PSU's 3.3v Rail directly connected to/as vdimm

[Rio71]

hi
i have follow situation
- fluctuate and to low vdimm
- 3,3v mod on psu works not on my hec300 lr-pt (link to thread )

theory...:
so i measured long time ago on an other board, and here again...
the vdimm-vrm using the 3,3v rail.
... but has no effect if vdimm-mod is high as possible. so vdimm-vrm is'nt not "in use" because its full through-steered.
only the vrm-own forward voltage (voltage drop) made a "lost" of around 0,15v - 0,25vdimm
for higher vdimm, modify the 3.3v rail is needed.
my thinking was "wy not bridging the vdimm-vrm and connected the 3,3v rail direktly to the dimm-slot."
so i soldered...

the mod:
... is simple. only connect the atx 3,3v (orange wire's) to vdimm-socket.
no opening trace or removing mosfet etc. is needed.
the wire sould be not to small. a lot of ampere flowing through.
(short calculation: "guess" 20 watt for 2 x 256mb made 6 ampere)
the little problem is: finding the correct pin(s) for vdimm !!!
(or the correct pin (leg) on vdimm-vrm-mosfet or vdimm blocking capacitor)

pro's are:
- stable and strong vdimm, especially under heavy load (memtest86 v3.0 #5)
- some psu's limiting the 3,3v-psu-mod with overvoltage protect.
- my psu (see above) need a deeper mod for upping the 3,3v rail...
- save power and produce less heat on vagp-vrm because 3,3v-rail is not so high.
- no heat on vdimm-vrm
- easy to mod

i have done this mod on asus p4p800-dlx and epox 4pdai. (see pics)
the red wire (along the socket) is not needed. it was only for research to "finding the last millivolt".
the vtt rising on both mobos "automatically" to half vdimm.
vdimm is now the same as 3.3vrail.
(fluctuating is pending from psu quality)
for me, in most win-applications 3,31v --> 3,31v and in memtest86 #5 3,31v --> 3,27v (meter on pin7)

again, watch out, vdimm is now possibly stronger and stable as before.
i will try this mod on is7-e too, but for a clean success an vtt mod is required.

Originally posted by enduracell
... I think though that a cap between 3.3 form PSU and groud before go to Vdimm will yield better results...

not needed. around the dimms are enough blocking capacitor's for this.
on p4p800: 3 x 1500µf and 1 x 1000µf










notice to measuring the real vdimm:
often asking... how measuring-point of vdimm ?
bards are differend, but the vdimm socket is every the same. (here for DDR1-Ram)
locate pin7 on any dimm-socket, here is the point to meter.
take a very closer view of socket and find pin 1, printed on it... count from here to pin 7.







disclaimer, disclaimer, blah blah.....

edit: bad english..
edit2: bad english..
edit3: updated/changed line "lost", "guess" and "ddr"

[enzoR]

some day if i feel crazy instead of doing 3.3, i'll do the same with 5v

[oc-rookie]

You are nuckin futs!!!!! Nice!!!

[enduracell]

This is one of the coolest mods I ever saw!
Thanx for sharing with us Rio71

[cold_ice]

Do you think it´s possible to do the same with the 5v rail? Then i don´t have to mod the PSU´s 3,3v rail, for more than 3,3v VDImm, right?

[Rio71]

Do you think it´s possible to do the same with the 5v rail? Then i don´t have to mod the PSU´s 3,3v rail, for more than 3,3v VDImm, right?

sure, bh-5 love high vdimm ... but 5 volt direktly as vdimm...
how xtreme are you ?

possible are adding in to the line of wire an series circuit of:
- two standard-diodes for around 3,6v
or
- one standard-diode + one shottky-diode for around 4,0v

[STEvil]

hm.. looks like this could be something to try on my 8rda+

I wonder if connecting the 3.3v directly to the output leg of the mosfet may be a better solution, though, since then you wouldnt have to poke around the RAM slots.

What do you think about that, Rio71?

[Slickthellama]

I love vmods like this. You guys are truly extreme...

[Rio71]

Originally posted by STEvil
I wonder if connecting the 3.3v directly to the output leg of the mosfet may be a better solution, though, since then you wouldnt have to poke around the RAM slots.

yes, connecting to the mosfet leg works too.
... if you know which is the correct one (leg)... this is most important.
but a soldering point of 3,3v rail is hard to find on top side of bord.

for me, i find (measured) the correct trace of vdimm and soldering on bottom side of board is much easyer.

many ways to do this mod...

[STEvil]

Well, I was going to use the AUX connector of the PSU and put the male end of the AUX plug going to the 3.3v mosfet ram mod and 5v mosfet mod so I can disconnect it all at leisure.. and have a use for that AUX plug... heh.

