p4gd1 torture chamber

[Speedy Gonzales]

Ok, I didn't get any answers to most of my previous questions yet so far. But maybe I'll still get some of them answerred, who knows. I'll just await and see what happens. When it's about using a turbo-pll to externally feed the PCI-E and SATA: I'm currently examinating the data sheet and feature list of the ICS954119 clock generator;
Datasheet: http://www.icst.com/datasheets/ics954119.pdf
Feature list: http://www.icst.com/icscs/PartSummar...119&mode=short
According to the datasheet, there are several pins on the clockgerator that are related to the PCI-E and SATA functions:

PCI-E:
--------------------------------------------------------------------------
Pin # |Pin name |Pin type |Description
--------------------------------------------------------------------------
17___|PCIEXT0 |___OUT__|True clock of differential PCI_Express pair.
--------------------------------------------------------------------------
18___|PCIEXC0 |___OUT__|Complement clock of differential PCI_Express pair.
--------------------------------------------------------------------------
19___|VDDPCIEX|__PWR__|Power supply for PCI Express clocks, nominal 3.3V
--------------------------------------------------------------------------
21___|PCIEXT1 |___OUT__|True clock of differential PCI_Express pair.
--------------------------------------------------------------------------
22___|PCIEXC1 |___OUT__|Complement clock of differential PCI_Express pair.
--------------------------------------------------------------------------
23___|PCIEXT2 |___OUT__|True clock of differential PCI_Express pair.
--------------------------------------------------------------------------
24___|PCIEXC2 |___OUT__|Complement clock of differential PCI_Express pair.
--------------------------------------------------------------------------
30___|PCIEXC3 |___OUT__|Complement clock of differential PCI_Express pair.
--------------------------------------------------------------------------
31___|PCIEXT3 |___OUT__|True clock of differential PCI_Express pair.
--------------------------------------------------------------------------
32___|PCIEXC4 |___OUT__|Complement clock of differential PCI_Express pair.
--------------------------------------------------------------------------
33___|PCIEXT4 |___OUT__|True clock of differential PCI_Express pair.
--------------------------------------------------------------------------
34___|VDDPCIEX|__PWR__|Power supply for PCI Express clocks, nominal 3.3V
--------------------------------------------------------------------------

SATA:
--------------------------------------------------------------------------
Pin # |Pin name |Pin type |Description
--------------------------------------------------------------------------
26___|SRCCLKT |___OUT__|True clock of differential pair for S-ATA support.
_____|________|_________|+/- 300ppm accuracy required.
--------------------------------------------------------------------------
27___|SRCCLKC |___OUT__|Complement clock of differential pair for S-ATA
_____|________|_________|support. +/- 300ppm accuracy required.
--------------------------------------------------------------------------
28___|VDDSRC_|___PWR__|Supply for SRC clocks, 3.3V nominal
--------------------------------------------------------------------------

Now what I don't quite understand yet from the data sheet and featurelist:
1. What is meant with "differential pair" when it's about the PCI-E and SATA?
2. What is the difference between the "true" and "complement" clocks?
3. What is meant with "tight ppm accuracy clocks for Serial-ATA"?
Or: what is "ppm"?
4. What is "undriven differential CPU, SRC pair in PD# for power management."?

Also and in the first place: would it lead to any problems I'd have to find a solution for, if I externally feed the PCI-E and SATA continuously with 100MHz, seperated from the FSB, PCI-E and SATA frequencies values that have been set in BIOS? And/or would I still have to set the PCI-E according to the formula FSB = 2.2 * PCI-E in order to let the clock generator / BIOS / motherboard "think" that the PCI-E frequency would have been set to the "correct" value? I hope there's still people around here, that have been working with turbo-pll's back in the days when motherboards didn't have PCI/AGP locks yet... If there's still some of those people around, then I'd hereby like to ask to those people themselves: PLEASE help me, your knowledge, experiences and skills might get very useful to me and other P4GD1 and maybe also P4GPL-X owners who might want to consider a turbo-pll mod for their motherboards. I have a feeling that I don't have to create a complex kind of turbo-pll, probably I can even just use a pll sawed out of an old motherboard. But I just need to know a few things before actually creating it, in order to prevent failure and damage to the motherboard and so on. Thanks very much for ANYTHING that might get useful.

[skönnbrink]

For those who with a P4gpl-x i did a mod, haven't seen any other done it... Shows how smal the difference betwen p4gpl-x and p4gd1 is

http://www.xtremesystems.org/forums/...d.php?t=100455

To flash the board to a p4gd1 bios, do i need to force the flash somehow?

