Dothan @ CT-479 - howto and fixes.

[Jericho]

Has anyone else experienced a brick wall with the FSB for boot/post ?

I'm hitting a wall at 245mhz FSB 1:1 (2Gb Ballistix , M740 , P4C800ed & 1023.001).

Its 100% absolutely error free at 245 in memtest(9hours full tests). But will not even post at 246mhz (regardless of multiplier used).

BUT - I can change the FSB in windows to 250 and its 100% rock solid with all tests ?

Anyone else experienced this.
Im thinking its just a BIOS bug.

[dasa]

yer i have p4c800-e latest beta bios and cant post over 226fsb but it works fine in windows at 250fsb..... even if i try to restart my pc with it semt over 226fsb with clockgen my pc wont post the sata hdd's dont spin up and the screen is off but everyting else seems to work so maby its a pci lock problem?

[Jericho]

yer i have p4c800-e latest beta bios and cant post over 226fsb but it works fine in windows at 250fsb..... even if i try to restart my pc with it semt over 226fsb with clockgen my pc wont post the sata hdd's dont spin up and the screen is off but everyting else seems to work so maby its a pci lock problem?

Just ried that - same thing. No re-start. Have to shut down and start up.

[vegeta]

I have the same experience with the p4p800se. I can't boot over 239.
And in windows without a problem. It's kinda annoying when you use memtest. If you do, you have to use dividers to get higher DDR Mhz's

[sky]

I dont have my Dothan sys but you can try this and see if it works. You need to disable Fastwrites in bios and in windows on GPU using RivaTuner. Actualy try first RivaTuner disable GPU Fastwrites then OS Fastwrites. It would decrease performance a bit but not much and should make oc'd system more stable. Would be intresting to know what gives more peformance -no PAT with FW or Pat on and no FW.
Hope it works

thanks, i'll try that when i get back home today.. will see what it does.
only thing i can say is that withou pat (pat being completely off) performance
is noticeable down from pat fully enabled or even partially enabled. partially
enabled happens at fsb > 200. cpu-z says it's enabled, well that's kinda true,
but using ctiaw you'll see it is only partially enabled. unless you can up the
ram to run at ddr500 on booting, you're not going to see pat again with these
high fsbs... when pat is off, i have to up the fsb quite some to get close to
what i got with it turned on. i even put a former cpu-heatsink on the nb...
no deal. if anything it worsened the oc (at least that's how it seems).

anyway.. what struck me as odd was the difference between the last gen
ati card (9800xt) and the well (now) last gen nvidia (6800gt). sure the nv
is faster overall, but what's that got to do with the nb and all? i mean, almost
40mhz of fsb-power lost only by using a different videocard - that seems
rather odd. where's the connection? i mean from the performance feedback,
3d2k1 offers, the 9800xt (when overclocked) wasn't that far off compared
to the gt, except of course in nature or dx9 stuff...

Check out this post: http://www.abxzone.com/forums/showp...8&postcount=320
I have a p4p800 pat hacked and it stops artifacting once I change performance mode to 01. Give it a try. Too bad it can't be changed permanently.

I hope I can upgrade to a dothan once I save up some money.

thanks! another thing i will try. i can't view the pic you linked to w/o regging
there, but i guess i know what you want me to try. so far i've been running
the tweaker to change the 2 settings below the "normal" timings. i don't
know what that setting is at right now, but i'm pretty confident it is at 00
right now. we'll see if it helps the situation. because what's the use of high
fsbs when you can't really use it (in 3d that is).




anyway.. another thing i found odd with my dothan combo was that when
i am running win2k, i can put out real good 1m and 2m superpi runs. but once
i try 32m, my times go to hell. the same config that can run a low 26 will take
a good 48m (!!) for 32m. that's win2k sp1. with the same setup and almost
identical settings (main stuff is there, memory timings, yadayada) on a winxp
pro sp1 install (former amd xpm install, blatantly misused with shot direct x9
install and all) i am able to pull low 29m for 32m. that is still not great, but
one hell of a difference. anyone know what could be the cause?
the win2k install was brandspanking new by the way... got the same slowish
result with the win2k install of the old xp-m...

