So did I, but that could have been VBT.
Since VBT and VTT are supposed to be the same, shouldnt there be a way to link the sensors maybe?
Printable View
So did I, but that could have been VBT.
Since VBT and VTT are supposed to be the same, shouldnt there be a way to link the sensors maybe?
hey guys,
so i finished modding my good board... :D
everything went just fine and after setting trimmer to 0 ohm vtt is now 1.52v (up from 1.49v) - vdimm in BIOS was 2.9v. what sucks though, is that i had to take mobo outside my case coz i used screwhole next to vdimm mod area to push wires thru. and now pc wouldn't even start with mobo inside the case... i need to figure out the way to put it back in :)
so, i don't know if i am at the point where ram limits me or what, but with 3.3 rai @ 3.8v, vdimm @ 3.5v and vtt @ 1.75 goldmemory generates no errors within 30mins only at 242FSB...
i need some more testing... i guess...
EDIT:There's very weird side effect of this mod... at 233FSB (3.35vdimm & 1.52vtt) goldmemory shows errors very quickly... but as soon as i make vtt exactly 1/2 vdimm - golmemory has no problem running for 30mins). so it looks like after the mod vtt always has to be 1/2 vdimm.... really weird...
and also, i haven't actually checked that before, but that little Dual-Mosfet that i take vtt measurements at gets REALLY HOT... i need to figure out how to cool it down... so far i have a fan blowing on it & memory slots...
better cool it down b4 it fry .... :D
Hey tictac...you got PM, bro. :)
Credit Goes to:Quote:
Originally posted by daos
who figured this mod out? i amv ery curious who finally figured this board out.
we need to give credit where it is due... :)
enduracell : main idea
me... :D : just give some opinion and double check Malves Vtt trace
bachus_anonym: post pic of dead board with winbond ic remove
Malves: find the vtt trace and vtt mod from the pic
bachus anonym : beta test
Thanx for your time and aford we work as a team here... :toast:
i like that... i hope it clarifies all doubts :toast:
Looks like it does. :)
Hey bachus you got PM at the other forum.
EDIT:
EDIT2: Just checked VTT on my DFI. It looks like VTT tracks even with modded vdimm. I set it up to 3.5v and VTT went to 1.777. Unless I got VTT at the wrong spot, which I don't think so.
damn... i just snap out my vdimm slot no 1 just to re check what uwackme said...Quote:
Originally posted by uwackme
So the feedback resistor is R750. It is located between DIMM1 and DIMM2 near the winbond chip end. It appears to be a 120ohm resistor, but my eyes even with the magnifier arent that good. It measures 120. The pad on R750 nearest the DIMM levers is Vtt, the end nearer the middle of the DIMM is attached to the winbond chip pin2. So this is the feedback path. There is a missing R751 that goes between Pin2 (feedback) and GND, but that would be the opposite of what we want.
Ultimate change would be to replace R750 with a pot, but that would be VERY hard based on location. Our mod so far...adding resistence into the feedback path....cut trace, reconnect through small resistence. Even my EC3001 soldering iron cant get in there without harming the DIMM sockets.
So at least we have a place on the top of the board to measure other than the DIMM. R750 end near levers.
ya....u r right the R750 is the terminal resistance of this vtt sense.... :thumbsup:
u dont need to snap the r750 off... we can solder a wire with a trimmer to R750 leg (near the missing r751.. away from the near lever) and ground it (same method as Motherboardfaqs vdimm mod theory)....
the theory is to increase the resistance
Another way of Vtt mod (Incase u dont want to cut any trace on ur PCB)
any other idea coming?
If we can find vtt pin on dimm solder point( we can solder vdd voltage to that pin ... same as Dirty_Punk vdimm mod did)
any 1 can find where is the vtt pin below the dimm?
Here the 2nd method(Alternative method of Vbt/Vtt mod)
Thanx to Uwackme.... :thumbsup: u make me snap off my dimm slot 1... hehehe
anyway its dead already no big deal... :D
Wait, I dont get anything for saying it would work? ;)
ic dfi have higher vtt voltage.... hehehe :D
edited: i just measure my R750 resistance its about 107 Ohm :toast:
can some1 measure the voltage on the pin i mark here.
If it is the right pin.... we can try solder Vagp/vdd to the vtt pin
same ways as dirty punk vdimm mod did
i hope this helps coz that's other thing i noticed.... at same 3.3v rail and same vdimm - vdimm fluctuates a bit... that wasn't the case before the mod. i guess i need to ground this trimmer now...Quote:
Originally posted by Malves
EDIT: Just to add. It's a good idea to ground the trimmer somewhere. Using the 3rd leg of it. It helps stabilize voltage.
thanx
OK, been testing for hours.