If i run AFUDOS i get this message when trying to flash
Error: ROM file ROMID is not compatible with existing BIOS ROMID

wanna try to find another IDE channel, se if that would work to. Just as with the mem-slot

[sky]

Chip mod 50kvr to 2nd leg from top on the left side of chip. Red circle is where to measure:

funny thing. if i try to get a reading there, the voltage my dmm shows
is 1.88 - 1.9v - this is on a board that isn't chipsetmodded, but has
the 3.3v -> vdimm mod in place.
also it doesn't matter what value i select for chips in bios (1.5 or 1.6v)
- it remains the same. anyone else have similar problems? seems a bit
odd - if that is indeed the real value it runs at, i wouldn't mind though


yesterday i primed a completely unmodded board at 265 - didn't check
what the actual chipset voltage on that thing was, but pci-e was at
122 and boot-fsb was 172 hm. at least that part works.. another non
modded board would only do 260 with same settings.. bummer (also
didn't check the real vchip on those two boards)

[ArneJohansen]

Have any1 tried the new nvidia 7950 gx2 in p4gd1 or p4gpl-x? Can`t se from the support list on nvidia homepage that it is supported. But i wonder if some1 here who has both can try and tell?

[Speedy Gonzales]

skönnbrink, great project you have going on there. I had a nice time reading about it and I wish you much success!

[Heartfixer]

Hello everyone.
I've just removed my dothan from the P4GPL-X flashed with P4GD1 bios (no big deal though) and replaced with a prescott E0 2.4A tested to be stable at 4.1GHz 1.6V on a P4P800SE.
The voltage fluctuations on vcore are huge, from 1.64 to 1.52V and the system is unstable. Has anyone managed to develop a droop mod for this motherboard?

Ok, we are all yelling here...

ANYBODY DEVELOPED A DROOP MOD?


Sorry for the yelling

[Speedy Gonzales]

While I was thinking about the ideas/theories of using a turbo-pll on the P4GD1, some other ways of that it might become useful came into my mind. At page 1 of this P4GD1 torture chamber thread, sayaa says the following at the Tips, Tricks and Tweaks section:

-------- TIP: the max fsb you can boot with is usually lower than the max fsb you can run in windows!
------------. thats why clockgen is very important on this board to increase the fsb in windows
------------. pentium m cpus initially boot with the default multiplier and fsb, then apply the fsb set in bios
------------. and then the multiplier set in bios. this means we cant boot at high fsbs because the stock
------------. multiplier with the increased fsb results in a huge cpu clock wich makes the cou crash the
------------. cpu, even though this high clockspeed is only applied for a fraction of a second.

and

-------- TIP: booting with 200fsb and increasing the fsb in windows with clockgen seems to give the best
------------. performence. once we boot with over 200fsb the bios probably automatically choses more
------------. relaxed advanced memory and chipset timings.

About using a turbo-pll for overcoming the problem of the board lacking a true and decent PCI-E/SATA lock, I'm still not sure yet. And since there doesn't seem to be anyone here, capable nor interested into contributing to that project, I'm not sure if I'll ever will. But one thing I am VERY convinced of, is that there's a big chance on that both of the problems that I quoted above, CAN be overcome by using a turbo-pll(?). My theories are that when you:

1.
replace the motherboard's crystal by one that has a frequency of only like 90%-ish of the default 14.3MHz one, so let's say a 12.0MHz one
2.
externally feed all the other frequencies that have to stay unaffected by the slower crystal frequency by using a turbo-pll
3.
use the "200" MHz FSB setting in BIOS

, this will result in that the motherboard will "think" it is in 200FSB mode, while in fact the true FSB will be +- 90% of 200MHz = 180MHz*. Then it should be possible to boot at the better performing "200FSB" mode even with the Dothans that have the highest multipliers. Because those Dothans will then not be overclocked so highly during boot-up that they'll crash, right?

With some of the more advanced types of turbo-pll's it's even possible to take off the onboard crystal and replace the signal that it used to generate by one that is externally generated by the turbo-pll. And also in a way that the frequency can be adjusted 100% variably, even on the fly during the motherboard's operation. So in case of using this type of turbo-pll / features, as soon as the motherboard would have booted up at 180FSB and in "200FSB" mode, the crystal frequency can then even be increased back to default 14.3MHz again in order to get the FSB "back" to 200MHz (or even higher, if desired). For example, in case of using phase-change cooling, by the time that Windows has then booted up at 180FSB in "200FSB" mode, the phase change system and CPU cooling unit may then have reached a temperature low enough for the CPU not to crash when the crystal frequency / FSB frequency is being increased again(?). I guess this type of a turbo-pll solution may be interesting to the people that would like to use and overclock the Dothans with the higher frequencies on the P4GD1, like the 770 or 780?

