DO NOT DO THIS MOD AND THE OCW MOD AT THE SAME TIME! There have been reports of some kind of problem when doing both. NEVER DO 2 Vcore or 2 Vdimm MODS AT THE SAME TIME, the Vcore could be 1.8V or higher! If you do the mods, your doing them at your own risk. The Rebels Haven Computer Forum will not be responsible for damaged systems or components... If your unsure about the mods, I strongly suggest you do not implement them...
Let's start with identifing the PWM chip. Someone with a board tell me what all the numbers are on the chip I have circled in this pic:
-------------------- Too Many Computers,... Too Little Time ..... Damn Phenoms! Posts: 25194 | From: Fire Island, NY | Registered: Feb 2003
This little booger is the PWM controller (7x7mm!!!)
It is a STMicroelectronics (www.st.com) L6711 PWM controller with internal VID DAC. The .PDF spec sheet is availible HERE.
Since the A64's use the "Hammer" DAC with internal pullup, i tried to short pins 24-23 (VID 0/5) and pins 25-26 (VID 1/2).
(yes i did get rid of the excess goop, i did not short 25-26 to that solder pad below the PWM)
Since the pins are a logic "1" when floating, and a "0" when grounded, i figured that shorting one pin to a pin that is a "0" would yeild both of them being "0". I did this in hopes that I would turn the 1.45v setting in BIOS (VID[5:0] 100100) to a 1.575v setting (VID[5:0] 00000). But this did not seem to work right. 1.4v setting in BIOS makes a 1.45v, and i cannot change to anything else without locking the PC up.
I will try removing the 23-24 short (+25mv boost) to see what effect that has.
Seems that either the PWM doesnt allow these voltages (possibly ASRock set the overvoltage to under 1.575v) or the BIOS sees an incorrect voltage and kicks it out, or the Float/ground doesnt work as expected. I might have to go directly to ground.
-Plack
-------------------- Lap: Dell Vostro 1500/1.6GHz C2D/4GB RAM/Nvidia 8400GS 128MB/15.4" Desk: E8400 9x433(3.9GHz)@1.35v/Gigabyte G33M-DSR2/4GB RAM/ATi 3850 256MB/37" Westinghouse 1920x1080 My Photo works on Deviant Art Posts: 1313 | From: Hephzibah, GA | Registered: Jan 2004
posted
Trying to fiddle and fool the DAC is looking for trouble. Those pins you shorting may have active pull ups and the chip may get damaged...
Pin 8 is the Feedback pin... From the other post:
"Every Vcore is a Pulse Width Modulated Switching Regulator. It has feedback resistors going to the error amplifer. 1st thing that has to be done, is identify the PWM Switching Reg chip and get a data sheet...
The pin that gets a resistor to ground, will be labeled "FB" or "Feedback".... I try and either measure the resistance to ground to see if it agrees with the "typical" circuit on the datasheet, or visually read the numbers on the resistors.. Most boards follow the "typical" circuit to the letter. That tells us whether to use 10K, 20K, or 50K...... I see most modders just try 50K and in reality, that probably will never cause a problem.. I just like to know what I'm doing...."
Posts: 25194 | From: Fire Island, NY | Registered: Feb 2003
posted
As always... polygon is right And i might have already toasted the DAC, but at least the 1.4v setting still works (measures like 1.43v on a DMM).
I put a 100K pot between ground and the top of the capacitor above the FB/Comp pins. There is only a resistor (1.2k) between VSEN and FB, but no resistor between FB and Comp (just the cap). I started with a 100K, just to make sure i had enough resistance to play with. Only need a 20K or so to make it work right in reality. I havent measured my exact resistance now, but it is well below 20K.
Voltage without the mod was 1.43v, right now i am sitting on 1.625! Everything adjusts fine from stock to past 1.65v. Got scared at 1.65v (as one should on a 90nm), and so never tried above that.
9x295 is now possible, and stable on my board 300MHz HTT is still unstable, regardless of core multiplier, unless i crank the memory way down, thereby making it useless.