Mod works fine, set 3.3Vdimm Vtt tracked 1.65V. My CH-5 1Gig of it ran 230Mhz 11,3,3,2.5 cpuint enb dual channel. Previous high was 220Mhz. This was on D23-3Dfire. Main system has D10, Ill have to pull it apart now. I think my 350W PS has overvoltage on 3.3 rail that Im just nibbling setting it to 3.6V, the main systems 530W supply has more headroom.
The Vdimm regulator NEEDS some margin to operate correctly. With RAIL at 3.3V and Vdimm at 3.3V yes you will see ALOT of instability because the Vdimm regulator (a little single voltage switching powersupply) is flipping out. You need to keep RAIL .2-.3V above Vdimm. You will see, as you gradually increase Vdimm turning the pot that suddenly the voltage goes from solid rock stable to wildly fluctuating. That is NOT good for stability and Happy-Ram. So set rail to 3.6V by modding your PSU...my Fortrons 350 and 530 PSU's both have rail adjustment pots.
On the Vtt mosfet... it's just like the one's under the "PLATE" on ATI 9700pro cards, and now especially could use some cooling, Im slapping a baby microcool sink on mine. Amazingly, the Vdimm mosfet stays pretty darn cool.
Im looking up the DDR DIMM spec now, need to verify that VREF is set correctly... that it tracks Vdimm. And what is value of VDD and VDDQ in the DDR spec. Spec I guess says the Vdimm mod changes "VDD" and "VDDQ" is fixed at 2.5V...hmmm
So pin 1 of the DIMM's is on the CPU side of the opposite end from the Vdimm regulator. Pin 1 is Vref, Pin 7 is VDD, pin 15 is VDDQ.
On the Valueram, VDDQ and VDD are equal, the in the JEDEC spec for DDR DIMMs a pin on the stick determines if VDDQ tracks VDD or not. I didnt look on the CH-5's but Winbond spec has VDD/VDDQ at 2.5V each default.
Pin 1 is Vref.... it tracks Vtt as it should, spec'd at Vtt +/- .04V. Must wonder if the Vref MAXIMUM spec is a risk with this Vtt mod. Since VDDQ max is 3.6V....Vtt(Vref) max is 1.8V....dont go over it.
Vref is spec'd at DAMN well better be 1/2VDDQ :rolleyes:
So all this time, with Vtt not following Vdimm correctly, we were out of spec with DDR design parameters.
grounding trimmer's third leg, like malves advised, to the screwhole prevents system from starting - i mean it starts but screen stays black...
are you starting at the high or low setting of the trimmer, and what size trimmer are you using? May need 1-50k.
can anyone build a completely new vbt divider circuit .... so that vbt will always stay 1/2 of vdimm voltage like MadCat1968 IC7-Max3 Vbt divider circuit.... :D
On the trimmer value...we are only inserting a small additional resistence into the feedback path for Vbt. So 10-22-33ohm ish amount is all you end up setting the POT to. The R750 we are "enhancing" here is only 120ohm resistor. We are adding a little bit more in series with this 120ohm. It doesnt take much to raise Vtt from 1.45V (2.9V bios setting) to 1.6-1.7V for Vdimm modified to 3.3-3.4V. Just do all this with Bios setting at 2.9V, and you can lower the setting and Vtt paces Vdimm correctly. All we have done really is move UP to a higher range....from 2.6-2.9(1.3-1.45) stock settings to a new range of 2.9-3.3(1.45-1.65) for example.
This additional resistence INCREASES the voltage on the FB pin...trying to tie it to ground through a resistor would DECREASE the FB pin value and LOWER Vtt. So a high value pot tied to Vdimm might actually work instead....bleeding voltage into the FB pin, the analog of increasing R750. This might actually result in Vtt tracking Vdimm rather than you needing to track it manually with BOTH pots. A "set and forget" 2nd pot. Someone should give it a whirl. A 50K pot for starters...set to max and slowly turned down while watching Vtt at a set Vdimm.
Ultimate fix for all this is a completely different Vdimm mod.
We need to find the point where the BIOS setting controls BOTH Vdimm and Vtt in parallel. Stock the board does have Vdimm/Vtt correct and insync and they track. The current Vdimm mod...normal or DirtyPunk...is done at point in the circuit AFTER the Vtt control is made. So Vtt no longer tracks Vdimm.
