-=Dfi Infinty&LanParty Rev B Faq and more=-

[The B]

Hello!

This is my firs post here. I’m having trouble with my Ultra Infinity. Specs:
Ultra infinity 230 MHz @ 1.6V Taiwan
2x512 Twinmos SP 230 MHz @ 3.2V
Mobile barton 2530 MHz @ 1.825V
6800 GT 450/1190
Tagan 480W 3.3-28A, 5-48A, 12-28A

All the main mosfets are sinked and system is water cooled.

At these settings it can prime for 13h. And so is stable. But as soon as I increase Vcore to 1.85 or more it freezes. Prime or any other program never report errors. And the freezing is random usually when surfing or when I’m away from computer. It used to be able to take 1.85V. What do you think might be the problem? Caps? All the big caps around the cpu mosfets are brown. If you want I can post a picture of the setup…

B

[trodas]

coop - thanks I now trying to insert as many caps on the board, as I can and some make the mobo unstable I managed to isolate 5 of them anyway, there for sure, unsure about two of them, however I think (as they aren't into the sensitive area) I should leave them alone I still waiting to get my SMD clearing caps and stuff for testing of my resusciated mobo, tough

About tweaker - well, it works well for me, however the auto-start / auto apply settings did not work on XP SP2 as well, as on XP SP1.0a. I looking for a fix, but unsucesfully yet...

Sorry for you burned one, mate. But I also have rev. A one. However the mosfets is reported to have the same specs (the last digits are package and minor stuff only, at least from what I hear), so the only way to explain the better overclockability of taiwan ones are the caps. But then again, in the vital areas, they are same. And in the less interesting ones, the taiwan ones used totally cap Teapos, while China ones use better OST ones. At least according to the specs.
Frankly it means - worser specs = better overclock. Insane as it sounds. So, either the Teapos are't as bad and the OST specs is way off the reality, or there is something else. The cap issue will be determined sooner or later, as I hope soon I'm get the rest of the caps and I will replace everything with hi-quality Panasonic FM caps. So, if my overclock get worser, then it explain everything. But I believe my overclock will be far better, so... we have to wait for the damn caps to see...


The B - hello there. I bet it is the caps, but a nice picture of them is recommended. If something just freeze, it is not the Vcore, but rather the AGP/PCI power supply caps (the ones that failed on my mobo too)... Sinked mosfets did not do you any good, unless you have fan blowing air on them.
I has completely heatsinked mobo, run fanless, and problems are only worser. Removing heatinks helped there.
So, close-up (and write us the type of the caps - eg. what is on them printed) on caps 1 to 3 and next on caps 7 and 8 according to this my draw:



Are necessary. Be prepared to solder them off and replace them with better ones, tough
It is a old mobo now, you know. Caps die.

[The B]

Thank you for your answer.

Caps 1-3 are 6.3V 3300 uF and 8-7 are 16V 1500uF. 9, 34, 32, 31, 33, 30, 29 are Teapo. The thing is that 9 (which is Teapo) looks just like 1-8 only smaller. But on 1-8 there seems to be no manufacturer written.

Pictures:
http://img204.imageshack.us/img204/8...ture0039mu.jpg
http://img128.imageshack.us/img128/9...ure00414jx.jpg

B

[The B]

So? What do you think is the problem? Which caps need to be replaced?

As you probably know the board is completely irreplaceable. So I don't want to mess with it unless there is a good chance of it working. I just want to prolong its life for a year (maybe less). Then I will get myself a proper system. If I kill it I’ll have to settle for Asrock 939 dual sata2 and a 3200+ Venice.


B

[trodas]

Sorry, getting burned with own reworks Still waiting for the final caps, that SUXX ...

Cap's 1-8 are KZG ones, and only Chemi-con made KZG line of caps, so... This is pretty easy to identify. Teapo are one of the worsest brands even in the bad caps brands - no wonder you having problems.
Cap 9 is used on the Taiwan mobos as Chemi-con as well, 1000uF 10V one. It is a cheap crappy OST on China mobos, witch may explain something about the overclocking ability, tough it has to do something with "just" SB and stuff, so I would never quess that it could be so important. Neverless I put there a 2200uF Panasonic FM and even bridge him from bottom 100nF ceramic cap for best filtering ever.
Such quality voltages my mobo never enjoyed even as new - ever I expect it pay me my work back with nice and stable OC

So, you problem is? Mobo acting weird as soon, as you go to or over 1.850Vcore, IIRC? 1.850Vcore is the breaking point of the voltage regulator used. DFI moded it to go higher (2V and up with futher moding), however the mod seems to be failing for you. Since this is impossible, as no other that ceramic caps and resistors aren't used in this mod, I would like to think that you got the same issue as me and your caps finally giving up.