I have made a few runs through SuperPi and Memtest86 running 9x295 at 200MHz divider (1:1) at 3-4-4-8, but nothing 3D is stable. So at least DDR590 is possible on this board, but i think i still need more volts for that.
I will post more info later, for now I am gonna play a few games to make sure this thing is 100%...
-Plack
-------------------- Lap: Dell Vostro 1500/1.6GHz C2D/4GB RAM/Nvidia 8400GS 128MB/15.4" Desk: E8400 9x433(3.9GHz)@1.35v/Gigabyte G33M-DSR2/4GB RAM/ATi 3850 256MB/37" Westinghouse 1920x1080 My Photo works on Deviant Art Posts: 1313 | From: Hephzibah, GA | Registered: Jan 2004
posted
Yep, its my patriot XBLK. As far as VDimm, it will involve ripping out the chip that already exists, its just too weird. Seems like it works backwards from what it should.
System was perfectly stable at 9x290 on 1.625v. Did crash at 9x295 (bout an hour into the game). I should probably just turn up Vcore a tad more, and 295 would be fine.
Can TCCD take 3.3-3.4v nonstop? Easiest mod is just to hook 3.3v rail to the memory... But i am kinda scared to run that high of voltage to it nonstop. Its not like the old BH-5 i dont think...
-Plack
-------------------- Lap: Dell Vostro 1500/1.6GHz C2D/4GB RAM/Nvidia 8400GS 128MB/15.4" Desk: E8400 9x433(3.9GHz)@1.35v/Gigabyte G33M-DSR2/4GB RAM/ATi 3850 256MB/37" Westinghouse 1920x1080 My Photo works on Deviant Art Posts: 1313 | From: Hephzibah, GA | Registered: Jan 2004
3.3V is OK for TCCD but I wouldn't go any higher... Just make sure there is a FAN blowing on the sticks....
Testing indicates that over 2.9-3Vdimm, there is little increase in OC ability with TCCD...
When my board arrives, the 1st thing will be to look at the Vdimm mod.... There has to be a way to adjust it up 2-3 tenths of a volt... I like 3V for Tccd. I know I'm getting all it has...
-------------------- Too Many Computers,... Too Little Time ..... Damn Phenoms! Posts: 25194 | From: Fire Island, NY | Registered: Feb 2003
In order to raise the VCore a proportional amount, we will mess with the feedback circuit of the PWM controller. If we lower the resistance of the FB (feedback) pin of the PWM, then we can increase the voltage out. We do this by using a resistor or potentiometer from the FB pin to ground.
I have measured the resistor RFB to be 1.13K on my board (measure it yourself if you can on your own board). Since we now know the resistance, we can calculate the voltage out.
I have calculated out the required resistance to meet certain values, show below: +5% = 21.6K +10% = 10.2k +15% = 6.4K +20% = 4.5K
So for example 1.45v * 105% = 1.522v 1.45v * 110% = 1.595v 1.45v * 115% = 1.668v 1.45v * 120% = 1.740v
Rather than have just a single potentiometer in the circuit, you will want to run a resistor in series with the potentiometer in order to limit the max VCore voltage. This is in case you put the pot in with the wrong setting, if you accidently turn it too far, or if the pot shorts out. In this case the voltage would only be a maximum of whatever the inline resistor allows. For most, you should put in the 15% or 20% limit.
The nice thing about this mod, is that you can turn down your VCore inside BIOS, and it will still change your VCore. This would allow you to set the Max Vcore, and then just turn down the Voltage in BIOS from there.
For instance, using the +20% resistor, you could set your voltage in BIOS to 1.175v and you would get 1.41v out - approximately original voltage. So if you wanted to use a solid resistor, this would work fine.
For those that want infinitely variable voltage, put the limiting resistor in series with a 10K or 20K pot. Just add your resistance together for the minimum voltage, and the max is what your resistor is.
Now for the mod: The place to solder the resistor wire to is here:
It will look like this when you are done:
Soldering tips: Use a small solid core wire for the mod. Pre-tin the wire, and then just apply heat for as short amount of time required to solder the wire to the cap. If you dont know how to pre-tin wire, then dont even attempt this mod. These components are extremely small, and require a decent soldering skill. If you have never soldered before, you need to try something else first. Thats just the way it is, unless you like to live very dangerously.
