***Volt Mod's For AsRock SataII***

[ochungry]

voltmod update

Ok, I went to radioshack and bought a 2x3 breadboard ($1.95) and soldered 2 sets of resistors. One set for the vdd and the other for vcore.
I have already done VID3+VID5 mods, but this time wanted to see @ what resistor value I can get rid of vcore fluctuation.
I had a feeling, the resistor value need to fall somewhere in the middle of the FB reading, and not the edges. What I mean is that the vcore controller has an accuracy of .5% that can fluctuate 16-32 mv.
Instead of being close to the edges (near 16mv or 32mv) I wanted to find out how much ohm needed to put it right in the middle (16....middle....32).
I soldered on the breadboard a 25K w/ 10-100 ohms in series. this gave me 25.1, 25.2, 25.3.... so on. Soldered some pins. Stripped an IDE cable and used individual lines and soldered header to one end(to push in the pins on bread board). Attached one end to 25k leg and soldered the other end to the cap (as shown by Rebel's mod) soldered the other wire to ground and a header on the other end, set my vcore to 1.275v(1.50v actual) and checked vcore from 100 ohm end(26K total in series) I found 25400ohm(25.4k) less fluctuating than below or above it. I guess I am very close to that middle point was looking for. I am thinking to divide the resistors to 10ohm interval and check vcore w/ smaller fractions.

Now about the vdd: I did the same thing w/ vdd, but only used 1k ohms up to 9K(in series) and used 10-100 ohms to end up w/ a 10K VR.
The vdd without the mod was @ 1.99v, and gradually raised it to 2.07v using 6.7k resistor. don’t want to go higher than that for now. I also found the middle ground for 2.07v vdd. from 6.4-6.9K vdd stayed the same. so I chose 6.7K.
Both mods are on the same breadboard. I guess you can call this breadboard a VR board, that is more accurate than a trimmer, no chance of overvolting by accident, and can know the resistance values.

Edit: I went to bios and lowered the vcore one notch to give me 1.58v. came to window and CPU show a bet more fluctuation. I lower my BD VR(breadboard VR) 100ohm and vcore stabilized.
Let me see how I can explain this. I guess the vcore regulator .5% accuracy does not exactly fall right on 25mv vcore increase. it has a 32mv range and 25mv increase will have some values left over(32-25=7mv) for the vcore 25mv increments. so I have to adjust the BD VR to compensate for that 7mv difference that may fall below or above the bios vcore adjustment.
I think this explains it why I had to decrease another 100ohm to make vcore stable as previous setting was. I really think If I solder a few 10 ohms and use smaller fractions the vcore might be more stable.
What if I solder a 1uF cap at each pin...hmmm?
I might do that. hope don’t fry the regulator. If you don’t see me post here for a good while, chance are I don’t have a mobo to access internet
Oh another note: there is still a bet of fluctuation. But very little and only jumps up 16mv and quickly adjusts. Much better than before that was 32mv and constantly up and down.

Edit2: I un-soldered a 10uF capacitor(small cylinder shape) from an old mobo, and soldered it between the 100 ohm resistors (grounded the ground end) and now the vcore is very much stable. It just fluctuates 16mv once in a while up and right back. Because the cap is 10uF, It takes about 10 minutes for vcore to stabilize. I guess it needs to fill up.

 

[jonspd]

< is lost care to explain only the vcore mod/vdroop? according to persivore 1,068 is what he used to get a stable vcore. Also on a side note depending on what vcore in bios, vcore, and cpu speed you are @ will depend on what ohm resistance work's best.

Or a pic would be great.

According to MBM5 which is wrong I get a .2 vcore flux.

jonspd

 

[ochungry]

< is lost care to explain only the vcore mod/vdroop? according to persivore 1,068 is what he used to get a stable vcore. Also on a side note depending on what vcore in bios, vcore, and cpu speed you are @ will depend on what ohm resistance work's best.

Or a pic would be great.

