Introduction:
Welcome to the third installment of MTB’s Overclocking Guides. In our previous article,
Overclocking 102: Taking Your Overclock to the Next Level, the concept
of clock dividers was explained, providing the fundamental knowledge needed to vastly increase system performance.
Through the careful selection of quality components, paired with a good understanding of BIOS settings, it is not
uncommon to see motherboards being run 66% faster than stock. Getting a system to run with 100% stability at those settings,
however, is another matter. In order to run a system that fast, changes need to be made in the operating physics of the
system itself. How so? This installment in our series looks at an integral part of overclocking: voltage adjustments.
NOTE: The information being provided in this article is for information purposes only. The author and the
MODTHEBOX.com website are not liable for any damage to person or property caused by attempting to modify any device
beyond the manufacturer's specified design.
Common Misconception:
I’d like to start off by addressing a commonly held view about increasing core voltages. Practically every major web
page for PC enthusiasts I’ve been on has a thread that basically says the same thing: "Increasing core voltages decreases
the life of your CPU". This, in itself, is a half truth. Most people will explain that if a processor has a useful life of
"x" amount of years, and increasing core voltages will decrease that value. This is simply not the case.
When a CPU dies of "old age", what has physically happened to the core has been the result of electro migration. Electro
migration, put in very simple terms, happens when the electrons running through the internal circuitry of a CPU stop
following their engineered paths, and start "jumping" onto paths they weren’t designed to take. The CPU has tolerances in
place to allow for this, but over time, electro migration becomes more and more of a factor, until finally the tolerance
levels are exceeded, and the chip fails to function as designed. In general, when a CPU is designed and tested, engineers
can determine the MTBF or Mean Time Between Failure of a given processor. What is interesting to note is the term "Mean Time".
This represents the average in a given group. In that group, there will be chips that fail much earlier than the norm and
there will be chips that last much longer than the norm. Overvolting a chip simply increases the chances of a chip failing
earlier than it should have, regardless of if the chip were only to last a few months or many years.
Understanding vCore:
If you ask anyone with some overclocking knowledge what to do when your overclocked system isn’t stable, 9 times out of 10 the
answer will be "up your voltage". As well, a system may not be able to run at a specific speed with stock voltages, but quite
easily runs that speed when core voltage is increased. We know it works, but the question is why does it work? Understanding
that will give you deeper insight to the inner workings of your system, and vastly increase your chances at successfully
creating and/or maintaining a highly-overclocked rig.
Let’s start off by understanding what vCore actually represents. Every one knows that computers work in binary; thousands of
one’s and zero’s being deciphered into something tangible that a user can work with. How does a CPU understand what a one or
a zero is? The answer is voltage. CPU’s contain gates, which can be compared to on/off switches. Those gates can have two
states, either a logic high (a one) or a logic low (a zero). It’s pretty easy to figure out that in most cases, a logic low
would be a voltage of 0V, but what about a voltage high? That’s where vCore comes in. The processors vCore value represents
a logic high, or the amount of voltage a gate would need to output a 1 instead of a 0. vCore values are dependant on the
CPU manufacturing process. Older processors might have run with vCore values well above 2 Volts, while current models like
the Pentium 4 chip run as low as 1.5 Volts.
In a perfect world, a value of 0V would register as a zero, and a value of 1.5V (for a P4 at least), would register as a 1.
As we all know nothing is ever perfect and because of that we have something called tolerances.
Tolerances & Overclocking
In our perfect world, anything not equal vCore should not be treated as a logic high. If we followed that model exactly,
just about every PC on every desktop would be crashing regularly. The reason: voltage is never a constant. Plug a
multi-meter into a Molex connector on one of your power supplies, or look at your voltages using a program like
Motherboard Monitor, and you’ll see how voltages fluctuate and deviate from the manufacturer’s specification. A 12V
rail on a power supply might only put out 11.85V, or a vCore setting of 1.8 Volts on a motherboard might have a true
value of 1.72V. Since those voltages can never be perfect, we have to compensate for it. The way that’s done is with
tolerances.
Suppose a gate which expects vCore to be 1.5 Volts gets a voltage of 1.42 Volts. Obviously, we would want the gate to
assume this is a 1, a logic high. However, under our model above, if it isn’t exactly 1.5 Volts, the system will treat
it as a logic low. There is where tolerances are needed since we can assume a value around 1.5 Volts should be interpreted
as a logic high. Now, that gate has a tolerance; a range of voltages that it will interpret as a logic high.
