AGP/PCI Clock Correlation

In a previous post, PRIVATE1964 disputed that the AGP clock(when
adjusted) does not affect the PCI clock speed, which as many have
proven is false.


If you agree with the fact that the PCI clock speed is raised in a
direct affect to the AGP clock, reply as such.



If someone OTHER THAN PRIVATE can dispute that this is not the case,
please do so.


I'd like to be proven wrong, as long as it's done by someone coherrent.

Isn't it a case where both statements could be true depending on the
motherboard / chipset and implementation by the manufacturer if the PCI/AGP
speed is synchronous with the FSB or not and then if the board has a PCI
divider. And if that divider could influence the AGP speed as well - since
that works from the initial speed of the FSB, but.....

I have not worked with a system though where when a PCI speed is not locked
at 33mhz but controlled through a divider and adjusting that divider also
changes the AGP speed. But I have worked on a system where the FSB would
increase the PCI speed and so used a divider setting to reduce it back close
to 33mhz but has left the AGP running above the 66mhz it is specified to run
at - usually, but not always, without problem as the most sensitive of the
two was the PCI bus running out of spec and in particular the IDE
controller.

As a side note I have found that adjusting an AGP bus speed from 66mhz all
the way to 99mhz has no beneficial effect as I once thought it would.



--
Regards

Morgan

No more noisy hard drive for me...

www.flyinglizard.freeserve.co.uk

I am an A+ Certified Technician and have been involved with computers for
over 20 years and I have never seen an AGP bus speed effect the PCI speed.

The PCI bus speed is determined by a divider applied to the fsb(CLK), the AGP
speed is also determined by a divider off the fsb(CLK) or can be locked to
66mhz for motherboards that don't offer the option in the bios.

If you run the fsb out of spec and you don't have the proper divider for the
PCI bus, then the PCI bus speed will be overclocked, but that doesn't effect
what the AGP bus is doing. The fsb(CLK) effects the AGP speed, but not with the
Nforce chipset.
The main point is if you have an option in your bios that can increase the AGP
bus speed, it does not increase the PCI speed also, but the only way to run the
AGP out of spec is to overclock the fsb. Even that is not 100% true anymore
because the Nforce chipset does let you overclock the AGP bus speed even if the
fsb is running within spec.


You have a main clock signal that the PCI and AGP bus speeds work off of with
dividers. PCI does not effect AGP and AGP does not effect PCI. The main clock
and dividers determine the speed of each one.

This topic is ridiculous. The same person that posted this topic claimed there
is no such thing as an AGP bus.

http://www.nvidia.com/object/LO_20010614_6672.html

As a side note I have found that adjusting an AGP bus speed from 66mhz all
the way to 99mhz has no beneficial effect as I once thought it would.


It does give you something in benchmark tests, but so little it probably
doesn't give you anything extra in actual 3D applications like games.

Running 3Dmark2001at 99Mhz gives me around 200 extra points over 66Mhz. The
gain is most likely not worth the risk of running the video card out of spec.

"teqguy" wrote in message
...
In a previous post, PRIVATE1964 disputed that the AGP clock(when
adjusted) does not affect the PCI clock speed, which as many have
proven is false.


If you agree with the fact that the PCI clock speed is raised in a
direct affect to the AGP clock, reply as such.

In my experience (more than 20 years working with computers), with an HP
S.T.A.R. certification, Microsoft MSCE 3.51, 4.0 +I, 2000 +I ,Compaq ASE,
and a Cisco Design Specialist designation... In the real world of business,
this isnt a matter worth contemplating.

However, as an enthusiast , hobbist, gamer, overclocker, whatever... its
academic.

All things begin with the FSB. Branched busses are based on factors of the
FSB. Branches are determined by dividers which are set in stone. You cannot
independently alter one bus without altering another. AGP will always be 2x
of the PCI. the PCI will always be based at1/3, 1/4, 1/5, or 1/6 of the FSB.
Period. And CPU and Memory Bus' are always the same.

Some board makers will allow you to increase the CPU bus, while locking the
PCI and AGP at the standard dividers for a given bus speed. But this does
not mean that you can alter only one bus frequency in complete isolation.
There is always a tether to another bus.



If someone OTHER THAN PRIVATE can dispute that this is not the case,
please do so.


I'd like to be proven wrong, as long as it's done by someone coherrent.

i'm not sure which side of your argument my comments support... all i can
tell you is that its accurate and truthful. Do with it what you will.

"nut cracker" wrote in message
...

