Voltages being reduced too much witle CPU is in use on X58 UD5

bhavv

bhavv

bhavv

Permabanned
Joined
14 Nov 2009
Posts
13,638
Im having this problem that I set a particular voltage in my bios, and windows loads up with that voltage in idle mode, but when I start to run a stress test it reduces by around 0.5v. In order to get my CPU load voltage to 1.396 to be stable under load without BSODs, I am having to set my bios CPU voltage to 1.475v :x.

I have tried disabling all the power saving and core / voltage altering features in the bios (C1, C3, C5, EIST and something else), but the CPU voltage still keeps on dropping as soon as any load is placed on it.

How do I stop this happening? I dont want to have to boot with ridiculous idle volts just so I can get a stable lower voltage under load.
 
I found out its the Vdrop / Vdroop, and that this board completely overdoes the voltage reduction under load.
 
Does your CPU run ok at stock settings, if your trying to Overclock your CPU theres plenty of info on the net google it > i7 920 ex58 ud5 overclock
 
It runs fine at stock settings. I just had to enable LLC to reduce the Vdroop and now I dont need to set my idle voltages so high when overclocking.

I never knew about Vdrop / droop, which is what held me back from pushing further than 1.3v before. I read gigabytes own I7 overclocking guide, but it never mentioned LLC and Vdrop / Vdroop at all. The Vdroop things were just making my overlocks over 3.8 Ghz unstable.

I were even considering getting the Corsair H50 or Coolit water coolers, but looking at reviews my current temps with the Triton 88 are just the same or a little better since putting on Arctic MX3.
 
Last edited:
So does it mean your cpu load voltages go up now instead of down? As I have LLC enabled, 4ghz 1.275 vcore in bios, when using linx it will go up to 1.31 (everest reading).

It says it stays the same in cpu-z so not sure which one I should believe.

RoEy
 
RoEy said:
So does it mean your cpu load voltages go up now instead of down? As I have LLC enabled, 4ghz 1.275 vcore in bios, when using linx it will go up to 1.31 (everest reading).

It says it stays the same in cpu-z so not sure which one I should believe.

RoEy
Click to expand...

See, the voltage reduction on the UD5 is so bad, that even with LLC enabled it still goes down.

With LLC Disabled, setting the Bios voltage to 1.475v gave me 1.396v under load in CPUZ. With LLC enabled, setting the bios volts to 1.425v gives 1.396v under load.

The whole idea of reducing voltages while the CPU is under load is ridiculous. I have no more crashes at 4.2 Ghz and 1.396v and can leave my PC switched on all day long without any problems :)
 
Last edited:
bhavv said:
The whole idea of reducing voltages while the CPU is under load is ridiculous.
Click to expand...

Fool.

Idle voltages just don't matter very much. High temperatures and high voltage simultaneously does serious damage to processors, but either by itself isn't so bad. Since temperatures drop quickly when it idles, it's no biggie.

That might be misleading. I'm insulting bhavv for his attitude to and evident ignorance of electronics, and on an unrelated note saying that I don't think idle voltages matter very much.
 
Ummm, reducing voltages *under load* is ridiculous because it makes overclocking become unstable. That is all. Lower voltages are good, but dropping them automatically by around 0.8v while the CPU is being used is ridiculous, as opposed to simply making the default voltage of the chip that much in the first place.

Im talking specifically aabout Vdrop and Vdroop here, not reducing voltages in the Bios or reducing default voltages.

Not having any Vdrop / Vdroop at all, and simply allowing an I7 to run at 1.25v all the time at stock settings isnt going to do anything to damage the chip. It is actually far worse in terms of stability to have the CPU drop 0.8v below its rated voltage specification while it is being used then to leave it constantly at the voltage set in the Bios.
 
Last edited:
If you draw a current from any power supply, the voltage at which the current is drawn decreases. Voltage drops under load, always.

Vdrop nearly = vdroop. They're both voltage lowering from it's open loop value, when in windows its a bit lowered by the drawn current and called vdrop. When you fire up prime95, current drawn goes up significantly and so voltage goes down more. This is not ridiculous.

