Volt changes dont stay on ECS 650i

I have been overclocking my e4300 and i got it stable to 2.7 then went to 2.8 and it would post but was not stable.
Anyway i tried to increase the volts from stock 1.3 to 1.325 or higher to make it stable but any change i make to the vcore doesnt actually change it. I can see in CPU-Z 1.41 that the volts are still 1.3 or actually a little lower at 1.28 and the bios agrees that it is at 1.3
Any ideas about this?

Its generally called a vdroop. Well basically you're board isn't very quality... Any ECS motherboard isn't. Mainly it just has issues accepting changes, and then hardware on the board isn't capable of mainting a higher voltage than its meant for, mainly the caps on the board.

If you are going to do some serious overclocking invest in a new board, recapping an ECS would be a waste of time.

Hopefully snodblatter will see this, he has plenty of experience with vdroops

Vdroop

bump it up again. It could be a wrong reading by CPU-Z too. My abit ip35-e has some major drop too.

Is this your board? If so, looks pretty damn sweet, those caps are Sanyo OSCONs, can't really do much better

From the looks of it droop shouldn't be too bad, looks like a fairly standard 4 phase design, it should work well. Only problem I can see is not enough caps on the primary side, but whatever, it should still be OK, certainly not bad enough to stop all OC attempts.

Personally I wouldn't trust the software readings in CPU-Z at all, it's really quite inaccurate. However, unfortunately I don't see any easy test points where you could try it with a multimeter and see what it's doing, so I'm not sure how you could verify how much droop you're getting.

Other than somehow probing the legs of the caps from underneath the board, all I can suggest is keep bumping it up and if it works fine, don't worry about it Droop is perfectly normal anyway, your board will live longer with it

Yup thats my board.

I tried setting the vcore up to 1.4 and down to 1.2 and neither register as changing. CPUZ and the bios both agree it is at 1.3
I understand cpuz may not be accurate but should i trust what the bios is telling me?

The other thing is that when i increase it, I am still not stable.

Is my board really a pos that doesnt really have the features it claims?
I mean all the reviews i read about it before i bought it seemed to be able to change the vcore.

Hey, truptrup,

First of all, never trust software readings. Do not trust what CPU-Z reports. Basically you should not trust any software reporting for neither voltages nor temperatures. How come? Let me explain.

In short, I've written an article on this topic explaining and describing why software temperature and voltage reporting is not reliable at all. You can read it from here. Summing the article up, each motherboard has a Winbond (or an alternative) Super I/O chip, which uses mathematical equation to calculate your temperatures and voltages. That Super I/O chip is linked directly with the sensors but unfortunately the formula that is used to calculate the reported temps and voltages might be way off.

However, if either the temps or voltages are way off then the user tends to be aware of this and ignores the reported values. For example, if Speedfan or MBM tells you that your -12V is steadily +5V, and that your CPU temperature is -30C while running quite finely your computer on air-cooling; it does not require an engineering degree to find out that the reported values are way off. The equation is screwed up, the reported values are miscalculated and so forth. Then the user decides that s/he does not relies anymore on the reported readings. In this case it is pretty straightforward.

Nevertheless, the problem occurs when the reported values are not way off, they are just one or two degrees/floating points mistaken. Reporting 11.8v instead of 11.4v or 58C instead of 64C, this is more dangerous because it does not look 'obviously wrong.' Therefore, the user tends to believe these reported values. Why these values aren't reliable? There are various reasons for that. Sometimes the equation fails because other motherboard is detected and/or other sensor-types and then the values are interpreted via a non-compatible formula. On the other hand, sometimes the sensors fail, too. Then the values are correctly interpreted and reported but the diodes and sensors failed miserably.