For long-term systems, dont run over 1.6v on a 90nm CPU, or 1.7v on a 130nm CPU. You could run more or less, but that is a personal choice. Longetivity will probably be affected greatly by going over these voltages. Dont forget to monitor temperatures, as they will go up extremely fast with these new voltages.
I will post a pic of my final VCore pot/resistor when i get the components on monday. I am just running it off a breadboard right now.
posted
Oh, and i tested 3.3v on the RAM. Didnt seem to do much other than make the RAM F'ing hot. Although I did manage to pass Memtest86 at 290MHz 2.5-3-3-8
Windows still crashed in 3D mode, even at 3-4-4-8.
VTT did track properly, exactly 50% of VDimm. Not sure what the onboard FET was up to... it still carried the same gate voltage as before
I tried the VCore mod, lets see what you can do with the VDimm. 2.8-3.0 would be nice...
What would really be interesting, would be to see some good ol BH-5 running 3.3v 2-2-2-5. Seems like the board just peters out on memory above 250-260... perfect range for BH-5.
-Plack
-------------------- Lap: Dell Vostro 1500/1.6GHz C2D/4GB RAM/Nvidia 8400GS 128MB/15.4" Desk: E8400 9x433(3.9GHz)@1.35v/Gigabyte G33M-DSR2/4GB RAM/ATi 3850 256MB/37" Westinghouse 1920x1080 My Photo works on Deviant Art Posts: 1313 | From: Hephzibah, GA | Registered: Jan 2004
posted
Oh, and to measure VTT, it is Pin1 on the Dimm slots. It is the top left pin on each slot. I measured the pin at the top of the mobo, on the slot closest the CPU.
-Plack
-------------------- Lap: Dell Vostro 1500/1.6GHz C2D/4GB RAM/Nvidia 8400GS 128MB/15.4" Desk: E8400 9x433(3.9GHz)@1.35v/Gigabyte G33M-DSR2/4GB RAM/ATi 3850 256MB/37" Westinghouse 1920x1080 My Photo works on Deviant Art Posts: 1313 | From: Hephzibah, GA | Registered: Jan 2004
posted
That is correct Speedy-J! I just chose to solder mine to the top of that Cap, since it was the easiest place to solder, and least chance of shorting something else out. If you are good enough, you could go directly to pin8... but i am sure not!
And dont forget that Polygon suggested I use the FB pin. I had completely overlooked that when i was looking at the spec sheets. For some reason i thought that the DAC would override eveything else
And yes, old-school mods are fun Its nice to have a board that doesnt require this... but where is the fun in that!?
posted
And for the heck of it... A view of this board in its current state (yes i am posting from this PC).
In the pic you can see along the top and on the left random tools and solvents used for this mod.
Top right is the breadboard that currently holds my potentiometer used for the VCore mod.
My optical drives are still in my massive case, but since i have 36" cables, they easily reach the PC (and yes my case is as tall as that desk!).
Thing that suprises me is how fast that Thermaltake Venus-12 reacts to changes in CPU load. Sitting at idle or minor useage, it will run around 2200RPM. But within 10-15 seconds of full load, the thing will spool up past 3500RPM. Being a gearhead myself, this is quite fun to hear, as it reminds me a turbo building boost. The same holds true when i unload the CPU... fan comes down rather quickly.
Below the board you also see a printed shot of the PWM controller i used to trace on as i was doing the mod. Also Memtest86, and yes, a legal copy of BF2. Speaking of BF2... it is about to see some use...
Topic Closed
Damn Phenoms!
![[Big Grin]](biggrin.gif)
And i might have already toasted the DAC, but at least the 1.4v setting still works (measures like 1.43v on a DMM).![[Smile]](smile.gif)
![[dono]](graemlins/fragend013.gif)
![[ThumbsUp]](graemlins/thumbsup.gif)
![[Roll Eyes]](rolleyes.gif)
![[banghead]](graemlins/banghead.gif)
