How to disable OVP on Asus SK8N MB?

[mslezak]

Got my Opteron 144 to 2.3ghz, but the OVP keeps turning my system off- usually in 5 seconds or less! 2.275 shuts off after a few minutes, only 2.25 won't do it (1.92V).

I looked at the datasheet for the vcore controller, the Analog Devices ADP3166 here at http://www.analog.com/UploadedFiles/...8ADP3166_0.pdf

It has a voltage limit between 1.7 to 1.9V it appears (under current limiter comparator). It also clamps the device to ground (called a crowbar) from 2 to 2.2V, seen on page 2 of the datasheet. I have seen some knowledgable OCers on this board, so I am asking for an opinion here to disable OVP.

The crowbar on page 5 is shown to be pin 6. This output "can be linked to an outside device to shutdown the system" - it outputs a logic level (high or low). It would make sense then that to keep it from activating, you simply disconnect it and connect to whichever signal (high or low) that keeps it from activating. So that would not be too hard I imagine.

Now the current / voltage sensing pins are where I need help. CSREF on pin 16 states that it senses the current and voltage for the crobar and current sense. Would it make sense to cut the trace at this pin, then put a resistor in between to lower the voltage the regulator sees?

There is another pin 15 ILIMIT. This is set with an external resistor to ground to set the current limit threshold. Then you could change the resistor value to set a higher current limit. Would this allow a higher OVP value?

Just looking for some pro advice before I start cutting away. Thanks all in advance!

[Tom Holck]

Welcome to XS mslezak

I don't think it work.

Many have tried to raise the corevolt on that MOB.

My advice: By a SK8V its faster, and then you can go to 2.080 volt on the core

Don't pass 2.112 volt, if you to the bord will die. (tried 4 times )

[mslezak]

Thanks for the reply. This mb has had a reading of 2.144V as the highest, and it was still fine after this. It did activate the OVP however. So you say the SK8V will allow 2.080V without activating OVP? Do any other boards you are aware of let you go any higher? Thanks.

[Tom Holck]

Originally posted by mslezak
Thanks for the reply. This mb has had a reading of 2.144V as the highest, and it was still fine after this. It did activate the OVP however. So you say the SK8V will allow 2.080V without activating OVP? Do any other boards you are aware of let you go any higher? Thanks.

Yes 2.080 is the OVP limit on SK8V. If you set 2.112 V right before bench, the volt drop a little, and the bord let you do that. But you must have extreme cascade-cooling, not even dry-ice is sufficient past 2 volt. My first -112C cascade can "only" handle 2 volt for superPi M32 in 20 minutes (64bit OS), then the temperature goes up.

I think SK8V still is the best FX bord. The new 939 bord have weak power-supplies, and AGP-lock problems. And difficulties with FSB higher than 250/260. (had 2 out of 5 SK8V that can do FSB 250+ fastest timings)

Have you seen OVP-mod for SK8V? The corevolt holding my old FX back. My cascade can handle 2+volt, and still do -120 C


I have had my new FX-53 939 on 3.5 Ghz, but it's not faster, because of the poor mem performance...

[Hell-Fire]

There may be an option, if you are willing to try it.

The CROWBAR does not clamp 12v to ground to shut off and protect the CPU until CSREF hits 2.1v.

CSREF is also tied directly to the lines fedding Vcore via the ADP3166, and a few other regulators that are part of the circuit.

Check the resistance of Pin #16 (CSREF) to GND. Lowering that resistance may allow you some more headroom with Vcore.

[Tom Holck]

I will try that have a few SK8V left to try on

[Tom Holck]

Hell-Fire:

Is that pin 16?

If it works my max oc will improve.

for now its 3360 bench all
and 3486 schreenshot

And still waiting for the x800xt

[mslezak]

Pin16 is on the other side of the top of the ADP3166, you are pointing at pin27.

Find the resistor that is between pin 16 and the output inductors and draw a few lines of pencil lead over it (between both ends of the SMD resistor - pencil mod), then try to clock higher. Should lower the resistance ever so slowly, changing the reference voltages used for the crowbar and powergood functions, and current sense amplifiers. You will effectively be raising the voltage reference value.