And yea, this board is not possible to replace today. Either you make it work, or you screw it up - you still can't use it well, as it is - or do you manage?

And well, replacing caps (if you can solder well) can't kill a mobo - unless you use QUALITY ones. The Panasonic FM I used are the absolute low-end. If you go to anything lover, your board die. As simple, as it it. I once added 3 not-good-brand caps to the Vcore ones in parallel: http://ax2.old-cans.com/s.php?p=badt...3&c=8&d=1&v=v2
...and under 20min, the mobo is gone and I still haven't fixed it, even after replacing the mosfets... Now is time to replace the Vcore caps as well, but the chances aren't big, IMHO. It looks like something in the mobo PCB was possibly destroyed or something...

Anyway, well, oscilloscope is on it's way, eh. We then SEE what I did to the voltages

And it is time for serious modification! Now - who need COM, LPT and SP/DIF ports, when use X-Fi?



It create more space for heatsink as well, as make the airflow better. So, why not?

I already get bored and removed the sound output ports as well, hehe. I did not even use them, since I using Audigy2zs and soon X-Fi Fatal1ty, so... And they does restrict the airflow, so for the sake of better cooling - off they go. And the board is still functional, LOL
Above this, I also removed every the remaining ten pieces of OST (one smaller Teapo between them, a 22uF for 25V one) 100uF caps between the PCI slots and one near the main power cord from PSU for 120uF 16V Panasonic FM ones. They are bit bigger and more bulky, tought I hate the idea of caps, marked as BAD CAPS and known to be prone for FAILING MISERABLY are on my mobo. Easpecially the cursed Teapo went off the mobo with big kick in the ****...

Now the only one caps that are originals there are the cursed 6 caps witch are going to be replaced as soon, as I get them and other that that - there is 12 10uF SMD caps, this one right to the bulky Panasonic:

...and on it is writen 3t, 10, 16V.
Since they are near the AGP, around the sound output and some around the SB I did not think much about replacing them, but I have 10uF 16V ceramic caps, witch will do better job for sure - ceramics is always better that electrolyte ones, by their far lower ESL and ESR (means also more ripple current they could deliver...). But it is worth? Can they be failing?
Especially those near bellow the AGP port will be very hard to remove/replace...

[trodas]

Hi guys. Using the Mushkin Hi-Perv Lev.1 ones, 2x512MB with BH-6 chips.

Or so I was told on the website, not tried to take down the heatspreaders myself to take a peek...

I wonder if anyone can suggest a good drive strenght and slew rates for my two used dimm slots 1 and 3

For empty ones I use recommended 1/10 (strenght/slew), however with AUTO values for the used slots I maxed up at 249/250Mhz. Semi stable. Sometimes it pull trought the bench, sometimes it does not.
Tried 4/6 (IIRC the strenght/slew should be more agresive for 2x512MB sticks) but hexus PI fast benchmark ended up with horrible BSOD

Suggestions wanted

[VulgarHandle]

Alright all you veterans, adventure in the planning.

Goal 1: Waste time untill K8L arrives, so I can decide between K8L and Conroe
Goal 2: Break as many records as possible for Socket A before it becomes the wife's internet virus collector.


Got 4 Aces up the sleeve.

Ace of Spades: Jinu117 VapoLi system
http://webpages.charter.net/jinwpark...e/ProductionA/
Ace of Hearts: Gainward BLISS GeForce 7800 GS+ SILENT 512MB GDDR3 TV-Out/Dual DVI (AGP) (24pipes, 7900GT GPU)
http://www.overclockers.co.uk/acatal...e_AGP_411.html
Ace of Diamonds: 2x74gb 16mb Western Digital Raptor (maybe 4x37gb 16mb WD Raptor if they come out in time)
Ace of Clubs: 2x256 Corsair PC3500 (BH5)


Got 4 Jokers in the deck.

AMD Athlon XP-M 2500+
AMD Athlon XP-M 3000+
AMD Athlon XP 2700+(tbred-b)
AMD Athlon XP 1800+(tbred-b)


Got 1 Rules card.

