Undervolting a Barton 3000+ dangerous?

[mista ting]

I'm on a Amd "Barton" 3000+/400fsb it came completly locked :cry: so since i dont wana oc via fsb agian, im thinking to just leave it at stock speed and undervolt it to decrease temps, is undervolting dangerous? whats the lowest volt i can go with the cpu still working good? (Abit NF7-S V2.0) :santa2:

[DarkPurity]

Undervolting CAN be dangerous, so be careful.

You don't want to screw too much with the amperage by playing with your voltage. Processors have certain tolerances, and while over-volting can be marginally cured with good cooling, you can't really do anything to prevent damage when undervolting.

[electrorcamd]

How does undervolting damage a cpu? I thought it would extend it's life since it would be running cooler. The lowest you can go will vary by the chip. Lower it, check for stability just like if you overclocked it, repeat until it is stable. If you want, since DarkPurity said it can be dangerous, raise it a little from the lowest stable voltage.

[DarkPurity]

If I remember correctly, which I probably don't since I don't screw with voltage much, but if I do:

Voltage is inversely proportional to amperage. Therefore, if your frequency remains constant, which it does to a degree, then the lower your voltage, the higher your amperage. In layman's terms, that means if electricity were water, and you were pouring the same amount of water down two pipes at the same time, the water will move faster through the narrow pipe than it does the fat pipe. You might be able to get more water into the fat pipe at a time, therefore getting more water through the pipe for the stress (an overclock). The narrow pipe will undergo more stress on its bends (switches) and could burst easily if the water moves too fast or it takes on too much at a time.

Thus, higher amperage can do damage. You could take 1289371629836 volts to the wrists and be ok as long as the amperage was low enough (not that such a thing would realistically happen).

[electrorcamd]

Okay, I just did not know that it really worked like that. I just always thought running the lowest stable vcore was best, as that would yield lower temperatures. I especially thought this because the Mobiles, which are the same cores as desktops, run a lower voltage safely.

[DarkPurity]

Lowest stable IS best, so you got that much right. I'm just saying be careful with it and make SURE it's stable, because sometimes the stress doesn't immediately appear. Usually it shows up in the form of failed boots or failed POSTS, or random illegal operations.

AMD recently released that the lowest safe tolerance for their recent mobiles is 1.0v, so I wouldn't take it lower than that, but there is definitely merit to clocking it down.

[Biggles 266]

Thus, higher amperage can do damage. You could take 1289371629836 volts to the wrists and be ok as long as the amperage was low enough (not that such a thing would realistically happen).

Of course if you did, whilst it would be the current that affects your heart, that sort of voltage would give you massive burns... which isn't quite the same thing as being OK.

[DarkPurity]

LOL thus the unrealistic figures.

Of course, I suppose if you strapped yourself to an oversized heat sink, like a tanker semi filled with mercury, you might be closer to OK ;).

[Dragonprince]

Undervolting the CPU is not dangerous at all. Long before you could damage anything the chip would become unstable. Also there's a reason they build regulation circuitry into the board....

[Graphicism]

Just go for it man, I did...

[mista ting]

ty for the info, thats one hell of a undervolt Graphicism, 0.992v wow!! :eek:

[Moto7451]

Just wondering but why are you staying away from overclocking the FSB?

[Jerryhey]

Just wondering but why are you staying away from overclocking the FSB?


Probably because the 3000+ isn't a good oc'er most of the time (low headroom)
and usually runs a bit hotter. Eh, my guess at least.

[mista ting]

because this barton is crappy for ocing my memory is top of the line stuff, if i increase the fsb by one mhz it becomes really unstable i tried diffrent ram same thing happens

[electrorcamd]

I'd say go ahead and get it to the lowest stable voltage, maybe raise it a small amount if you are worried about it. I don't think your 3000+ could go below that 1.0v safe limit set by AMD, unless you underclock it. My 1.6GHz Duron only got down to about 1.2v stable, and that uses less power than a Barton, so I wouldn't worry about it damaging it.

[doublejack]

If I remember correctly, which I probably don't since I don't screw with voltage much, but if I do:

Voltage is inversely proportional to amperage. Therefore, if your frequency remains constant, which it does to a degree, then the lower your voltage, the higher your amperage. In layman's terms, that means if electricity were water, and you were pouring the same amount of water down two pipes at the same time, the water will move faster through the narrow pipe than it does the fat pipe. You might be able to get more water into the fat pipe at a time, therefore getting more water through the pipe for the stress (an overclock). The narrow pipe will undergo more stress on its bends (switches) and could burst easily if the water moves too fast or it takes on too much at a time.

Thus, higher amperage can do damage. You could take 1289371629836 volts to the wrists and be ok as long as the amperage was low enough (not that such a thing would realistically happen).


Hmm. Somehow I'm skeptical this is correct. Ohm's Law states:

Voltage = Current (amperage) * Resistance

So clearly if resistance is constant regardless of vcore and clock frequency changes then the lower the vcore the lower the current becomes. So if this resistance assumption is correct then undervolting couldn't possibly cause damage from too much current, as current would decrease as vcore is lowered.