"teqguy" wrote in message
...
In a previous post, PRIVATE1964 disputed that the AGP clock(when
adjusted) does not affect the PCI clock speed, which as many have
proven is false.


If you agree with the fact that the PCI clock speed is raised in a
direct affect to the AGP clock, reply as such.

In my experience (more than 20 years working with computers), with an HP
S.T.A.R. certification, Microsoft MSCE 3.51, 4.0 +I, 2000 +I ,Compaq ASE,
and a Cisco Design Specialist designation... In the real world of
business,
this isnt a matter worth contemplating.

However, as an enthusiast , hobbist, gamer, overclocker, whatever... its
academic.

All things begin with the FSB. Branched busses are based on factors of the
FSB. Branches are determined by dividers which are set in stone. You
cannot
independently alter one bus without altering another. AGP will always be
2x
of the PCI. the PCI will always be based at1/3, 1/4, 1/5, or 1/6 of the
FSB.
Period. And CPU and Memory Bus' are always the same.

Above statement applies to real world of technology. You wont find a HPaq,
Dell or IBM server that will allow you to monkey with that. On the
otherhand... for consumers:

There are technical deviations from this, like VIA Appolo Pro133A chipsets
allowing you to run PC133 memory at 133Mhz while you have a 100Mhz FSB
processor.... My KT333 chipset on my Abit KX7 allows me to have a 200/233
CHIP with 333 DDR.

My KT400 chipset allows me to run my memory out of sync with the CPU bus,
but only at speeds faster than the CPU bus.

But DDR and RAMBUS are muliples of the FSB on the board. Most current CPU's
"pump" the FSB to give you the internal clock of the CPU. So, an 800Mhz FSB
P4 is really a Quad 200Mhz FSB. AMD's are doubled (thanks to the Alpha EV
bus architecture, 333Mhz Bartons are (2x 133 FSB).. and on and on.


Some board makers will allow you to increase the CPU bus, while locking
the
PCI and AGP at the standard dividers for a given bus speed. But this does
not mean that you can alter only one bus frequency in complete isolation.
There is always a tether to another bus.



If someone OTHER THAN PRIVATE can dispute that this is not the case,
please do so.


I'd like to be proven wrong, as long as it's done by someone coherrent.

i'm not sure which side of your argument my comments support... all i can
tell you is that its accurate and truthful. Do with it what you will.

But this does
not mean that you can alter only one bus frequency in complete isolation.

That's not true. I can set the AGP bus speed on my NF7-S to anything I want up
to 99Mhz and it does not effect the PCI bus one bit.

You cannot
independently alter one bus without altering another.

That's not true, if that was the case I would not be able to set the AGP bus
speed in my bios on my NF7-S to anything I want up to 99Mhz which I can do and
have done.
It did not effect the PCI bus speed at all.

That is a fact!

If it also raised the PCI bus do you honestly think my hard drives would run OK
at 50Mhz because that would be the speed the PCI bus would be running at
according to you? Unless it was locked at 33Mhz which it is.

If you increase the fsb, the PCI bus and AGP bus speeds will also increase.
Unless the PCI bus uses a divider to limit the speed to 33mhz. The same goes
for the AGP bus, but it would be limited by a divider.
If I change the divider for the AGP bus so it runs faster, that does not effect
the divider for the PCI bus. The PCI bus will alway run at the speed of it's
own divider.

Why would the PCI bus be effected by what the AGP bus speed is doing?


All things begin with the FSB. Branched busses are based on factors of the
FSB. Branches are determined by dividers which are set in stone.

That it correct, but changing the divider of one does not change the divider of
the other one.

I'm done with this. I'm not gonna argue about a simple fact, and I do mean
simple fact about computer buses.
So you can all go on and discuss it to death, I just don't care.
The fact remains plain as day, increasing the AGP bus speed does not effect the
PCI bus speed like the person who posted this topic has claimed.
If you think the PCI bus is effected by raising the AGP bus speed then maybe
you should go back to school
and learn some more.

Case Closed!

You cannot
independently alter one bus without

Here's some more proof.

http://www.rme-audio.com/english/tec..._tweaks_01.htm

Scroll down until you see this

"though the AGP bus speed can be manually configured independently of the
locked PCI bus speed)."


I rest my case morons!

so much for being done and case closed, eh?

your example is taken, but that is not the standard.

"PRIVATE1964" wrote in message
...
You cannot
independently alter one bus without

Here's some more proof.

http://www.rme-audio.com/english/tec..._tweaks_01.htm

Scroll down until you see this

"though the AGP bus speed can be manually configured independently of the
locked PCI bus speed)."


I rest my case morons!