DFI Lanparty NF2 Ultra-B

Some of the rules include:
1. TicTac 6/19 bios
2. Vcore Vmod
3. Vsens Vmod
4. All sinked up
5. Water cooling/ramsinks for vidcard


Now, if anyone wants to add some rules, let me know. I won't be doing the Vdimm Vmod, as I don't think I'll need more than 3.3v, never have to this point, and I can't solder, nor do I know anyone that can. The other Vmods are done with grabbers. I also have no plan to replace caps.

Thanx for your input,
John

[coop]

Well, if you want to get some records, you have your work cut out for you. You can have a goodtime with it though before it becomes "the wife's internet virus collector". (I like that)

[trodas]

I dubt he can score very much, w/o replacing the aging useless caps, and since he can't solder, well... Other than that, the 3DMark01 SAE record is IMHO easy to break. Just get into 260MHz FSB, 3Ghz+ CPU and 7800GS overclocked card and the record is ours
Planing to get into that too, just not having the time right now... But can claim No.1 in AXP 3DMark01 in sig could be nice

[Max Tractor]

are the 4 sata ports locked on the infinity

[trodas]

VulgarHandle - so, what's the progress?

Max Tractor - sorry, don't know. Never used SATA, hate it...


TWO questions for all


1) What is the APX O/C WR (World Record) clock on air?


2) Speaking about mods, how much resistance one should add there:



To get +0.2 - +0.25V apart from what one set in bios?
I mean - how much give 0.1V or something, so I should just solder there a resistor and not a VR (Variable Resistor), as all I need is about 0.2V raise for extreme OC

[VulgarHandle]

awaiting socket a brackets from jinu

[VulgarHandle]

killed my mobo...done with socket A...dream lost '4x4' can't get here soone enough

[trodas]

Damn, that suxx. What did you do? Might could be repaired, tough?


BTW, anyone have some answers to my questions?

[Sharptooth+]

I have the same question as Trodas. I'm trying to stabilize 250FSB (10x) with my LanParty NFII Ultra B using Mushkin Redline XP4000 (2x512) UTT CH-5 in dual channel mode. At this time I'm using Hellfire 3EG Rev3.

I get as high as 236FSB and 240 is somewhat stable when I loosen the alphas. However, 250 still remains elusive even with loose memory and alpha timings.

At this point, I'm testing each Mushkin stick individually. I just want to see if either stick can run at 250FSB (10x) and pass SuperPI 32M.

For far, I've tested one stick (slot#1) and it passed. I'm testing the other stick and when I get a result I will post back.

With one stick I used a vdimm of 3.3, Vdd 1.8 and 11-2-2-2-2.0;13,15. I left the alphas and drive strength and slew rates on auto.

It seems to me that running in dual-channel with these memory sticks may simply require more than 3.3v. But if the other individual sticks passes SuperPI 32M, then I have to believe the northbridge memory controller simply can't deal with the load of two of sticks at 250FSB.

That's the theory at this point.

--Fernando

*Update* The second individual stick also completed SuperPI 32M. So, both sticks (individually) can do 250FSB (10x) with SuperPI 32M. This was done with each stick in slot#1 (closest to CPU). Now, I'm testing both sticks in slots 1&3. After I get a result, I will post back again.

*Update#2* Using dual-channel mode (slots1&3) SuperPI 32M failed in the first loop. This occurred even when I bumped up the vdd from 1.8 to 1.9. Vdimm was 3.3v. This appears to support the argument that not all northbridges are the same. It seems there are some variations because some LP-B users out there can do 10x250 with UTT CH-5 or BH-5/BH-6 just by 'burning in' the ram with command per clock enabled. I did try with cpc-off and it didn't make any difference. 250FSB (10x) in dual channel is not going to happen under these conditions.

[Sharptooth+]

killed my mobo...done with socket A...dream lost '4x4' can't get here soone enough

How did you kill your LP-B?

--Fernando

[Sharptooth+]

Hi guys. Using the Mushkin Hi-Perv Lev.1 ones, 2x512MB with BH-6 chips.

I wonder if anyone can suggest a good drive strenght and slew rates for my two used dimm slots 1 and 3

Suggestions wanted

I was told that when the drive strength and slew rate settings were on 'auto' they are 4-11. But I haven't been able to confirm that myself.

Have you tried loosening the alphas for 250FSB? I've had some success using the parameters in sig I found from the forums.

