GA-P31-ES3G Manual Vcore Problem

I'm overclocking my E5200, or trying to. I set voltages to Manual, Vcore
to 1.35, and in the Health BIOS screen it shows near that, I have C1E
and EIST disabled.

In Windows cpu-z and Gigabyte Easy Tune show at or near 1.25, like they
do when voltages aren't on Manual.

I have the chipset drivers installed. Could they be resetting the
voltage? Or is it a bad BIOS?
--
Ed Light

Better World News TV Channel:
http://realnews.com

Iraq Veterans Against the War and Related:
http://ivaw.org
http://couragetoresist.org
http://antiwar.com

Send spam to the FTC at

Thanks, robots.

Ed Light wrote:
> I'm overclocking my E5200, or trying to. I set voltages to Manual, Vcore
> to 1.35, and in the Health BIOS screen it shows near that, I have C1E
> and EIST disabled.
>
> In Windows cpu-z and Gigabyte Easy Tune show at or near 1.25, like they
> do when voltages aren't on Manual.
>
> I have the chipset drivers installed. Could they be resetting the
> voltage? Or is it a bad BIOS?

As far as I know, the boost is added, after the range limited VID
is applied. The Intel processor supports EIST, and a range from
"low" to "high" for the control registers on the processor. VID
controls the voltage, and FID the multiplier. And FID and VID changes
are used by EIST, to conserve power when the processor is idle. The
range of allowed FID and VID is limited, to help prevent people
from overclocking.

This is no good for overclocking, because the range limit Intel
applies to the VID register, doesn't allow it to be adjusted
far enough to do the job.

So instead, a "boost" has to be applied to Vcore some how.
On my previous motherboard, there is a pin that sums control
currents, and it can be used for "boosting". I don't think
my old board supported "undervolting", only boosting is supported.

VID ----------------------- Vcore Regulator ---- final voltage equals
register +------ (Boost pin) range of values from VID
| plus fixed positive offset
GPIO ---- summing ---+
pins device

Now, one thing this results in, is in terms of control
registers, there are two registers. There is one register
for the VID, plus a second register (somewhere) controlling
the boost via some general purpose I/O pins.

The hardware monitor (measurement circuit), monitors the
resulting summation of VID plus boost.

Is that good enough ? Not really. The thing is, as the CPU
load goes up, the Vcore actual voltage drops a bit. On
previous generations, the slope of the load line, allowed
about -0.150 volts when the CPU was running 100%. So if
you were running Prime95, the Vcore measured value could be
shifted by -0.150.

If you check the measured value, when the system is idle,
then that will get you closer to the "true value" or sum
of the VID + boost. On an Asus motherboard, it is
traditional for this to be off by about +0.060 volts (overshoot).
So if you saw 1.41V measured at CPU idle, you might suspect
the "true" value to be 1.35V. Asus tends to allow their Vcore
to overvolt a little bit at idle.

I'll try a few experiments on my (Asus) motherboard, and
post back what I see. I can't do that without rebooting.

Paul

Thanks, Paul.

I did notice exactly that droop during Intel Burn Test. It was intermittent.

This board lets you set Vcore to specific voltages, from tiny to
massive. I don't see anything else there.

With EIST and C1E disabled, shouldn't the voltage increase endure into
Windows? Or is there something in Windows that could reduce the voltage?

--
Ed Light

Better World News TV Channel:
http://realnews.com

Iraq Veterans Against the War and Related:
http://ivaw.org
http://couragetoresist.org
http://antiwar.com

Send spam to the FTC at

Thanks, robots.

Ed Light wrote:
> Thanks, Paul.
>
> I did notice exactly that droop during Intel Burn Test. It was
> intermittent.
>
> This board lets you set Vcore to specific voltages, from tiny to
> massive. I don't see anything else there.
>
> With EIST and C1E disabled, shouldn't the voltage increase endure into
> Windows? Or is there something in Windows that could reduce the voltage?
>

OK, I tried a few experiments here.

First, I used RMClock, to get the particulars for my processor.

E8400 333*9=3000MHz, FSB1333 (333*4), low multiplier gives 2000MHz.