The only question is, does the resistance change? This I don't know the answer to but I suspect it stays pretty constant. It probably goes a little higher which results in more leakdown current (especially noted as an issue for 90nm chips). However, even if it does go up it would have to change larger proportionally to the change in voltage for the inverse relationship (lower voltage yielding higher current) to work. I'm doubtful this is the case unless we're talking about extreme ends of the voltage spectrum. But I don't know this as fact, it's just guesswork on my part.

[Dragonprince]

Your oversimplifing but basicaly correct. V = I*R is only valid in simplier circuits, as circuits become complex, like in a CPU, inductive and capacitive issues cause errors in computing V = I*R that have to be handled. For the most part you are corrct in that there are not any varistors in a CPU that im aware of, but there are plenty of gated circuits that depend on certain voltages being present or they fail to open and close properly...

[Petr]

doublejack is correct. When you decrease voltage, current decreases as well. Resistance depends on temperature. The higher the temperature, the lower the resistance of transistors (and higher of copper interconnects). So if you decrease voltage and current decreases because of that, the thermal output decreases too (Watts = Voltage * Amps). So the temperature is lower, resistance decreases. That means Amps decreased but somewhat slower than linear because Voltage (decreasing) = Amps (decreasing) * Resistance (decreasing). On the other hand resistance on interconnects decreased what means better conductivity on interconnects and better stability as a result - CPUs are often limited by interconnects' speed (remember Thoroughbred A0 vs B0).

[IrQ]

doublejack is correct. When you decrease voltage, current decreases as well. Resistance depends on temperature. The higher the temperature, the lower the resistance of transistors (and higher of copper interconnects). So if you decrease voltage and current decreases because of that, the thermal output decreases too (Watts = Voltage * Amps). So the temperature is lower, resistance decreases. That means Amps decreased but somewhat slower than linear because Voltage (decreasing) = Amps (decreasing) * Resistance (decreasing). On the other hand resistance on interconnects decreased what means better conductivity on interconnects and better stability as a result - CPUs are often limited by interconnects' speed (remember Thoroughbred A0 vs B0).


All of this i for the most part right.

Also this is right:
Originally Posted by DarkPurity
If I remember correctly, which I probably don't since I don't screw with voltage much, but if I do:

Voltage is inversely proportional to amperage. Therefore, if your frequency remains constant, which it does to a degree, then the lower your voltage, the higher your amperage. In layman's terms, that means if electricity were water, and you were pouring the same amount of water down two pipes at the same time, the water will move faster through the narrow pipe than it does the fat pipe. You might be able to get more water into the fat pipe at a time, therefore getting more water through the pipe for the stress (an overclock). The narrow pipe will undergo more stress on its bends (switches) and could burst easily if the water moves too fast or it takes on too much at a time.

Thus, higher amperage can do damage. You could take 1289371629836 volts to the wrists and be ok as long as the amperage was low enough (not that such a thing would realistically happen).


But you also have to take in to effect that the transistors in a processor are so small and low powered that they will work with almost nor current at all. It takes a LOT less energy to open a 1ft by 1ft wodden gate then it does to open a 10ft by 10 ft wooden gate. Basic laws of physics take into play there. Since the component is physically that much small and electricity is nothing more then a physical force it will require much less force to open that gate in a transistor in a processor then it will to open a transistor in a radio.

On the same though volts*watts equal your total amperage. Your wattage will stay the same but if you lower your voltage you lower you ramp draw too. That can be dangerous as electronic devices need so many amps or they fault and dont work.

It would be kind of like a brown out slowly kiling your electronic device. It won't happen right away like a energy spike would but it will happen over time. Adding more amps is not an issue but to few is.

But at the same time you have to remember electronic theory is still in it's infancy. Every rule you've seen here is still a guess technically and nothing has been full blown proven. For god sakes we still don't even know if they electronc flow from the negative pole to positive pole or from the positive pole to the negative pole. We THINK it flows from negative to positive but the next big discovery can change that next year and might change back a year after that. It's still in its infancy and everything is a guess pass that. Electronic theory is based on statistical theory which is also the most incorrect and least accurate math Man has created.

[Petr]

If you decrease voltage, Watts decrease too. So this statement "then the lower your voltage, the higher your amperage" is against basic law of physics (Ohm law). If it stayed the same then why increasing voltage when overclocking would increase thermal output? And the other way if you lower voltage, then thermal output decreases.

decreasing voltage => decreasing Watts
increasing voltage => increasing Watts

since
Watts = Voltage * Amps
and
Voltage = Amps * Resistance

then Watts = Voltage * Voltage / Resistance

If Resistance stays the same (and it does to some degree), then lowering voltage results in lowered Watts in faster rate than linear.

[Quailane]

You might find this interesting.
http://www20.tomshardware.com/cpu/20041001/index.html

[mista ting]

thanks for the info ^_^

[=ACID RAIN=]

Hehe.... I have my 2100+ tbred overclocked slightly and undervolted at the same time. Given that the CPU runs at 2200 at default voltage, I figured I could run 1800 (200x9 for good fsb speed) with less than default. Works great, so I made it my gameserver when I upgraded my game rig. On watercooling with the fan at its lowest possible speed, it runs about 35C. Perfect in my opinion, and very quiet.