For example try these:

8-2-2-2-2.0;14,16;5,4,5,3,3,5,4,;e,e,f,e;4-11,4-11,4-11

You could substitute the empty ram slot with 1-10 as you suggested.

Also try this:

8-2-2-2-2.0;14,16;5,4,6,3,3,4,5;e,e,f,e,;4-11,4-11,4-11

Again you can substitute the empty ram slot with 1-10 if you like or you can simply leave all of them on 'auto.'

Report back after you tried this.

Thanks!

--Fernando

[Sharptooth+]

Dropping the FSB from 250 to 245 (10x) didn't make a difference with SuperPI 32M (2x512 dual-channel with CPC-ON). It still failed (first loop).

I dropped it down from 245 to 240 (10.5x) and currently testing SuperPI 32M again.

Will report back with results!

--Fernando

[Sharptooth+]

Okay, 10.5x240 passed SuperPI 32M. The question now is, will it Prime for more than 8 hours.

I'll report back...

--Fernando

[Sharptooth+]

10.5x240 errored in Prime95 in 5mins. So, I'm back to 11x236.

--Fernando

[VulgarHandle]

killed it during inuslation process for phase change...

what kind of cooling do you have for your north bridge? i use a Vantec Iceberq, copper....

i also sinked the mosfets..while I could only stable 252fsb with water and radiator on a/c vent, it was still possible, using 1.8v chipset, 3.1v ram(corsair 2x256 bh5, dual-channel mode, cpc on, all timings in sig, and alphas/strengths at auto), and 1.950v vcore(252x10.5)

for daily use, you can see what i used in my signature

as my own update, i'm, ehem, awaiting funds for a new board(another lp nf2 ultra-b), only one i could find is a refurbished one...says it still has manuf. warrenty

[trodas]

Sharptooth+ - you obviously need to concentrate on two issues - Vdimm area and Vdimm mosfets cooling AND north bridge cooling. I getting the idea that your mobo w/o modifications in the caps and cooling can't handle prime stable more that 236Mhz dual channel. Remember, 2x512 is bad for nF2. All records are done with 256MB sticks for this reason...
Futhermore for the CH-5 stick are might the drive / skew settings inapropriate, but I can't suggest anything in this area.

VulgarHandle - sounds good that you don't giving up, dude

Yet no-one answered me, what is the AXP OC WR on air...

[Sharptooth+]

Sharptooth+ - you obviously need to concentrate on two issues - Vdimm area and Vdimm mosfets cooling AND north bridge cooling. I getting the idea that your mobo w/o modifications in the caps and cooling can't handle prime stable more that 236Mhz dual channel. Remember, 2x512 is bad for nF2. All records are done with 256MB sticks for this reason...
Futhermore for the CH-5 stick are might the drive / skew settings inapropriate, but I can't suggest anything in this area.

Actually, I have a Swiftech MCX159-A on the northbridge and my vdimm and mosfets all have Microcool heatsinks on them. So, it's not my cooling

2x512 can be bad for NFII boards but you would be surprised by some I know who can do it with cpc-on and 250FSB!

My particular board does not use the Teapo caps. So, I don't think I have to replace any of them.

--Fernando

[Lastviking]

Sharptooth+ - you obviously need to concentrate on two issues - Vdimm area and Vdimm mosfets cooling AND north bridge cooling. I getting the idea that your mobo w/o modifications in the caps and cooling can't handle prime stable more that 236Mhz dual channel. Remember, 2x512 is bad for nF2. All records are done with 256MB sticks for this reason...
Futhermore for the CH-5 stick are might the drive / skew settings inapropriate, but I can't suggest anything in this area.

VulgarHandle - sounds good that you don't giving up, dude

Yet no-one answered me, what is the AXP OC WR on air...

Hmm maybe this:





Cold air..i´m not 100% sure that this is the WR with air.

[trodas]

Sharptooth - yes, I too managed CPC on 250Mhz FSB with 2x512 2-2-2 timings - but it is not very easy to do and my mobo aren't standard anymore, so...





Lastviking - Pic not visible, please use http://www.slibe.com/ or ImageShack or anything reliable, thanks Anyway, the link suggest 2881Mhz? Only? I managed 2900 and it won't seems to be giving me any troubles, so if this is the case, I gotta get the Big Typhoon now and I want my WR then
Of course we probably need more research for aircooled WR on AXP...

Any estiminate of the resistor to increase the Vdimm by 0.22 - 0.25V ...?