On my SK8N, this appears to be marked R42, which connects to pin22 of the ADP3166. The other resistor it could be is R38, which comes from the trace coming from the coil in the upper left hand of your picture - take your pick, they appear connected to each other. On your board, you can use a continuity tester to find it with one probe on pin16 and then poke the resistors in the area until you get a beep. Or use the diode checker - mine gives 0.00V when you touch both probes together, then put one probe on pin16 and poke the resistor ends in the area until you find another with this reading. That's how I found R42 and R38.

I edited your picture above with Pin16 marked and also the location of R42 on the SK8N (looks really similar to SK8V), but I can't upload it. Let me know how and I'll post it for you.

[Hell-Fire]

Originally posted by Tom Holck
Hell-Fire:

Is that pin 16?

If it works my max oc will improve.

for now its 3360 bench all
and 3486 schreenshot

And still waiting for the x800xt

No sir....I marked #16 for you in the pic below.

Agree with what was stated above. Do a contuinity test to find out what resistors/capacitors are connected to pin #16. If you can actully see and follow a trace coming off #16 then that would be easier obviously.

Yes, do a resistance between whatever we can find connected to #16 and GND. You will have a hard time getting a good reading if you try to do it with #16 unless you have uber tiny probes on your multimeter.

[Tom Holck]

Thanks Hell-Fire and mslezak,

But:

I have made a delikate soldering to pin 16 with a thin wire.

Booted op at -120 C safe volt (1.75) everything OK.

then I tried to messure the resistance of pin 16 on the wire?

different messure in different selektion?

0-2k it was 16 ohm
200k it was 150 kohm ??

and the computer instantly froze when I messured. I booted up 3 times ..
Now the bord is dead? Last time I was in bios in monitor status, it seens at it was overvolting when I messured on the wire?



I have a new board ready, but something is not right

Edit: New bord core/mem-volt modded ready for aktion
Comments?

[mslezak]

Try to reset your bios with the jumper and pulling the power cord out for an hour or so. Sometimes that brings these guys back.

You have to be really careful around that unit. If you accidently connect 2 pins together, protection circuits shut it off.

You don't really need to know the resistance, you just need to change it. Resistance measurements are best taken when the device is OFF. By you simply attaching your probes to the circuit when it is on, can be changing the resistance of the area you are trying to measure. If your multimeter provides less resistance than the pathway on the board, it chooses your pathway!

Please either post your email or send me an email to nocman43202@yahoo.com and I'll send you my modified picture you sent me to show you where the resistor is located on the SK8N. As I said, that particular area appears to be arranged very closely to the SK8V (and why wouldn't it be - same manufacturer, same K8 power module - ADP3166). On another board, with the power OFF, use the continuity tester or, if you only have resistance, place one probe on pin 16 then press the other on the resistor SMD ends in the area until you find 0ohm resistance. From my findings, Pin16 appears to be connected, possibly inside the ADP3166, to pin22. Just to be safe, reduce the resistance of the resistor closest to Pin16 with 0ohm resistance, to be sure you are changing only that value.

If you are trying to measure the resistance of that resistor, put probes on either side with the power off. Note it should have a code on it as well (such as 103 = 10X1000= 10K). Now draw a pencil line over the top, connecting both tin leads of the SMD resistor. This is graphite and will serve to reduce the resistance of the resistor. You can now remeasure the resistance and record. Then boot up and try a higher vcore value. If it doesn't work, turn off, add another line, reboot and repeat procedure. If it appears you are reducing the max vcore, this is just as benificial information. If that is the case, please post and I can give you the quick procedure to work with.

I'll be joining you as soon as my SK8N gets back from Asus!
Speaking of, has anyone seen the great Peltier deals on ebay? I just got 170W- 220W dual peltier setup for under $13!!!! Major OC time coming...

[Tom Holck]

Thanks mslezak, I now when a MOB is dead
no help resetting bios disconnekting etc...

There was no error or pins connected.

The resistanse from pin 16 to ground when turn of is 5 ohm?

I dono if the corevoltmod is the same on SK8N and SK8V. I newer had a SK8N! But your welcome to post picture of the resistor on the SK8N (mail tom.holck@email.dk) that might give me some idea where to look on the SK8V..