According to MBM5 which is wrong I get a .2 vcore flux.

jonspd

jonspd,
I am still working on it. I will drw some schematic when done. I don’t have a digital camera to show actual work.
basically, I am using Rebels mod+OCW vcore mod+ my mod(VID5) and in one piece .
I changed the total resistor(rebel's vcore mod) to a 15K+ 5-100 ohms+3-50ohms. I replaced the 10 uF cap / a 1uF(small cylinder kind) and soldered one leg to 15K-100 ohm junction. the voltage at the junction reads 83.5mv. So, the cap is receiving that much voltage. the cap's other leg is connected to ground. Right now, I have set vcore to 1.275 in bios, and CPUz reads 1.616. the FB ground is connected to 3rd 50 ohm resistor(from ground side).
The total resistance is 15K+5*100 ohm + 50 ohm=15.550k.
VID3 and VID5 are also connected to SGRND. all and all I can get up to ~1.75v. But just using 1.616 now.
Also I connected a 1uF cap to the VDD Resistors. Its 3K from the source.
I know it all sound confusing. But it is not really. Very simple once you see the schematic drawing.
About persivore's mod for vdroop, I think you can just solder a resistor right on top instead of removing the existing. as long as you know the value of existing, you can use ohm law to get near what should be.
Mine is a "06B" written on top of it which is 1130 ohm. One could solder 2 wires to each end of the existing resistor and use something like I made or just a 50K VR/trimmer and get it close to 1068ohm as persivore's.

 

[ochungry]

Ok, I drew an schematic of the vcore breadboard VR.
I changed a few things. I went back to 10uF capacitor, because it workd better.
The VDD can also have similar layout, but less R. The VDD only requires 5-200 ohms in series, and 5-1K . a 2.2K will give me 2.08 v, .09v increase from 1.99.
every 200 ohms decrease is ~ .01v

Screenshot

 

[jonspd]

Cool kinda like a switch board you set by how many resistor's you using the amount of ohm resistance you get.

2.07-2.08 work's best for me. Anything more then 1.52 on cpu it doesnt like also.

What is your max 1:1 and 5:6 setting.

Jonspd

 

[ochungry]

Cool kinda like a switch board you set by how many resistor's you using the amount of ohm resistance you get.

2.07-2.08 work's best for me. Anything more then 1.52 on cpu it doesnt like also.

What is your max 1:1 and 5:6 setting.

Jonspd

Opteron 146 was running 2.85 prime stable. could do Spi 32M @ 2.98ghz but could not pass prime.
venice 3000 runs @ 307 1:1 w/ patriot xblk.
cant do better even if using dividers. This ULi 1695 highest is 300mhz, IMO. can go higher than 300mhz, but something will not be stable. its either SB, NB, or just the way the board is made.

 

[jonspd]

Can you post ever setting you have for 285 prime stable?


and how you can get 32md stable @ 2.98/

If you like I can post up all I have done and see what the difference is maybe I have room for improvements.



Jonspd

 

[ochungry]

Can you post ever setting you have for 285 prime stable?


and how you can get 32md stable @ 2.98/

If you like I can post up all I have done and see what the difference is maybe I have room for improvements.

JOnspd

Jonspd

Just those settings I posted a while back in new asrock thread(post #490).
only dif is that I set SB and SB burst length to auto. NB to 800 MHz and 16 bit burst length.
I am also using 2.95 vdimm and 1.616 vcore.
I bet you need to increase your vcore.
using 04xx bios. enable ram idle, and manually set the preamble and async(in bios under mem settings).
disabled sata , enabled sataII and in sata mode.
defrag HDD before prime. put page files in a another partition(10gig only).
I posted my results w/ screenshots about 1 ½ month ago (in that new asrock.... thread).
BTW, I have RMA'ed my opty 146 and using Venice 3000 right now.
The reason for RMA was because I noticed a black rubber like strip along one side of IHS, and wasn’t too happy about it. I think the IHS was not mounted w/ precision and some sealing material sipped out. was running unusually hot too. I don’t think I had a good batch.

 

[ochungry]