[Highland3r]

Has anyone else experienced a brick wall with the FSB for boot/post ?

I'm hitting a wall at 245mhz FSB 1:1 (2Gb Ballistix , M740 , P4C800ed & 1023.001).

Its 100% absolutely error free at 245 in memtest(9hours full tests). But will not even post at 246mhz (regardless of multiplier used).

BUT - I can change the FSB in windows to 250 and its 100% rock solid with all tests ?

Anyone else experienced this.
Im thinking its just a BIOS bug.

Would agree on the bios bug. Cant boot over ~ 225-230 here.
Whats interesting is that if you boot to windows, then restart the pc with the fsb set below the "no boot" fsb (if that makes sense) it'll restart fine.
If you set the fsb in windows ABOVE the "no boot" fsb then it WON'T reboot properly. It just sits there at a black screen.
ie, boot at 210 fsb (standard boot with 2*256, 12*) the reboot, everything will be fine. Boot at 210 again, drop multi and set 255 fsb, it'll run tests fine, prime95 and whatever. However restart the pc (shutdown is fine) and it hangs on boot.
Sounds very much like a bios bug. Have played with the 400/533 jumpers that didnt fix, very strange problem!

[olaf63]

Big ????? also here, since I had a vga monitor connected I managed to work rock solid at 2612 mhz (201x13).
Then I bought a new 19" lcd monitor with dvi and from there I experience many craches and I had to lwer the cpu frequence to 2520 (210x12).
The oly change was DVI connection usage!!!!

Ciao Olaf

[dqniel]

This is a bit off topic-

What is the difference between a 478 pin and 479 pin 533 dothan? I mean, obviously, there is the difference in the amount of pins...but are they both compatible with the same things? Was the 479 variant just a later version created to stop morons from trying to put 478 P-M's in socket 478 boards?

Differences between these two? http://processorfinder.intel.com/scr...+on+selections

[marvt74]

Has anyone a guide to fit a mach 2 to the ct 479? I've read that some modding is needed just need to see how much is needed before i splash out

[sky]

while it's true the dothan *socket* has 479 holes, a dothan actually has only 478 pins
and your link shows a dothan with pins (yup, that's the 478 thingy) and a dothan with
balls (bga - ball grid array) - that's the 479 "pin" type.
the bga-piece needs a special socket or a special adaptor to make it run on the asus
adapter. the other is just your regular dothan. i'm running a sl7sm (760, ppga) and a
sl7em (755, ppga) and boy they don't have balls, but pins and work without troubles
in your regular i855 mobos or on the asus adaptor...

@marvt
look around here, it's either in this very thread or in a dothan thread close
by, where hipro showed his prommi mount on a dothan setup.

[dqniel]

while it's true the dothan *socket* has 479 holes, a dothan actually has only 478 pins
and your link shows a dothan with pins (yup, that's the 478 thingy) and a dothan with
balls (bga - ball grid array) - that's the 479 "pin" type.
the bga-piece needs a special socket or a special adaptor to make it run on the asus
adapter. the other is just your regular dothan. i'm running a sl7sm (760, ppga) and a
sl7em (755, ppga) and boy they don't have balls, but pins and work without troubles
in your regular i855 mobos or on the asus adaptor...

Thanks for the clarification. I just bought a new 478-pin 730 for $152 on ebay and I just wanted to make sure it would work with the CT-479 before I bought it

Now it's just time to decide on the $90 p4p or $170 p4c :\. Seems the p4c is getting better results but I don't know if I wanna dish out the extra $70...I'm a poor college student you know Seeing as though I have junk ram at the moment, I'm leaning toward the p4p since I won't be able to pull amazing speeds with Crucial Value RAM regardless of nb capability.