Thanks, though. I'll add this to my list of stuff to do... its getting larger by the day..

[cold_ice]

Originally posted by Rio71
sure, bh-5 love high vdimm ... but 5 volt direktly as vdimm...
how xtreme are you ?

possible are adding in to the line of wire an series circuit of:
- two standard-diodes for around 3,6v
or
- one standard-diode + one shottky-diode for around 4,0v

agree, 5v is too much
I hope I´ll have time to do this mod in the next few days

[enduracell]

I asked Çipro5 if I could connect the 3.3 line directly to the big mosfet that gives Vdimm on the �F7-S (the middle that reads the 3.3 line pin) and he told me that I could if the base of the mosfet doesn't measure volts when the mobo is off. So I mesuered the base of the mosfet when mobo was off, and guess what? it gives 3.3volts from the standby (I think its because of the ACPI feature ,S3 and the like). So I m wondering if would be a problem to connected to vdimm directly (with Rio way or on the upper left leg of the mosfet) and disregarding the fact that the mosfet takes volts when the mobo is off...

[Slickthellama]

Originally posted by cold_ice
agree, 5v is too much
I hope I´ll have time to do this mod in the next few days

yeah but at 5V i would pay to see what happens.. 1-1-1-3 @ DDR600

[enzoR]

LOL... i ment like a regulator circuit to produce adjustable 0-5v directly to the ram.... hmmmm makes me think about the lab PSU's in school..... you know, i actually saw some japanese dude change all his 3.3v lines to 5v.... always those japanese....

[Holst]

Jeff has done this before on a faulty 8rda+ he had.

One point to note you should remove or disable the VDIMM supply on the motherboard before attaching the 3.3v line.
Without doing so you may damage the boards voltage regulation and you may also introduce spikes or noise into the VDIMM supply.

It may also be worth soldering some extra capacitance over the VDIMM slots to cover you against spikes or drops on the 3.3v line.

[STEvil]

Holst - would removing the VDIMM mod (so the trace is cut) effectively do that?

I need to beef up the power to the mosfets that supply my CPU and redo my VDD mod before I play with this, maybe I should PM Jeff and make sure what i've got planned is right before I start soldering like a madman... heh.

[Dirty_Punk]

i have tryed it on NF7-S and works perfectly, no vdimm disabling, vddr up to 3.47v and no problem...
the only thing is that with VDD mod and VDD up to 2v i get only 0.04-0.05v drop from idle to full load, but with vdd mod disable and vdd set to 1.5v (the best settings for my mobo) i have a gret vdimm drop, like 0.12v from idle to load....
why?!?!?

[Rio71]

but with vdd mod disable and vdd set to 1.5v (the best settings for my mobo) i have a gret vdimm drop, like 0.12v from idle to load....

mmm... could be an "reading-drift" in your hw-mon software.
both have nothing to do with... normally..
try an multimeter... ?

yes, disabling the vdimm-supply and adding extra capacitance makes the mod near perfect.
an possible way could be simply lifting the right leg of mosfet(s)...

[Dirty_Punk]

Originally posted by Rio71
mmm... could be an "reading-drift" in your hw-mon software.
both have nothing to do with... normally..
try an multimeter... ?

i'll try next week because i can use the TEST rig only in the week-end... howether i'll try also to measure VTT to see if it is always at 0.5*VDDR or if it doesn't change...
howether gained about 12mhz in fsb thanx to this "stabilization" mod!!
with normal VDDR mod i get oscillation of 0.12-0.15v on vdimm, now only 0.03v!!!

[enzoR]

wow... good that its better for you!

[pelikan]

Originally posted by Dirty_Punk
i have tryed it on NF7-S and works perfectly, no vdimm disabling, vddr up to 3.47v and no problem...
the only thing is that with VDD mod and VDD up to 2v i get only 0.04-0.05v drop from idle to full load, but with vdd mod disable and vdd set to 1.5v (the best settings for my mobo) i have a gret vdimm drop, like 0.12v from idle to load....
why?!?!?

Have any pics for us please?

[Crankster]

Oh man, i think i'm gonna have to recruit some friends for a LN2 5 v into BH-5 killing spree with a Mobile... can you possibly feed the NB 3.3 volts? =)

[HKPolice]

Plz post update for IS7!!

[HKPolice]

Plz post updates, and confirmations?

[Boyne7]

could this mod be done on an a64 board so that you could run as much vdimm as you want (up to the max of the 3.3v line) to your ram without having to worry about frying the a64 cpu do to high vdimm voltage? if so this could be extremely handy to us a64 users