The approach taken by load line calibration and the like is to jerk the supply voltage up and down to try to maintain a constant voltage regardless of load. So vdroop is still present, you just pull the starting voltage up and try to keep pace. I don't think this is a good idea, faster voltage changes lead to greater spikes in the applied voltage. Too high may be dangerous, too low may cause instability. Intel don't seem to rate it as sensible either.
 
bhavv said:
Ummm, reducing voltages *under load* is ridiculous because it makes overclocking become unstable. That is all.
Click to expand...

You might want to read up on vdroop. If the voltage didnt drop, your overclock would also be unstable, as when load instantly ends, the voltage will instantly rise beyond the boards ability to reign it in, and your cpu will get voltage spiked way beyond what you have it set too in the bios.

http://www.anandtech.com/cpuchipsets/Intel/showdoc.aspx?i=3184&p=5

Overclocking brand boards sometimes have a bit more capacitance bolted on, but you then see them get slated in reviews for being less power efficient than other manufacturers boards using the same chipset.
 
Ridiculous may have been the wrong word, but it is still bad for overclocking because you dont want your voltage to decrease when overclocking.
 
Voltage has to decrease when load is applied. Its the laws of physics.

You are suggesting the board realise that its dropped, and then correct it back up. However what does it do with all that charge when Intel burn test suddenly comes to an end? It kicks the top off that Chinese capacitor is what it does.
 
I see. Im not an engineer or a physicist :).

I fail at maths or anything with calculations and numbers.

But it has been speculated that the Gigabyte X58 UD5 has a larger voltage decrease under load then other X58 motherboards do.
 
Last edited:
JonJ678 said:
Fool.

Idle voltages just don't matter very much. High temperatures and high voltage simultaneously does serious damage to processors, but either by itself isn't so bad. Since temperatures drop quickly when it idles, it's no biggie.

That might be misleading. I'm insulting bhavv for his attitude to and evident ignorance of electronics, and on an unrelated note saying that I don't think idle voltages matter very much.
Click to expand...

What do you think gives you the right to speak to people in that manner?

Who the hell do you think you are, with your big post count?

RoEy
 
Care to include what I was replying to, or does that blunt your point somewhat?

Every motherboard manufacturer out there is making boards that show vdroop. Rather than wonder why he condemned it as a ridiculous design and looked no further into it. I consider this foolish.

Post count indicates a reasonable amount of free time rather than extensive knowledge, for obvious reasons. I'd guess a third of my posts on here are useful. I'm not really interested in a needing a "right" to speak to people, the idea is faintly ridiculous. I'll judge you on what you write and hope you do the same with me.
 
They're the same thing? :confused:

Interested if anyone has any thoughts on what a large vdroop indicates about the overclocking potential of a motherboard. I think they'll be nearly independent but curious if anyone knows more details than this.
 
Last edited:
For any given resistive conductor the voltage drops at the end of the circuit from a combination of amperage, source voltage, ambient temperatures and conductor material. Its the reason power grids use higher voltages to reduce current leading to voltage drops. And why electronics use voltage regulators to attempt (transients) to hold a voltage to a loadline.

All is not as it appears anyway, you set a maximum voltage and the GTL circuits will stay under a set point to avoid overshoot during state changes. The mosfet regs may not supply what you're asking for either, depending on quality and design. And the voltage sensors and bios calibration can be miles off, a multimeter is only way to be sure exactly what is going on. Just remember Vdroop is a functional part of the Intel design to avoid spikes. Concentrate on the full load voltage needed for stability and forget about the off duty values.

Vdrop is the term generally used to refer to the 'apparent' difference between the bios set voltage and ambiguously the bios hardware monitor value or idle voltage in the OS. Vdrop=0 can happen, but it's normally an offset.
 
Last edited:
No need to come out calling people fools jon. The dude didn't know, fairplay to you for telling him and making a good point...but no need to start insulting.
 
You must log in or register to reply here.
Back
Top Bottom