I wonder what you people here think about these theories. So here I go again: if there's anyone around here who does have the skills and knowledge, please give me some feedback on this matter. Anything is welcome! Thanks in the advance.


*Assuming that when the PCI-E, PCI, SATA and other peripherals are being ran at 90% of their stock frequencies (at least during boot-up), they will still be able to work fine. For example, back in the time when motherboards didn't have AGP/PCI locks yet, they usually worked fine when being ran with the AGP and PCI at respectively 60MHz and 30MHz.

[e.lectronick]

Hi, Speedy.
I've been interested in PLLs for another reason; I've been trying to figure out how to sync up my video capture card clock with that of my digital audio capture card. This would be to eliminate drift between sound and video when transferring my Laserdisc collection to digital format. It ocurred to me when I read your post that this project would involve essentially the same sort of reclocking or dividing of the original FSB as it would with my capture cards. I did a quick Google search for "turbo PLL" and found this site:
http://www.overclockers.com/tips745/
It's a bit old, but it covers the issue you were discussing.
I also found this article on the way the i915 chipset relates to the PCI express clock:
http://www.xbitlabs.com/articles/chi...fastest_3.html
I haven't read it all the way through, but I can't imagine that it'd be impossible to do something about the problem if enough people put their heads together.
DIYaudio.com has a number of people who reclocked their CD/DVD players as if it was nothing. I think many folks are more afraid to mess with their computers because they look so intimidating. I know that's my problem. I'm gonna try and get over that phobia. Its just wires and chips.
-Erik.

[Speedy Gonzales]

Hi e.lectronick. Thanks for your reply. I do indeed see the similarities between the mod you're looking for and the one I'm investigating the possibilities of. However, since it's off-topic on this "P4GD1 Torture Chamber " thread here, I'll send you an IM related to this matter soon. Also, I'd like to advice you to start a thread dedicated to this subject. That way more people might notice it. And afterall, just like you siad yourself: I can't imagine that it'd be impossible to do something about the problem if enough people put their heads together.

[Nanok]

If enybody wont to sell there ASUS P4GD1 plz pm me

[e.lectronick]

Hi,
Does anyone know whether its necessary to perform the VCORE mod on the P4GD1 if you plan to use the CT-479/Dothan CPU?

I am about to to all of the HiPro5 mods on the adapter, and I know it gets its own power from the 4 pin connector, but I'm still wondering if there is any benefit to having the VCORE mod installed anyway?

Thanks, and sorry if this is a dumb question.

-Erik.

[caligula]

For those who with a P4gpl-x i did a mod, haven't seen any other done it... Shows how smal the difference betwen p4gpl-x and p4gd1 is

http://www.xtremesystems.org/forums/...d.php?t=100455

To flash the board to a p4gd1 bios, do i need to force the flash somehow?

If i run AFUDOS i get this message when trying to flash
Error: ROM file ROMID is not compatible with existing BIOS ROMID

wanna try to find another IDE channel, se if that would work to. Just as with the mem-slot

Try the old school parameters, /n /pbnc
Worked for me so I have now P4GD1 bios on P4GPL-X.

[m0tek]

Finally got my hands on one , here on israel - and since then - ONLY PROBLEMS

Ive Installed a 480w Thermaltake Truepower , a X1900Xt and a 3.0 ghz P4 , 2 Gb of Certified ADATA memory (Already Memtested for 24 hours , no errors)

Before that - i had a p4p800-e deluxe that was workin quite well.

Anyways - i reinstalled Windows ofcourse + all drivers , im lettin u know now , i had this cpu overclocked when i was on H20 , but now im on air cooling , waitin for a Danger den parts to arrive , so im on stock clocks.

So - after installing ive encounterd a few errors.

1.Computer just freezes after somthing like 15-30 minz - and its not heat connected (picture jus freezes and i have to turnoff-on the computer)

2.i got 2 Ide hd's , 1 Cdrom , and 1 Sata Hd

Cdrom on Primary IDE (if its not there it doesnt work for some reason - even tho i messed with jumpers) , 2 Hd's on Pri Raid Ide , Sata is on First Sata ..
the Hd's (Ide) + Cdrom seems to work well , the sata drive is detected within windows But shows up as an Unformatted Drive (like a floppy with no disk inside) , when i access it , it says "The disk is not formatted..." u know the drill , so ive figures theres somthing wrong with either the mobo or the drivers...


