DFI NF4 the UTT/VX/BH5 KILLER READ IF YOU HAVE DFI

[AMDnewbie]

I just experience weird things with the tmsp pc3500 now....
unable to work after memtest #5 90 loops and consistent freeze with #8.

after tonnes of errors with #5 after 90 loops...it can't even pass #2.

it can't be the MB since UTT CH-5 working with ease but this BH-5 ram just behave very differently with others i had tested so far....after memtest die...

http://www.xtremesystems.org/forums/...ad.php?t=63314

[pazza316]

Put it in the Orange Slots!

[coop]

OK i know i already posted this thaugh in another thread, but why did DFI even implement the 3.3 / 5 switch why not just run the vdimm of the 5v like some epoxes (im not a epox fanboy)

if it was for the inexperianced, they coould have hidden the high vdimm options through bios or just make us flash modded bioses with unhidden features..........

Because they saw alot of people modding their boards to be able to run higher voltages through the rams. Only having 2.8v to 3.2v or on some 3.3v did not cut it, and the increased performance must have been worth the risk of a mother board to many. They saw a need for some and accomodated. It states in the book that running the high side is risky, and I would not like to go back to cutting traces or soldering. I also know how it is to loose hardware that cost good money and that you develope a "likeing" to as well and I empathize. I think it may be a good idea to start a list of data from people who have had this happen and include Rev. # from the boards, maybe place of purchase, type of rams - both by manuf and chip type. I think it is early and not a good thing to blame someone for giving what is asked for.

[Millyons]

Naw you didn't get what i mean, i ment why not just run the vdimm of the 5v rail and have the options up to 4v for vdimm, i didn't say anything about lowering vdimm

[coop]

I acknowledge what you were trying to say. It could have been related to the heat issue running low volts from the 5v line. There may not have been another way to do it than the way they did and still be affordable.

[felinusz]

The DFI Lanparty 'B' and DFI Infinity nForce2 motherboards both fed VDIMM off of the +5VDC rail.... I think he's wondering why DFI didn't do the same for the nForce4 boards. The NF2 DFI boards didn't have any VDIMM related or VDIMM voltage regulator heat issues of any kind... what prevented DFI from implementing an identical setup on the nForce4 boards?

BIOS options allowed for up to 3.3V with these older motherboards. On my old DFI Infinity NF2, I ran my VDIMM at 3.6V cleanly with a stock +3.3VDC rail, after a simple VDIMM voltage regulator Potentiometer mod.



The UTT/BH-5 memory death at around ~3.2V is peculiar and troubling. I'll be sticking to the yellow slots and the newer BIOSes that work well with them after reading this .

[coop]

I can't address the question in your first paragraph. I don't know. It seemed simple enough and worked well, up to 3.3v. I did the vdimm mod as well. My rams are in the yellow slots and I'm using the 510-2 bios. That combo is working really good atm.

[felinusz]

It was rhetorical, don't sweat it

Yeah, the 510 BIOSes apparantly work quite well with high voltage memory and the Yellow slots that aren't causing RAM death. For now everyone's best bet is to just use the different slots and a newer BIOS.

I've got no clue why RAM is dying. Personally, I'm of the suspicion that Bacchus is correct... UTT is a gamble. The BH-5 that's dying seems to have been overvolted a very great deal, 3.8V for day-to-day use for example is perhaps too much even with excellent active cooling. I personally would not run my BH-5 past ~3.6V for day-to-day use, unless it saw large stable gains from further overvolts (which some RAM apparantly does!).



As for checking DIMM #2 on these boards for power spikes...

Checking that DIMM slot for a VDIMM voltage spike at boot is extremely simple, there's no need for anyone to wait for hipro5 to crack out his likely un-faulty board. You also really do need to check something like this on a board that is confirmed to have the RAM killing issue in the first place, or the readout is meaningless to the situation.

[Millyons]

I don't have DFI but im really interested in this thread, so someone who has it can check the exact voltage each slot is getting at these points.........this is for all DDR mobos

[hipro5]

It was rhetorical, don't sweat it
Checking that DIMM slot for a VDIMM voltage spike at boot is extremely simple, there's no need for anyone to wait for hipro5 to crack out his likely un-faulty board. You also really do need to check something like this on a board that is confirmed to have the RAM killing issue in the first place, or the readout is meaningless to the situation.

I had one and I sold it......I'll ask for esdee to bring his......So I just DON'T own a "likely un-faulty board" as you say....

[AMDnewbie]

just like to share....

my ram die in yellow slot after 90 loops of #5 in memtest. Die is in a sense that it can't be use anymore as it won't pass memtest in default anymore.

it were freezing in #8 memtest previously so i decide that maybe it need burn in to pass it...afterall its new bh-5...and all this happens on 200Mhz 2-2-2-5 only.

rma it and get second pair...

still unable to pass #8 on similar senario...this time no #5 loop anymore...
swtich to orange slot after some advice but still un-stable under cl2-2-2-5 no matter what i do...till i disable bank interleave then it can scale up to 235Mhz.

after some kind soul advice again....really great thanks...
I run this thing w CL1.5-2-2-2 w bank interleave....200Mhz fail again....but it works from 220Mhz and above...now I am at 235Mhz w 3.2v burning in...

so this utt chip might be the problem in the end....just my 0.02c..

