P4 and Vcore - Explained

[mandacat]

P4 and VCore - Explained

There always seems to be confusion about Pentium 4 “Vcore”. What I'm about to discuss can be confirmed by downloading a copy of the CPU datasheet (whichever CPU you have) from the Intel website.

On a P4 Mainboard, “VCore” is often referred to as the Vcc (DC Voltage) supplied to power the CPU and is identified by the CPU’s internal “VID” settings. VCore is supplied to the CPU by an on-board regulator. This on-board, “phased” regulator gets its power from the Power Supply’s Regulated +12 Volts and is most often identified by its multiple MOSFET Transistors, Small Donut / Ferrite Transformers, and High Heat output. The on-board regulator sets its output voltage, automatically, from a circuit that reads the “VID” settings in the CPU.

VCore on a .13 Micron Pentium 4 CPU is generally specified to be on the order of 1.500 - 1.550 Volts (although there are some newer 2.8 Gig + units which may vary somewhat from those specs).

It is very important to understand that Intel sets their CPU “VCore” specification at “0” CPU current ( NO CPU Load). Actually, this condition is never visible since reading the VCore will cause some small load on the CPU.

The best we can do for reading VCore at the lowest CPU load is when the CPU is close to “IDLE” in Windows. At this very low load, VCore should be at its HIGHEST SETTING but, since the CPU is still under a small load, the VCore should actually be BELOW spec. On a properly designed mainboard, the VCore reading should be from .025 -.050 volts BELOW the “VID” (specified CPU VCore). This means that a CPU, which has a specified VCore of 1.525 Volts, should actually read between 1.475 – 1.525 Volts at IDLE in Windows.

More important now is the specified VCore behavior as the CPU Load increases. As the CPU load increases, the increased CPU Current should result in VCore Dropping proportionally. In other words, the more the CPU Load increases, the more VCore Decreases. This behavior is specified by Intel, is designed into the regulator, and is NORMAL.

As an example, take an Intel P4 2.53 Gig., 533 CPU, with a Vcore (VID) spec of 1.525 Volts. With “NO LOAD” VCore should read from 1.475-1.525 Volts. At its Full Load of 52.5 Amps, VCore should read between 1.346 – 1.425 Volts. Reference the Intel CPU datasheet.

VCore can be easily measured by connecting an accurate Digital Voltmeter (DVM) from the regulator output to ground. This eliminates possible Bios or Software conversion errors and is a good way to validate the output of the on-board regulator.

[beware430]

Originally posted by mandacat
P4 and VCore - Explained

There always seems to be confusion about Pentium 4 “Vcore”. What I'm about to discuss can be confirmed by downloading a copy of the CPU datasheet (whichever CPU you have) from the Intel website.

On a P4 Mainboard, “VCore” is often referred to as the Vcc (DC Voltage) supplied to power the CPU and is identified by the CPU’s internal “VID” settings. VCore is supplied to the CPU by an on-board regulator. This on-board, “phased” regulator gets its power from the Power Supply’s Regulated +12 Volts and is most often identified by its multiple MOSFET Transistors, Small Donut / Ferrite Transformers, and High Heat output. The on-board regulator sets its output voltage, automatically, from a circuit that reads the “VID” settings in the CPU.

VCore on a .13 Micron Pentium 4 CPU is generally specified to be on the order of 1.500 - 1.550 Volts (although there are some newer 2.8 Gig + units which may vary somewhat from those specs).

It is very important to understand that Intel sets their CPU “VCore” specification at “0” CPU current ( NO CPU Load). Actually, this condition is never visible since reading the VCore will cause some small load on the CPU.

The best we can do for reading VCore at the lowest CPU load is when the CPU is close to “IDLE” in Windows. At this very low load, VCore should be at its HIGHEST SETTING but, since the CPU is still under a small load, the VCore should actually be BELOW spec. On a properly designed mainboard, the VCore reading should be from .025 -.050 volts BELOW the “VID” (specified CPU VCore). This means that a CPU, which has a specified VCore of 1.525 Volts, should actually read between 1.475 – 1.525 Volts at IDLE in Windows.

More important now is the specified VCore behavior as the CPU Load increases. As the CPU load increases, the increased CPU Current should result in VCore Dropping proportionally. In other words, the more the CPU Load increases, the more VCore Decreases. This behavior is specified by Intel, is designed into the regulator, and is NORMAL.

As an example, take an Intel P4 2.53 Gig., 533 CPU, with a Vcore (VID) spec of 1.525 Volts. With “NO LOAD” VCore should read from 1.475-1.525 Volts. At its Full Load of 52.5 Amps, VCore should read between 1.346 – 1.425 Volts. Reference the Intel CPU datasheet.

VCore can be easily measured by connecting an accurate Digital Voltmeter (DVM) from the regulator output to ground. This eliminates possible Bios or Software conversion errors and is a good way to validate the output of the on-board regulator.

Does this explain why Vcore needs to be increased in OC'ing situations?

[freecableguy]

What you're describing is called droop. A regulator's output voltage will decrease proportional to load (current). This is a great concept will dealing with multiple power supplies/generators as it allows for load sharing (beyong the scope of this topic). This is, however, NOT ideal when dealing with a single power supply and a single load (CPU). I am surprised to hear that Intel has inherently designed CPUs to work this way. The job of a voltage regulator is to keep voltage as close to constant as possible over ALL ranges of current output. If CPU load where to increase drawing more amperage and the supply voltage were to decrease the situation would be one in which the power level (P=VI) would remain close to unchanged. This is NOT good for any sitauation (especially for OCers) as this can cause massive instability. Like I said, I have a hard time chocking this down. I'm going to have to read over the Intel whitepapers and get back to you on this one. For the record though, I do agree with the first point - a no-load voltage should be the highest output of a voltage regulator. True. But with a properly designed regulator this voltage is very close to constant for all load conditions.

-kjb