I did the vldt mod, thanks to persevere's ingenuity. I am using 1.21 for now. Dont know what ldt voltage is good for this board. have to do some tests.
I am using my breadboard VR same as the one did for vdd, Vdimm and vcore.
Here are some #.
For vldt:
37K resistor increase to 1.20v
22K to 1.21v
17K to 1.22v
15K to 1.23v
12K to 1.24v
10K to 1.25v
For Vdimm, if you solder a resistor right on top of 1.2k resistor(OCW vdimm picture)
The resistor to use:
1k gives 2.80v
750 ohm gives 2.85v
500 ohms gives 2.90v
300 ohm gives 2.95v
220ohm gives 3.00v
180 ohm gives 3.05v
The 180 ohm almost gives you same value as if you remove the existing resistor and soldering a 150ohm resistor(as was modded by "ED", OCW Forum). Because according to ohm law: R1*R2/R1+R2= (1.2*.18)/(1.2+.18)=.216/1.38=156 ohm
If cant find 180 ohm resistor, a 150 to 200 will do it, w/ just little voltage variation.
Just make sure the resistor is marked "102" on top of it(1K). if not, then you have use the ohm law(above formula) to get the voltage you are interested.
You can also Use a Trimmer as per TechPowerUp, or BreadBoard similar to wha I built, w/ pins and resistors in series.

 

[jonspd]

I think VLDT help's htt speed and possible cpu overclock.


Great info you got there OCH.

 

[ochungry]

Thank you jonspd.
With vldt mod and increasing it to 1.22 I am now able to overclock Venice 3000 to 2790mhz(310X9) , where could only do 2700(300x9) passing Spi 32M.
I am using 166mhz divider though. Need to get some good TCCD's that do >300mhz, and I am all set. For some reason cpuz shows the speed @ 2789 (309.9x9). but I am happy. I am going to increase the FSB and see what I come up with.
Here are my settings:
FSB=310
Ram=166/333 mhz
3-4-4-8, 1T
Idle cycle=enabled
read preamble=5.5 (depends on ram's. use your ram's spec)
Max async latency=7ns (depends on ram's spec.)
NB=600, 16 bit
SB=auto, burst length=auto
vcore=1.616v
Vdimm =2.91v
vdd=2.08v
vldt=1.22v
AGP aparature=32mb
SataII=Sata mode
04xx.bin bios
Using black dimms
Higest cpu temp=39c
fan blowing on rams
modded NB and SB heat sink w/ fans
fan blowing on vid card.
That's about it. Dont know if any other info matter here.
280000000000000000000000000000000000000000008
EDIT: Oh boy, for the first time I am able to go into 2800's w/ this venice 3000. Now I can clock up to 312x9=2808mhz.
Thank you Persivore and thank you jonspd for starting this thread.
The same settings as above, but 312mhz FSB. Screenshot of Spi 32M and CPUz validation.

 

[jonspd]

Nicely done.

So I take it that the Vldt helped alittle

Guess I need to go get a soldering iron.

 

[ochungry]

Nicely done.

So I take it that the Vldt helped alittle

Guess I need to go get a soldering iron.

It sure did, and a lot. Before, I couldn’t post beyond 320 w/ lower multiplier.
Guess what, I now am able to use 8x multi and go as high as 350mhz FSB/HTT, to clock 8x350=2.8ghz on this Venice 3000.
Had to use 166mhz divider for the memory and increase the timing to 3-5-5-10 to get 280mhz mem speed. Here where I need good set of rams. Where are you DDR700 .
With a good set of TCCD modules and 2.5-3-3-7 can produce amazing Spi result. But as is, I have HTT overclocked 75% and Venice 3000 is more than 55% overclocked. This is not High HTT @ very low multiplier (5x as was done by an OCW forum member) just for the sake of high HTT. Here is the Screenshot.
I also had to change vldt to 1.24v and vdd to 2.10v.
vdimm is still @ 2.91 volt, vcore @ 1.616v
I am going to prime it now.

 

[jonspd]

Holy tit batman.

Well I guess I need to get on the move.


Keep up the good work and hopefully you will be able to get some DDR 700

Have you tried 133 divider with lower timing's? (my 133 divider acts weird)

 

[ochungry]

Holy tit batman.

Well I guess I need to get on the move.


Keep up the good work and hopefully you will be able to get some DDR 700

Have you tried 133 divider with lower timing's? (my 133 divider acts weird)

Thank you. Yes you need to hurry up and do the vldt mod. made a world of difference. This thread is getting hot now
I haven’t tried 133 divider yet. The reason I didn’t was because it would have put the ram to ~ 231mhz, and thought maybe lower Spi score. But I will do it now just to see. If you use lower multi for opteron and good rams, I bet you can break your record.

EDIT: Ok, Here it is @ 133mhz divider and 3-4-4-8 timings
Screenshot. Going w/ A64 tweaker to use 150 mhz divider. That is if I dont Bsod it, since mem controller uses its own even #.
This is Sisandra using 166 divider(3-5-5-10), which isn't bad at all.