By the way, I came across this discussion of PAT performance increase a few minutes ago while googling. Kind of interesting, but I don't know how relevant it is: http://www.xbitlabs.com/articles/mai...-p4p800_8.html

[Jericho]

Highland3r

Thanks , then we can confirm its a geniune bug thats affecting everyone - hopefuly it will be fixed with a BIOS.

dqniel

P4C800ed without a doubt - save save save . Its worth waiting an extra week.
Its not just the fact the P4C is slightly faster, it is also a better layout.

[dqniel]

Highland3r

Thanks , then we can confirm its a geniune bug thats affecting everyone - hopefuly it will be fixed with a BIOS.

dqniel

P4C800ed without a doubt - save save save . Its worth waiting an extra week.
Its not just the fact the P4C is slightly faster, it is also a better layout.

I make 70 in one night of work, but I save for other things...like the ridiculously priced gas that gets me to work along with college tuition. Who knows, maybe I can find a used p4c800ed for only a bit more than a new p4p800se.

[Jericho]

I pulled a bare P4C800ed board off ebay for about the same as a new P4P800se, so yeah its possible.

I have used both P4P and P4C btw with this dothan, so Im talking from my own experience.

[marvt74]

Thank you Sky, my xtpe borked so seems like a good time to switch since i cant use that as an excuse not to anymore

Bring on the dark side

[sky]

update regarding my post from earlier today. i guess it is time to spread some
thanks to railer and ironcheff. being the lazy smartass (?) i am i incorporated
both their recommendations in one go and went upping the clocks.
so with both fixes in place i was able to up the fsb once again to >300 and was
actually able to bench aquamark3 at 9x306 (2754 mhz) with the geforce mildly
clocked at 400/1150. so i guess we can state that this works. which of the two
fixes finally made this possible, i don't know. suffice to say that it works. this
was done running the mem 3:2 (not good). anyway, i had to ump the pci-agp
fix up to 36/72 in the process, since the rig went into supercrawl mode once
i passed 304 mhz fsb at those settings i used.

now what i need to figure is how to make this work with my ram. so far i've been
able to go up to +240 on the ram - which really isn't that much of a feat, con-
sidering it's bh5 (2x kingston 512mb sticks, them being doublesided doesn't help).
but then again, with this p4p800-se my vdimm drops considerably under
load. vdimm mod is done via lm358 and resistance is set to 0, so whatever is
left of the 3.3v rail after some regulation, goes to the dimms. by bumping the
3.3v to about 3.5v on this last try, i had about 3.27-3.3v working the rams
under load. idle is ~3.4v...

i need to stabilize the vdimm, because 0.12v max droop under load is not to
be tolerated. makes benching some sort of a gamble. ideas to make this
happen are appreciated. btw psu is an ocz powerstream 520w - and i doubt
its 3.3v line takes a dive under load, when it didn't buckle running my xpm
@3.1ghz loaded.

[bhoward]

hipro5 .... and sorry for the long post

If you don't mind, I have a few questions about your Prommie mounting ....

1) What did you do for the back?
2) Are you using the back heater pad? (i assume you are because of the low heat output of the Dothan)
3) When you attached the Intel cradle, did you add any spacers to raise it a bit? I was think of doing that to decrease the force being applied by the CT-479 heatsinks holddowns.... I read that it is a lot.
4) Those are the holddowns from the CT-479 heatsink?
5) Did the heatspreader allow you to apply more pressure than normal?

Here is my plan

1) Using a #8 nylon screw thru the Intel cradle, washer(or 2 or bolt), thru motherboard, thru the Prommie backplate capped with a washer and nut.

This will raise the cradle holddown notches around 1/8" to 1/4".

2) Apply a thin foam tape around the processor core to act as a shim

3) LOTS of seal string like you have

4) using the same hold downs from the CT-479 heatsink.

5) Add another layer of neoprene all around the Prommie head

6) Add another layer of neoprene on the back

I am a little afraid of condensation because of the low heat load geneated by the Dothan ..... even at 3.4g (i wish! )

I am hoping for 2.9 to 3.1g from a 730M on a P4P800SE for everyday use.....Dreaming?

