8RDA+ high Vcore shutdown fix!

[Spartacus]

The 8RDA+ is a great board, about the best around for AMD overclocking right now. Even so, a lot of dedicated mobo abusers (myself included) have had a problem where the 8RDA+ would just shutdown at Vcore settings over 1.9v even though the board is rated to provide 2.2v Vcore.

The symptoms of this problem are that while running the CPU at 100% load (Prime) the system will totally shutdown. The system will often boot fine to the Windows desktop even with Vcore set at 2.0v. The shutdown can happen instantly upon starting Prime or sometimes not for 30 minutes. Generally, the higher the Vcore is set, the quicker the shutdown. My 8RDA+ would run Prime for quite a while at 1.925v but would shutdown eventually. Setting my Vcore for 2.0v guaranteed a shutdown in 30 seconds or less.

This problem makes it impossible to achieve a decent overclock because the 8RDA+ shuts down much too soon. I spent a lot of time looking at CPU temp problems, even going so far as to disable the AXP's internal diode for temp sensing. High CPU temp causing the shutdowns didn't make sense though because I was able to run the same 2400+ cpu at 57C (!) perfectly stable on my old KR7A. The 8RDA+ was shutting down with CPU temps only around 40C. Next I thought that maybe the 8RDA+ MOSFETS or voltage regulator were overheating. Nope, same symptoms with cold air blasting at the power section on the mobo.

OK, I've covered a lot of possibilities but no payoff yet. I started looking at the IRU3055 Vcore voltage regulator. It's job is to control the switching of the power MOSFETS providing power for the CPU. Looking at the datasheet:

http://www.irf.com/product-info/data...ta/iru3055.pdf

I found that the IRU3055 also has overvoltage and overcurrent protection. Hmmm, that's sounds like something I need to look at. I originally thought that maybe cutting or lifting pins 15 & 19 might keep the overvoltage circuit from shutting down the system. Then I realized that there is no overvoltage condition causing the shutdown. The voltage in the MBM log is always fine at shutdown. Ok, that leaves the overcurrent protection circuit.

Pin 16 of the IRU3055 is called "OCSet". It needs an external resistor to provide the threshold for the overcurrent shutdown. So Epox would need to pick a resistor value to protect the power section of the mobo (and CPU) and make sure that the value they select will allow high enough current for overclocked conditions but still shutdown the system before a meltdown. Could that be where the problem is? YES!

Due to mobo component variables and operating temp variables, the value of the resistor (I didn't measure it) that Epox selected is causing the system to shutdown when it should not. In short, the IRU3055 is being programmed to be over protective.

Here's the fix: Cut the trace leading to pin 16 on the IRU3055.
I used a flashlight, magnifying glass, and an X-ACTO razor knife to cut the trace. In my system, I pulled the power supply aside and could see the IRU3055 very clearly. It took about 30 seconds. Sorry, no pics. I don't have a decent digicam available.

Now the IRU3055 has no idea how much current is flowing through the MOSFETs. It will happily supply current until everything overheats, your system catches on fire, and burns your house down! That's the way every hardened overclocker wants it . YEAH!

Now I don't really think that will happen, but if it does, I will laugh at you and call you names! If you choose to do this mod, it is ALL AT YOUR RISK! This mod will also void your warranty (duh). The "right" way to do this mod would be to replace the resistor with a value that will allow more CPU power, but still provide some protection. Just cutting the trace completely disables overcurrent protection. You have been warned! The circuit is not a feedback/correction loop though, only a protection circuit.

Good luck guys!

NOTE: If you don't feel comfortable doing board mods and cutting traces, PLEASE don't do this mod yourself. Get help!

DDTUNG: Thanks for posting so many of your mods, I posted this mod here first.

[Jeff]

Nice find. :thumbsup:

[Wolf]

Good job on finding the fix but there is no reason you should have to do that to get the board to run at 2.2v. Looks like Epox screwed up the power regualtion again on this board. I was actually suprised when my 8k3a+ didn't have any power issues like the 8k7a or 8kha but it looks like they are back to their old ways. If it were me, I would have rmad it and waited for a better revision to come. I think it's rediculous you have to cut a trace on the board for it to run over 1.9v but I think your fix will make a lot of people happy that were having problems.

[DisposableHero]

hell yeah! nice work...

cant i just cut the pin on the chip? that way i could fold it back down making it look like nothing ever was changed if something went wrong?

[Spartacus]

Thanks guys.

I agree Wolf, I'm pissed that this was necessary too. I already RMA'ed my first 8RDA+ for the same reason. The fact is though that the 8DRA+ is the best AMD board around right now. I was determined to make it work or smoke it. If I had smoked it, I probably would have gone back to my ultra-stable KR7A-R, or maybe sprung for a KX7-333.

I wish Epox and the other mobo makers would build better boards. I wouldn't mind spending another 15% for the mobo if that's what it took to make the freakin' boards work!

[Spartacus]

Originally posted by DisposableHero
hell yeah! nice work...

cant i just cut the pin on the chip? that way i could fold it back down making it look like nothing ever was changed if something went wrong?

You can barely see the cut trace anyway. You'd have to really know what to look for.

I wouldn't cut the pin, it could snap off. If you cut the trace you can undo the mod by just soldering across the trace or use a short single strand of bare wire to connect the pin with the resistor again.

I've been running since earlier today with no shutdowns.