3.Ive updated to 1007 Beta ...might that be the problem for the freezing? should i revert to 1005 or 1006?

Thx for ur time

a very frustrated Gamer.

[Kardahs]

Hi everybody

I have purchased a P4GD1 some weeks ago and thanks to this thread everything's working fine with my Dothan CPU @13x235 Mhz

However the board won't post when I have 4 x 512Mb sticks in the 4 dimm slots. The 4 sticks work fine 2 by 2 in either of the 2 dual channel slots and it's just when their all togheter that it isn't working.These ' sticks used to work perfecty togheter on P4C800E-Deluxe and P4P800SE.

Has anyone come across the same problem ?

Thank you in advance for your help

[m0tek]

hey , ive fixed my prob
turns out to be bad mem

anyways -
i do have now 2 x adata , 2 x twinmos

alone - pc 3200

all togather - 2700 , quite resambles ur prob

[Speedy Gonzales]

Ok, we indeed seem to be shouting across each others here... So I guess I'll just shout my stuff too.

About my turbo-pll ideas... again. :P I would appreciate some direct feedback, about some things that still aren't 100% clear to me yet. For example, why is it particularly more interesting to boot with 200+ FSB rather than just booting somewhere between 167MHz and 200MHz FSB and further increasing it to 200+ FSB under Windows? I'm curious at this because I'm currently looking into wheather or not it may be worth using a turbo-pll to merely achieve being able to boot with an FSB of 200+ MHz set in BIOS when using a Pentium-M 780, without that such a setting will lead to the too high cpy frequency of 17*200=3400MHz or higher. Also, I'm curious as to wheather or not the Pentium-M 780 may be capable of reaching higher overclocks than its 'slower' sister Dothans that have the same core yet are officially rated for lower end cpu frequencies.

If I'm correct, then I believe that a turbo-pll might also solve that problem, namely(?), for example as following. Let's say that the signal of the onboard clock generator crystal would be replaced by one that has a 10% lower frequency than the default one. In accordance with that, this will obviously lead to a 10% lower overall system speed. The turbo-pll could then be used to parallelly generate the default frequencies of for example the USB signal and system time clock signal, so that it would stay possible to control the PC and also no longer be needed to keep on correcting the system time. Anyway, when the overall system would be 10% lower, then the 200FSB setting in BIOS will lead to that the FSB frequency will be 200MHz - 10% = 180MHz in reality, and to the not-too-high-to-run cpu frequency of 180 * 17 = 3060MHz. But (and this is what I am wondering about being true or not) while still maintaining the whatever cause for the 200+ FSB BIOS setting being superior to a BIOS setting within the range of 167FSB-200FSB(?) And then once the PC has booted and Windows is running, the multi can be lowered and the FSB can be increased 'back' to somewhere in the (real) 200+ FSB regions again, right?

By the way, yes, it will also lead to an initial 10% drop of other frequencies too, like for example that of the PCI which will then be 33MHz - 10% = 30MHz PCI. But to stay with this example of the PCI frequency... it's possible to compensate that decrease in frequency by actually setting & locking it to a higher frequency in BIOS: 36,3PCI in BIOS corresponds with 33PCI in reality (because 36,3MHz - 10% = 33MHz). But even when not correcting this, 30PCI wouldn't lead to trouble either for as far as I know. I've ran the PCI at 30MHz many times before, with several different types of mainboards and during long periods of usage and so far I've never experienced any problems with that at all. It can be interpreted as off-topic, but to be more specific: for example the Asus P4T, P4T-E and P4T533-C. These are old RDRAM legend boards that didn't have AGP/PCI locks and easily adjustable PCI deviders available. I've ran the PCI on 30MHz on those boards for a long time of usage, in order to get significantly higher FSB speeds (150FSB) at the cost of slightly lower AGP/PCI frequencies (60AGP/30PCI), while at the same time preventing that the AGP and PCI frequency would get too far out of specs while doing so. Never have I had any problems with that occuring.

Anyway, before getting even more off-topic and even nostalgic... Is there anyone here on this forum, who is able to summarize and/or correct some of the information that I'm missing or not fully understanding well? And is there anyone who can tell me wheather or not this turbo-pll is all worth considering bringing into practice? Oh, and please quit bombarding me with private messages about it... NOT! ROFL! Thanks in advance, for any possible useful feedback.