[OzSnoal]

I have this problem too. The stick closest to the processor is the one that misbehaves for me.

If there is a voltage spike to 5V on power up onto the DIMMs, this could definately cause strange behaviour. Spikes are nasty things, depending on the rise time and the duration of the spike, it can cause latching of the 'FETs/transistors' which make up the ram - caught in the wrong state could even be damaging. It is definately not something that someone can say 'is unlikely to be the problem'. The spike should not be there, period. If it is, it's bad hardware design, period.

If someone could direct me to an easily accessable point to measure (don't give me a VDimm pin, way to small to get your probe onto), I have access to a scope and a multimeter.

[Shroomalistic]

Ive had the same problems but not on the dfi nf4. Mine was on the dfi ut 250gb. Killed 4 sticks of twinmos 3200 utt and 2 sticks of old school ch-5. I believe it was the spd chips also cause letting it sit for a day or 2 would let it run again but not for long. I could run memtest on them for days with no errors but as soon as I restarted they would boot till I pulled them out and let them sit for days in again. Now I have some newer twinmos utt and they have the larger spd chip in a different spot and havent had a problem with them. Maybe we should all check our dead sticks and see if they have the larger spd chip or the smaller one.

[mrlobber]

What's important is that when switching from 3.3v to 5v or the other way around, user should LOAD DEFAULTS first, so vdimm goes down to say 2.8-2.9v. Then go to BIOS and shut it down there, drain caps with POWER BUTTON. Then Clear CMOS, unplug everything, including power plug, switch the jumper.

Just to warn, exactly this procedure (clear cmos + simultaneous vdimm jumper switching) yielded me 2 dead ram slots on my DFI SLI-D that were closest to the cpu (one orange, one yellow). With dead I mean that the board didn't POST when memory was in any of these slots (tested both with TCCD as well as UTT), while the other 2 slots still worked fine at least to 270+ htt. So I personally wouldn't do these 2 things simultaneously anymore.

[Millyons]

If someone could direct me to an easily accessable point to measure (don't give me a VDimm pin, way to small to get your probe onto), I have access to a scope and a multimeter.

Its quite easy but you can't do it if you have heatspreaders on ram then its hard.........

[hipro5]

Take a look on the circle in this photo.....On your DFI there IS NOT a capacitor in that place......You can measure there the Vdimm.......

[Nugz]

WTH. they took one of them off to save a little cash? omg
that is a cap that helps reduce voltage spiks. aint it hipro5?

Im missing that cap Rev. AA0

Peace Nugz

[hipro5]

WTH. they took one of them off to save a little cash? omg
that is a cap that helps reduce voltage spiks. aint it hipro5?

Im missing that cap Rev. AA0

Peace Nugz

All companies do the same thing....

[eclypse]

Hell i have the same revision board and i 2 dont have that cap! Whats it for?

[hipro5]

Guys......Take it easy.......I HAVE soldered that cap........

It's also for Vdimm....

[Nugz]

i got some of those i on a Few (2) IC7's i have here and P4P800 and a P4C800 thats dead.

tell us if it helps with the startup voltage spikes.

Like i said. " hipro5 is on it, we might have a fix soon"

Maybe a hard mod. void the warenty. But i dout if it's the way they made the board they be telling everyone. " send your board to us and we'll fix it" <--- thats alot of cash for DFI to be going back and fix'n there problems.

So yeah. If it's going to be a hardmod to fix it. then so be it.

Thanks hipro5 for looking into this.

Peace Nugz

[felinusz]

Has anyone tried contacting DFI about this, aside from the person who started a thread on the DFI forums (the thread that was locked)?

I just PMed RGone on DFI-Street about this thread, hopefully he'll come over here .

[eclypse]

Guys......Take it easy.......I HAVE soldered that cap........

It's also for Vdimm....

heh nice.. Notice any differences with the extra cap?

[Magnj]

yea, should we add them in?. Also FYI Smart Guardian is off on my Vdimm by .06...which is a pretty fair amount.

Also is it suggested that i use 510 BIOS and yellow slots for OCZ VX?

[OzSnoal]

Guys, I believe I have found out why we are having a cold boot problem.

I posted this over at OCZ forums:
http://www.bleedinedge.com/forum/sho...9515#post89515

Working scenario:
Basically when you shutdown your PC and power it up again, Vdimm moves up from 0V to the voltage you set it to in the BIOS. (Measured with a multimeter).

Failing scenario:
HOWEVER, if you pull out the power plug, reinsert the plug, DO NOT switch on the PC yet. Hmmm, what is that 2.7V doing on Vdimm?? Switch on the PC hmmmm, still 2.7V? And of course you now have the cold boot problem. The Vdimm never reaches the value I have set in the BIOS.