Don't you think Hell-fires idea: some resistanse from pin 16 to ground is the right metode to get the OVP higher. Do I have to find a resistor connected to pin 16?

PS: Peltier? I only use Peliter to test the cooling on my compressorcooling

My coolinghead cover the area we are talking about, not easy to test when booting at -120C and 2.112 volt...

[mslezak]

SORRY JUST ADDED, PLEASE READ!!!

In response to your post, using a resistance to ground - this won't work for this mod. You are modifying the voltage sensed from the output inductors, just before the CPU. There is a resistor inbetween the CPU output voltage and the CSREF pin. If you bridge a resistor to ground at this point, you are shorting the output of your motherboard to ground!!!! That is a sure way to fry it!!!!


Try using the R38 location first. Run pencil lead over it a few times, connecting both tin ends over the resistor marking (103 or whatever the numbers are). Otherwise, if you just want to be able to adjust it with the condensor attached, solder a thin wire to either end of this resistor. You will need the resistance reading of it as well. The equation is 1/(1/R1 + 1/R2) for resistors in parallel. So if R1 is 100K and you add a 100K resistor, you get 1/(1/100 + 1/100) = 50K, be extremely careful! You should find a potentiometer with a value near infinite resistance that will start your resistance at the original value (R1), then you slowly decrease the resistance with the pot to up the OVP. This means you should use an EXTREMELY HIGH VALUE POTENTIOMETER - say 1 megaohm or higher (1,000,000ohms = 105marking on the potentiometer). Check to make sure the resistance is at the maximum when attached to your two soldered in wires, then boot and slowly reduce the resistance. Then get in the bios and up vcore. Repeat and chart your increase.

http://www.1728.com/resistrs.htm

As an added safety measure, you should put a resistor inline with your potentiometer, such as the value of R1, so that even if you accidently set 0 resistance, your value drops to half the default. But you may need less resistance, we haven't tested yet, so it would be best to just be careful to set your megaohm pot at max resistance between the two legs you have soldered to your two wires.

[mslezak]

I may have to move this to another topic, as it may be too far off the original subject. Forgot to mention, if you have many processors laying around, you may want to remove the IHS. It dropped my under load temperature by 10C on my Opteron 144. I have pictures of your FX-51 with it's IHS removed and it looks nearly identical. I was hoping the unlocked multiplier mod could be discovered by a few brave OCers by doing the mod. It is really hard to do - I had to use sandpaper and emery cloth to file a razor blade to about 1/2- 1/4 of usual thickness, then apply pressure all the way around the perimeter of the IHS seal about 2mm inward slicing the adhesive material. You have to start on the corners and hopefully get it cut and slide all the way around. Don't do ANY prying, regardless of what others have said. It took 2 days of very careful cutting and pushing before it came off! May enable you a higher OC if you can't up the OVP. Also we can measure resistance values on the processor to find what they changed at the factory, although it appears 4 value capacitors are lined all the way around the CPU core, with 2 locked in blue epoxy. Seems the blue are the processor multiplier/ default frequency select is my guess. I'd be curious to get my hands on a 244 to see what the difference is as well, could save a bundle this way converting 1x to 2x! They do have the 940pins supported for 3 more years if I remember right, way past the end of the 939pin, so we could benefit from this information for some time.

The other chip you may want to get your hands on, which may outperform the FX-53 in overclocking, is the Model 146 HE $417 off AMD's website. I think www.monarchcomputer.com is the only store that has an HE in stock, and they appear to only have the 246HE at $709. Otherwise, a search on google will yield OSK146ATBOX under $500, but not in stock for at least 2wks. http://www.computerhq.com/AMD_OPTERO...id-462024.html ships to Denmark $556.78 US Dollars total. It is the low power 55Watt 2ghz Opteron (89W standard), which, based on previous AMD low-power models (such as the mobile XP), can be taken much higher than the standard models. I have gotten my Opteron144 up to 130W or so at 2.3ghz. In comparison, this HE processor would be at 80W at 2.55ghz, or usable with the standard heatsink and fan (in theory) even at 116W at 3ghz! 170W at 3.5ghz 249W at 4ghz estimated ... I believe your condensor will do 300 or 400W of heat reduction or higher, so imagine the possibilities!! -Matt