Relationship between voltages

[GraySplatter]

So I've been testing various settings with my E6600 and 680i for stability and have come across something I don't understand.

At stock voltage, I can run a reduced multiplier on the CPU with a FSB of 1466, Orthos stable for 8 hours. I can also run 1200 FSB at the full 9x for a 2700MHz CPU 8 hours stable at stock.

Just trying to get stable at 1333 FSB 9x for 3.0GHz is driving me nuts. I've raised the Vcore all the way from 1.2 to 1.4 and still only get about 10 minutes on Orthos. But when I bump up the Vfsb from 1.2 to 1.3 and Vspp from 1.2 to 1.25, it stabilizes.

If the FSB can run at 1466 with a reduced multiplier on the CPU, why do I have to bump up the voltage on the FSB to get a slower FSB with the higher multiplier to run stable?

And yeah, I know, I can bump these voltages up a lot to go higher. I'm just going through the process of seeing how everything is affected. That's how us n00bs learn!

So what is the relationship between the voltages on the CPU, FSB, northbridge, memory, etc?

And how about that Vdroop from 1.4 to 1.33ish under load!?!? Sheeeesh!

Jeff

 

[Roguestar]

You're multiplying the FSB by 4 because it's quad pumped. It's easier to think of (and explain to other people) if you refer to its normal speed.

Eh, long story short (with parts missing too) the vCore raised allows the CPU to run faster but generates more heat. To increase the FSB stably you need to increase the vFSB (but only by a tiny bit, usually +0.1) and to run most high-end RAM even at stock they want +0.3v (because they work best at 2.1 but the DDR2 specification only states 1.8). Northbridge voltage I upped by 0.1 on the recommendation of others but I have had little to do with it since to be honest.

 

[GraySplatter]

Ok, so using the real speed numbers, why can I run a FSB of 367 with CPU x6 multiplier stable at stock voltage, but have to increase not only the Vcore but the Vfsb to get FSB 333 with CPU x9 stable?

There must be a technical explanation for it, but I haven't run across it in my browsings.

 

[DrMrLordX]

Originally posted by: jeffwegher

So what is the relationship between the voltages on the CPU, FSB, northbridge, memory, etc?

The short answer here is: most people wouldn't know. I don't. Maybe some engineers from Nvidia could clear it up, but under normal circumstances there is no real relationship between vcore, vdimm, etc.

And how about that Vdroop from 1.4 to 1.33ish under load!?!? Sheeeesh!

This could be part of the problem here. If your vcore is drooping by .07v under load, you may be getting vdroop on other components as well. Whether or not this is a problem with your PSU or your motherboard is open to speculation; I don't know which PSU you're using, so I couldn't even begin to guess.

 

[GraySplatter]

I have the Seasonic M-12 700. So, I doubt that's causing a problem -- but I haven't put a multimeter on it, either.

Just odd with the voltages. I'm pretty sure I have a weak E6600 for overclocking which may be the main problem.

Jeff

 

[DrMrLordX]

If you've got a hefty PSU like that, my guess is that it could be a motherboard issue (at least, the vdroop is).

 

[Aluvus]

The answer is simpler than you might think: processor frequency.

At a 6x multiplier and 367 MHz (quad-pumped) FSB, the processor is operating just over 2.2 GHz. At 9x and 300 MHz, it's at 2.7 GHz. At higher frequencies, the processor needs the data coming in on the Front Side Bus to change state faster, because it needs to sample that data more often. The voltage change needs to be quicker and easier to pick up.

And the way to do that is to make it move across a larger range. Because it is swinging across a broader range, it takes less time to reach a value that the CPU can interpret as definitely a 1 or definitely a 0. After all, the Northbridge is attempting to make the voltage change a greater amount in the same amount of time, if you keep the FSB speed the same.

When you increase Vfsb, you are increasing the high signaling voltage that the Northbridge uses (the low signaling voltage stays at ground) to send data to the CPU, and therefore making life easier for the CPU. Assuming you don't go too high, anyway.

 

[DrMrLordX]

Interesting. I figured it was board-related but I didn't expect anything quite like that. Thanks for the explanation Aluvus.

 

[GraySplatter]

That is a great, straight-forward answer and it makes complete sense.

Thanks!!