Bottom-line is that you should never rely on your sensors and diodes if possible. After serious and continous testings if you've come to a conclusion that the reported values are approximately correct then you can rely upon them in the future. However, at first you need to test them manually! Grab a multimeter and measure your voltages on your own. Grab a thermal diode and stick it on your CPU, chipset, HDD, whatever device or IC. Then you can see within your eyes the correct results. Trust me, you'll be amazed. Also, integrated voltage monitoring sensors aren't updating that much frequently as they should, therefore, sometimes vdroops are missed (when the voltages drop and increase on a frequent basis, or due to high computer load or idle).

Check this guide on how to use a multimeter to measure your voltages. On how to measure your temps with external temperature diodes and sensors refer to my aformentioned article, even though, the process of installign a new sensors is pretty straightforward.
---------

In your case, you want to measure your vCore, not only your PSU voltages (5v, 12v, 3.3v). I'd advice checking your PSU voltages, too. Where to stick your multimeter probes to measure your vCore? Now this is a tricky question. You could reverse engineer the VRM circuitry (voltage regulation modulation) of your motherboard and then find out where it's appropiate to stick your multimeter probes, and ultimately find the accurate voltages. Sounds scary? It's not at all, let me explain.

Usually there are multiple electrical components near your CPU socket. There are MOSFETs that is followed by an inductors and a series of capacitors. In general, this is called the VRM circuitry, and this is from where and what feeds your processor directly. This is where the vCore is generated and maintained. Unfortunately, there is no de facto standard VRM design. Each board has its very own (or customized) VRM circuitry design. Therefore, there isn't a place where you can stick your probe and measure voltages. That place varies on each and every board. How to find out where is it? Read on!

Virtually, measuring the vcore in real-time implies a tricky process. Why? Because it is a relatively too low voltage but the curents that flow through are relatively large. The current is distributed from the vCore regulator to the bottom of your CPU via a copper plane. However, in your case, the board you own looks similar to a de facto standard 4-phase design. The basics of a 4-phase VRM circuitry are reachable through the World Wide Web... a little research via our friend, Google always helps.

You need to understand the main concepts on how a VRM circuitry works. The vCore regulator gets its feedback from a point of the plane, and according to these received values it tries to maintain (or generate) the appropiate voltage as seen from this feedback point. Furthermore, this voltage is not contact, therefore nor does the readings of it should be so-called 'stable.' This is where what we call 'vdroop' might happen. Almost always there is a significant difference between the 'load line' and the 'idle line.' Load happens when the processor is on load (cca. 100% CPU Usage), and idle when it is on idle (cca. 0% CPU Usage). You can call these as 'minimal' and 'maximal' voltages, too. The "on load" voltages stand for the 'minimal,' while the "on idle" represent the 'maximal.' Therefore, neither your or my vCore are stable. It is fluctuating. And calculating our average does not worth because we shouldn't rely on that. Average is just that, an average.

So, let's exemplify with a real-world scenario. You set 1.50v vCore in the BIOS. You boot up your system, you start off Prime95 stress testing, and voila, your real vCore is barely bordering 1.36v. You stop stress testing and then it averages around 1.48v. Sounds weird? Not at all. That's the vdroop. It is perfectly normal. It's the healthiest way. Preventing a board from 'vdrooping' you are drastically decreasing its life-span.

Okay, now you've understood the concepts. How to measure it by yourself? Let's see a few workarounds...

Firstly, you can measure the vCore on the output leg of that inductor. Some boards have copper planes behind the processor socket (on the back of the board); when the back of the socket is inside then they're locked on the inside/center of the socket. That's where you should see several capacitors (ceramic) that are connected in parallel. These are located between the vcore and the ground plane; therefore, you can put your DMM probes across one of those caps, too.

Some people advice measuring the vCore prior the inductor, that is, either on the legs of the MOSFET or one leg of the inductor. This is incorrect according to Johnny Bravo (kudos!). And if he says so, we should believe him. Nah, no doubt he's right. That's because measuring on this incorrect place the voltage fluctuates (turned on and off) several hundred times per second. Therefore, depending on the quality of your DMM, you will measure an average of this. On how accurate this average indeed is, we can say that it's roughly your vCore indeed, but that's not what we want. We want more accuracy!