[DDTUNG]

Spartacus

Thanks for the info. I've taken a quick look at the data sheets and it appears that your way will work. However, a better way would be to replace Rset(2.2k) with say a 4.7k resistor. Since I don't own the board I cannot tell which resistor Rset is, but perhaps someone will take the time to find out which resistor is connected to pin 16 and post back here.

DDTUNG

[Spartacus]

Thanks DDTUNG,

I agree, the better way would be to change the value on the resistor. I think the resistor number was R678 and it's connected to pin 16 by a short trace. I envisioned a long testing process to find the right value to allow more current and still be functional for overcurrent protection. And how would you test to verify overcurrent protection? I hate the smell of burnt transistors.

Going up in resistor value would decrease current on pin16 so would that increase or decrease the actual shutdown threshold?

I'm happy with it the way it is. It's been running fine all day now with no shutdowns or other surprises. If somebody does take the time to re-engineer the circuit and come up with a better solution I'd be interested to see it. I'm not sure that the proper value for one board will work right on another board though. That's apparently what Epox did, but it didn't work.

Thanks again DDTUNG, it's fun to converse with a legend.

[DisposableHero]

is there anyway you could possibly take a POS picture of where u cut

or even draw it?

[DDTUNG]

Spartacus

The data sheet gives a formula for VOCset which leads me to conclude that the proper way is to increase Rset.

VOCset = 160m(icro)AxRset - Rds(on)xiL1

When VOCset turns negative the system restarts.

My thinking at this point would be to replace Rset(R678) with a 4.7k or 5k VR set initially to 2.2k(assuming R678=2.2k, which you can read off R678) and turn UP the resistance of the VR gradually to increase the trip current level.

DDTUNG

[DisposableHero]

is there really any point for this current monitoring... how would it get that high anyway...

and adding a resistor is no easy task let me tell you

[Spartacus]

Originally posted by DDTUNG

My thinking at this point would be to replace Rset(R678) with a 4.7k or 5k VR set initially to 2.2k(assuming R678=2.2k, which you can read off R678) and turn UP the resistance of the VR gradually to increase the trip current level.

DDTUNG

I knew you'd have the right answer.

That makes sense, just turn up the pot little by little until the shutdowns stop. Now why didn't I think of that.

[Spartacus]

Originally posted by DisposableHero
is there anyway you could possibly take a POS picture of where u cut

or even draw it?

The only digicam I have access to is my daughters toy camera. Trust me, it's no good to see any kind of detail work.

If you can post a close-up pic of the chip on the 8RDA, I'll grab it and draw in where the cut in the trace goes.

Let me know.

[SupaMan]

how deep do u have to cut? just skim it? or do ya actually have to do some pushing?

[Spartacus]

You have to use a little pressure but not much. The trace is copper foil thinner than a piece of paper and really just sits on top of the board. Four or five passes with the blade should do it.

Try to imagine if you glued a thin strip of tin foil to a flat piece of plastic. It would take just a few cuts at lite pressure to cut through it.

[muzz]

WTG Guys......... :thumbsup:

[Spartacus]

Thanks muzz!

Here's a screen cap showing the system running fine with Vcore at 1.975v. This would have caused a shutdown within 2-3 minutes before the mod.

[bendside]

OK, so I've got the same problem that you had, except much worse (unstable in prime95 at 1.75 volts!!!). I looked at the pdf file you linked to, and I think I found the chip on the mobo, but I want to make sure I'm seeing the right thing. It is way up in the top left corner of the board, right below the psu, right? And the 16 pin is the one on the top, the 3rd one from the left? And the trace would be the lighter green strip running to the pin? Sorry if I sound like a dolt, but I've never done something like this before.

[Spartacus]

This should help a little...

[bendside]

So all the trace is is the strip that's a diff color? And was my description of the location of the chip right? Sorry, and many thanks for the diagram.

[Spartacus]

You should really RMA that board if it's that bad. Is that running at stock speed?

If it is, please register over at AOA forums or AMDMB forums and drop a note to the Epox guys. They should hear about this one.

To answer your questions though, yes it sounds like you have the correct chip. It's the 36pin chip near the big MOSFET transistors and the coils, right behind the PS/2 connectors. Make sure you verify the part number before you do anything and have somebody with some experience give you a hand if possible.

The trace itself is a thin layer of copper foil with a protective layer of green stuff called solder resist.

[Spartacus]

Originally posted by bendside
OK, so I've got the same problem that you had, except much worse (unstable in prime95 at 1.75 volts!!!).

Just want you to know this will not fix instability in Prime95, as far as errors and things like that. We are talking about a system that can run Prime just fine but will completely shutdown under heavy CPU load.

If the only problem is errors in Prime DON'T DO THIS MOD!

[bendside]

Yeah, i know. The prob is a complete shutdown, not roundoff errors or anything like that. Thanks for the heads up, though.

[bendside]

It's not at stock speed, but it's at a speed where, with stock voltage, it is perfectly stable through hours upon hours of prime 95. It's pretty clear that the voltage is the problem.

[TechTones]

I have the same no can do over 1.90 problem. My 2400+ did 2500mhz in the 8K3a+ at 2.10v On this board is just shuts down over 1.90 I put a fan on the vcore VRM. Now it BSOD over 1.90

The highest I can take this chip is like 2200. Did cutting that trace allow you to OC the CPU higher?