 

[jonspd]

Also see what you can get 1:1

I have to go buy a soldering iron 1st I do all my mod's via smd grabber is possible. Going to ask around and see if VLDT can be dome on one of the pin's under where you do the mod @.

Off to work be back 7 am sharp.

 

[ochungry]

Jonspd, 1:1 I can only do 307mhz w/ this hynix modules and very loose timing (3-6-6-11). I could do 295mhz 3-5-5-10. This 2x256 kit were bought as a temp. replacement for the Patriot XBLK (2x512) RMA'ed a while back. Patriot sent me new set and one of the stick did not do any good so I sent it back am still waiting for a good set of TCCD.
But this hynix kit is performing well so far, even though timings are very loose.
Now about Overclocking w/ lower multiplier: I think I have reached the 352 mhz FSB limit. can do 354mhz but not stable and wont pass Spi.
the difference between 166mhz divider at 3-5-5-10 and 150mhz at 3-4-4-8 is minimal, as posted the screenshot (post #55). This is the Screenshot of 352mhz FSB using 166mhz divider @ 3-5-5-10 timings. next I am testing w/ A64 tweaker choosing 150mhz divider and 3-4-4-8, and see if any better.

EDIT: prime95 was running overnight and I just turned it off into 10 1/2 hrs.
So it appears 8x352 is stable. Screenshot

EDIT2: One thing I noticed looking @ cpuz, was that cpu has used a 10x divider. so I chose 166 divider in bios and 345mhz fsb to get 2760/10=276mhz
where is near this kit's speed(275). 166(.83) calculation is not exactly how cpu divides it. .83x345=286 is higher and I didnt use it before. But now after seeing the cpu divides it differently, I used this divider and now cpuz shows the mem speed @ 276mhz. This is better and got better score

 

[jonspd]

Going to radio shack tommorrow to get a soldering iron and some vr's.


Guess I might play around with some soldering
It's been a while. Going to practice on a dead geforce 2.

Jonspd

 

[ochungry]

Going to radio shack tommorrow to get a soldering iron and some vr's.


Guess I might play around with some soldering
It's been a while. Going to practice on a dead geforce 2.

Jonspd

Good. make sure you get the narrowest shaft and point. also get a strip of wick to soak up excess. and works good when you remove smds. it will soak up the solder from where it is soldered to. very thin solder roll, and a can of tin. you can buy the plumber's tin from home depot is cheaper.
always apply tin to both ends you solder. just expose the wire about 2-3 mm from its insulator, tin it, apply solder to it, tin it again. tin where you solder to. tin the tip of soldering iron(when it is hot and well cleaned), apply little solder to the soldering iron tip. and touch and move off quick. you don’t need to hold the iron too long to where soldering. quick action will do it. the 2nd try is harder if didn’t do it right first time, and becomes messy. so make sure you do it right the first try. practice on some small smd's w/ above tips, you'll see its piece of cake.
The wick may not be needed if you don’t over apply solder. just eyeball how much solder you need to apply to the tip of iron. if the iron tip is clean and you tin it, solder will stick to it. otherwise you will have hard time keeping it @ the tip, that little bit of solder you need. not too much thu. always remember this- not too much solder needed, and always first try(application) should do it. and clean iron tip. always tin before applying. Tin is your best freind.
when done use alcohol and Qstick to clean the tin resin residue .

Edit: Forgot to mention that if you’re planning to solder a wire to mosfet where you check voltage, you should apply a dab of solder to the mosfet's leg before the wire, because the legs are thick and may take longer to melt the solder into wire. applying solder to the leg before attaching the wire, soldering the wire will be much easier. Scraping the surface of mosfet legs w/ an exacto knife will also help. no need to scrape off other locations, they are not as thick as the mosfet legs.

 

[jonspd]

Ok I got my xp-90c in the mail. Then I went shopping for a 25 watt soldering iron, some .032" silver solder, extra tip, pack of assorted pot's and trimmer's, some tinned flux from Home Depot, and a variety of sand paper for lapping and sharping tip.

Only thing I missed was the wick to remove solder and my mx518 to show up.


I dont think I'am going to hook a wire to anything for voltage read-out's.


It may be next week before I get everything done will post result's as they show up.



Jonspd