Johnny Bravo advices measuring the voltage after the inductor. On older boards you can do this by measuring it from one of those legs of an inductor that points toward your CPU socket. He explains that this won't work for the newer 'cube' inductors. The workaround is by finding either an empty socket for a smoothening capacitor and use the positive terminal, or using the positive side of a SMD electrolytic capacitor. If you cannot approach neither of these places on the top side then try on the rear of the board. This way you will measure the accurate vCore.

Sometimes you might have access to the capacitor leads of those caps, which are following that one inductor I was talking about. This most often is on the rear of the motherboard. Measuring here is a bit tricky but using needless of the DMM the process can be done decently. Analyze your board, and decide which place is easier to measure. You can link this technique with the aforementioned one, and once you're working on the rear of the board, you can as well try Johnny Bravo's advice. Compare the measurements.

Another way to measure it is using your Winbond Super I/O chip. Remember? We've discussed about our I/O chip at the beginning of my post. In general, that aforementioned Winbond Super I/O chip has 8 channels to measure the voltages. IIRC, it has approx. 100 pins. Out of those pins each stand for something; in our case, one of those is for the vCore. The easiest way to get into the depths of it is to read the specifications of your Winbond chip, and then go and search Google for its official datasheet. It contains a ton of valuable information regarding that exact I/O chip you have. Therefore, all you need to do is grab out your DMM and have fun!

However, this needs a little bit of work-experience with electronics because the pins of the I/O chip are very tiny. And you want to measure accurately your vCore, not something different, right? As long as you can correctly use a DMM and to accurately place the needles/tips on the correct pins, then the job is done. Don't forget that you need the official datasheet because neither I, nor do you know the internals of our Winbond chips just like that.

Another alternative way just for the sake of it is measuring the vCore through the VCC_SENSE and VSS_SENSE. In case of a Pentium 4 processor (Prescott) check this datasheet from Intel, turn to page 45, and notice where those pins are located. You can approach the situation using this technique if and only if you have access to the back of your motherboard. Oh, and by the way, this is the professional way to do it. This is how laboratory engineers are doing it. This is from and how automation machineries are testing each board.

Have fun!

PS: Moral of this story is that you need to measure the voltages with a Digital Multi Meter on both idle and load.

EDIT: Here'a an edited version of Snod's attached image. I've sketched it in a few seconds. The image could have been larger, though, because I couldn't see the markings on the PCB/board nor on the chips/components.

Alright everything is working. thanks for all the help. That post madhyena really helped with understanding vcore and vdroop. Thanks man.

I ended up putting in the bios from 4-03-07 not the newest because the newest one wouldnt let me overclock at all.

Then i read this and it helped me alot with disabling features that would hurt stability and understanding the things i was changing better.
EDIT: forums wont let me post the link to the pdf that had that info so just google 'nforce_680i_sli_overclocking.pdf' It really helped oc n00bs like me.

Also i read that this board is much more stable at multiples of 100/133/166 so that helped me picking frequencies.
Hope this helps other people with this board i really like it now that i understand overclocking better as it does work well with that update.

ECS gave me this answer to my problem about the most recent bios update. I really enjoyed it. Thanks ECS
"We set a limit on the CPU overlocking features to avoid damaging the CPU since over-clocking doesn't really bring significant system performance boost, it's just a number game."

Can you explain this a bit further please, perhaps with some examples? Thanks.



Typical response from ECS. They bring out a great board and then ruin it with BIOS "updates" like the latest one, which I agree with you, kills overclocking completely.

What i mean by multiples of 100 OR 133 OR 166 i mean setting the fsb.
EX:
MY stable fsb have been
1100
1200
1300
1333
etc

Ok, thanks