Since it is 2.6...2.7...2.8...2.9 I'd venture a guess it is a 2 bit register setting in the Northbridge. Now does it leave as 2 digital bits that power a voltage divider network to drive the regulator's control input. Or does the northbridge output an analog vltage used as reference for the regulators. IE: 0,0 = 2.6V; 0,1 = 2.7V; 1,0 = 2.8V; 1,1 = 2.9V. And we modify it to be: 0,0 = lowest new V....... 1,1 = highest new V. All the while Vtt tracking correctly.
If we can find the point where both are together...output of northbridge to the regulators...and alter THIS with a mod, then both will simply track, but start at a mutual (and correct) higher default setting.
It's really now just a matter of finding that point. Requires completely reverse engineeing the Vdimm/Vtt regulator design. Not too hard, I'd be a lot easier if I had a DEAD board to use as a road map....and pull parts off as needed. The basic prototype is sitting in the Winbond/etc data sheets for the chips involved.
In the meantime, guess today is RiptheSystemApart...AGAIN day :slobber:
:eek:
Did the mod today but didnt solder wires yet:cool:
http://muljas2.mine.nu/dani/Tietokon...0VBT%20mod.JPG
Cheers to a man that certainly knows his stuff.Quote:
Originally posted by uwackme
On the trimmer value...we are only inserting a small additional resistence into the feedback path for Vbt. So 10-22-33ohm ish amount is all you end up setting the POT to. The R750 we are "enhancing" here is only 120ohm resistor. We are adding a little bit more in series with this 120ohm. It doesnt take much to raise Vtt from 1.45V (2.9V bios setting) to 1.6-1.7V for Vdimm modified to 3.3-3.4V. Just do all this with Bios setting at 2.9V, and you can lower the setting and Vtt paces Vdimm correctly. All we have done really is move UP to a higher range....from 2.6-2.9(1.3-1.45) stock settings to a new range of 2.9-3.3(1.45-1.65) for example.
This additional resistence INCREASES the voltage on the FB pin...trying to tie it to ground through a resistor would DECREASE the FB pin value and LOWER Vtt. So a high value pot tied to Vdimm might actually work instead....bleeding voltage into the FB pin, the analog of increasing R750. This might actually result in Vtt tracking Vdimm rather than you needing to track it manually with BOTH pots. A "set and forget" 2nd pot. Someone should give it a whirl. A 50K pot for starters...set to max and slowly turned down while watching Vtt at a set Vdimm.
Ultimate fix for all this is a completely different Vdimm mod.
We need to find the point where the BIOS setting controls BOTH Vdimm and Vtt in parallel. Stock the board does have Vdimm/Vtt correct and insync and they track. The current Vdimm mod...normal or DirtyPunk...is done at point in the circuit AFTER the Vtt control is made. So Vtt no longer tracks Vdimm.
Since it is 2.6...2.7...2.8...2.9 I'd venture a guess it is a 2 bit register setting in the Northbridge. Now does it leave as 2 digital bits that power a voltage divider network to drive the regulator's control input. Or does the northbridge output an analog vltage used as reference for the regulators. IE: 0,0 = 2.6V; 0,1 = 2.7V; 1,0 = 2.8V; 1,1 = 2.9V. And we modify it to be: 0,0 = lowest new V....... 1,1 = highest new V. All the while Vtt tracking correctly.
If we can find the point where both are together...output of northbridge to the regulators...and alter THIS with a mod, then both will simply track, but start at a mutual (and correct) higher default setting.
It's really now just a matter of finding that point. Requires completely reverse engineeing the Vdimm/Vtt regulator design. Not too hard, I'd be a lot easier if I had a DEAD board to use as a road map....and pull parts off as needed. The basic prototype is sitting in the Winbond/etc data sheets for the chips involved.
In the meantime, guess today is RiptheSystemApart...AGAIN day :slobber:
Thanks for all your help man.
i'm gonna look this thread over again tomorrow, when i'm not tired, and see what exactly we're doing with this mod.
way to go Stang_Man..... :D GL :thumbsup:
Bachus_anonym
I just check my dead board and compare it ... there are missing mosfet in ur board...
i'm confused here... do u remove it .... what mod is that for?
here my dead board(NF7-S Rev2.0) mosfet near the sata connector :confused:
Vbt=Vdd mod.... delete
result:
2 dead board (Mistamasta from AMDMB.com & asus from lowyat.net)
1 Safe (Mickey Padge Xtremesystem.org)