[illmatik]

Man, I wish I didnt just throw out the heatspreader from a busted p4 last week after seeing hipro's mounting trick. I guess I can scalp this 1.7 that I was going to send out for pin repair seeing I dont have any other p4 boards to use it in... Nice one hipro, now I have mounting methods for both my vapo and my mach II!! Now I only need another 740!

[sky]

2) Are you using the back heater pad? (i assume you are because of the low heat output of the Dothan)

just some thought.. i guess the heatpad on the rear of the board may not do
anything to avoid condensation, as it is a good way off where the real cold is.
think of how the setup would be:

on top you have the evaporator of the prommi, below that maybe the ihs of
an old chip or some other means of stopping the evap from tilting on the tiny
dothan die, thus chipping it. then the chip itself, after that the adaptor.
up to this point it's more or less a "regular" prometeia install. BUT the adaptor
has a socketlike thing on it's bottom that fits onto the actual real socket of
the board.

what i'm trying to say is that you have a huge area of air underneath the
adaptor that needs taking care of before thinking about insulating the actual
rear of the board. if you don't insulate up there, like hipro did, you may not
need to put anything on the back of the board. overall you have about an
additional 0.8-1.0 cm of material (adaptor) between the backside of the mb
and the evap head.

this is just a thought. may not mean much, but might be worth thinking
about

cheers

[bhoward]

sky

The rear of all motherboards must be insulated whenever you are colling with sub-zero temps. The CPU --> it's pins --> adapter --> it's pin --> socket --> it's solder provide a direct channel for the cold to the read of the motherboard. If not insulated, frost will quickly appear there and kill the motherboard (at least!).

I am asking if the rear heater is being used to finish evaporating the coolant so that no liquid is returned to the compressor which would kill it in time. i.e. make the CPU feel hotter to the prommie.

My 2.4a is plenty hot enough to achieve this since it is providing around 140w (my guess-a-ment). The Dothan under full does around half of that.

[dqniel]

does anyone else get graphical glitches in 3dmode when pat is enabled running past a certain fsb?

my setup is the following, all stock aircooling for the time being:
760 on p4p800-se, bios 1009.003
2x 512 khx3200 bh5 @3.3v
pov gf6800gt with default ultra-heatsink
ocz powerstream 520w

so with PAT enabled i used to boot at 167 1:1 and upped the fsb in windows. i could go well past 220 that way, stable and all. just when i ran am3 - because it always was the first to show glitches - i get some artifacting like the gfx-card gets to little juice, black triangles and all. that is at stock clocks on the gfx-card. so i started dropping the fsb in 5mhz steps until it reached around 180 where it would almost stop. iirc 178 was the setting where it finally ran without artifacts.

back when i ran the same speeds (fsb, ram & overall clocks on the cpu) with my 9800xt, i could do fsbs of >220 mhz and would get no such errors.

so then i decided to step it up some and booted at 201 mhz fsb with PAT disabled or partially enabled. that way i could up the fsb to almost 290 mhz (mem 3:2 or 5:4 - didn't matter) without getting glitches on the geforce. only when i went even further, like 300 mhz or even 310ish on the fsb, i would get artifacting again - still with the gf at stock clocks.

then again, i could also boot at 251 mhz 1:1 with PAT enabled (since on the p4p800se you get pat back on when selecting ddr500 as ramspeed). that would work good for everything BUT graphical apps. i haven't tested how far i would need to drop the fsb to stop the artifacting, but my guess is that the border for that happening would be about the same 175-185 mhz fsb it used to be before with the 6800gt installed.

this seems rather strange to me, but i have heard of this from other folks around using current gen cards - not necissarily all nvidia. some x800's where in the pack iirc.

so if anybody has a workaround for that, i'd really appreciate that. from the looks of it, this seems to be somewhat nb-related, with slacker timings or something on the nb when PAT is disabled. even increasing the vagp didn't really do any good...


