point
 

Advertisment