/edit

this phenomenon is indepedent of the cpu-clock. with PAT enabled i got it at 2800 mhz (200x14, 1:1), 2400 mhz (200x12, 1:1) or even 2000 mhz (200x10, 1:1) using the same fsb-speeds and voltages for all runs or at 2760 mhz (306x9, 5:4) with PAT partially enabled...

I was recently told by somebody that it might be a problem with 2x512. He said that apparently Dothan's, no matter what you do, just don't like 2x512 :\. He said 2x256 is the way to go if you want anything respectable. Maybe you should see if you have some 256 sticks lying around that you could test to prove this theory

[hipro5]

Has anyone a guide to fit a mach 2 to the ct 479? I've read that some modding is needed just need to see how much is needed before i splash out

The "simplest" way........

[hipro5]

hipro5 .... and sorry for the long post

If you don't mind, I have a few questions about your Prommie mounting ....

1) What did you do for the back?
2) Are you using the back heater pad? (i assume you are because of the low heat output of the Dothan)
3) When you attached the Intel cradle, did you add any spacers to raise it a bit? I was think of doing that to decrease the force being applied by the CT-479 heatsinks holddowns.... I read that it is a lot.
4) Those are the holddowns from the CT-479 heatsink?
5) Did the heatspreader allow you to apply more pressure than normal?

Here is my plan

1) Using a #8 nylon screw thru the Intel cradle, washer(or 2 or bolt), thru motherboard, thru the Prommie backplate capped with a washer and nut.

This will raise the cradle holddown notches around 1/8" to 1/4".

2) Apply a thin foam tape around the processor core to act as a shim

3) LOTS of seal string like you have

4) using the same hold downs from the CT-479 heatsink.

5) Add another layer of neoprene all around the Prommie head

6) Add another layer of neoprene on the back

I am a little afraid of condensation because of the low heat load geneated by the Dothan ..... even at 3.4g (i wish! )

I am hoping for 2.9 to 3.1g from a 730M on a P4P800SE for everyday use.....Dreaming?

1. I used the Promie's default back panel but crossed with the holes.......Just with the seal string like an AMD mounting.....
2. Yes....
3. No spacers at all....
4. Yes they are...
5. I gained almost 50MHz more on air with the heat spreader and about .....nothing on phase change........

You CANOT use a heat spreader when using a prommie and the holddowns from the CT-479 heatsink....Too high and it won't fit....I change the mounting of the prommie now....I'll post some photos later on about that.......ALSO when trying to pull out the prommie's head from the Dothan when I had the seal string, I pulled out the CT-479 adaptor too and broke two pins of it......Soldered back....
So no more seal string for me anymore

I wish you to hit over 3100MHz with the prommie.......I couldn't managed to do so.....At 3025MHz I can bench ANYTHING - even 3D - BUT at 3031MHz I got freezings......I've changed mobos from P4P800-SE to P4C800-E and the exact same things happens.....3025MHz MAX no matter what.......From -20*C full load to -32*C full load........NOTHING........3025MHz MAX.......

P.S. P4C800-E mobo is way TOO faster from P4P800-SE........

[bhoward]

Thanks for the info hipro5

Hoping to see your new pic of the mounting

PS If you raise the cradle using a nylon washer between the motherboard and the cradle, you could ease the pressure applied to the CPU allowing you to use the heatspreader...??

[Highland3r]

Highland3r

Thanks , then we can confirm its a geniune bug thats affecting everyone - hopefuly it will be fixed with a BIOS.

dqniel

P4C800ed without a doubt - save save save . Its worth waiting an extra week.
Its not just the fact the P4C is slightly faster, it is also a better layout.

Ran some more tests on another Dothan system (another P4C800-e Del) same thing, 250 boots fine, 255 wouldnt boot, system runs fine in windows to ~ 275. As tested before, resetting the system at 275 = no boot. Rebooting its fine. Resetting at 250 though and all is fine
No boot on 8* multi on this system either.