6X 1.1000 V
9X 1.2625 V

CPU-World says that processor can span 0.85-1.3625, but it looks
like Intel has it limited as shown in RMClock.

So first, I tried with no boost. The assumption is (since I'm running
with EIST and C1E disabled anyway), that the VID register is already
pegged at 1.2625 volts. And the Vcore chart in RMClock has that
value printed on it as well.

Applied Measured Speedfan
CPUZ

1.2625V 1.224V Idle 1.22 etc
1.216V Load

I won't show any more Speedfan, because the rounding to two digits is
kinda a waste of time. Note that there is hardly any change under
load, and I've only got a 65W dual core installed. The measured voltage
value is "on the low side" by 0.0385V .

Now, I apply some boost. Any value set in the BIOS, higher than 1.2625
for this particular processor, would require boost applied somehow.
So a second register must have a boost value entered in it.
Some of the Asus boost circuits in the past have been pretty sloppy,
but this one seems pretty good.

Applied Measured
CPUZ

1.3250V 1.280V Idle
1.272V Load

In this case, it is on the low side by 0.045V.

Applied Measured
CPUZ

1.3625V 1.320V Idle
1.312V Load

In this case, the measured value is low by 0.0425V.

And in each case, RMClock still thinks the register is
set to 1.2625V. So the extra voltage is not a result
of the CPU register setting for the VID pins, but is
determined by whatever is used to add a boost.

RMClock says 1.10000V is the lowest register value I
can use. CPU-World claims the E8400 ranges from 0.85V
and upwards. The Asus BIOS won't allow that value to
be dialed below 1.10000V . Which means there is no
"undervolting" capability on my board, only boost.
And the boost is used, whenever voltages above the
range limit in the CPU, are asked for.

Since CPUZ seems to be reading the measured voltage,
if you turn on boost in your BIOS, you should be
seeing the effects. In fact, you should be able to
measure them in the hardware monitor BIOS page, after
doing a Save and Exit with the new Vcore voltage dialed
into the BIOS. So if the value stubbornly won't change,
when you know it is above the max value listed in
RMClock, then your boost is either non-existent,
the BIOS is wrong for the board (i.e. it's loading
the wrong GPIO bits to set the boost voltage),
or something along those lines. The BIOS setup screen
doesn't have to match reality, as I've had boards
where the memory timings shown in the BIOS, are not
actually being loaded into the hardware. So the
BIOS setup is not necessarily "correct" in all cases.
It is one of the reasons, I rely on CPUZ to verify
my RAM timings, rather than believe what I'm seeing
in the BIOS.

You can get RMClock 2.35 here. That will allow you to
verify the "range limited" value of your CPU VID register.

http://cpu.rightmark.org/download.shtml

Does EasyTune allow the voltage to be modified while
the system is running ? Or does it only monitor ?

Maybe you can see if EasyTune knows how to change it.
And that indeed might require a driver. You need
the equivalent of "GiveIO" or the other mechanism
is the BIOS passes an ACPI object to the OS, and the
driver uses that to make the changes. On an Asus
motherboard, that would be something like a System
Device called "ATK0110 ACPI Utility". That is a pseudo-device
passed by the BIOS to the OS. I don't know what
the equivalent would be for a Gigabyte motherboard.
You could try Device Manager, or Lavalys Everest, and
see if there is a strange "System Device" present.
Right now, mine seems to be using a Windows driver
of some sort. I don't think I have any utility here
loaded right now, to access ATK0110 and test it.
(It might be AI Booster, but I'm not installing that
kind of crap on my system :-) When I overclock, I use
the BIOS, as bad as it is.)

Paul

Paul,

I don't understand it as well as you.

But I downloaded RM Clock.

The Intel pdf for my E5200 gives a VID Voltage Range of 0.85V – 1.3625V.

RM Clock says 1.1 to 1.275.

The BIOS says that it belongs at 1.275. But it sets it at 1.25,
according to Gigabyte Easy Tune, and slightly lower according to cpu-z.

I can select a Turbo mode in the BIOS, which gives about 1.35 - but - it
overvolts my memory past spec by .1 to 2.1, and I can't turn it down
manually (only up).
--
Ed Light

Better World News TV Channel:
http://realnews.com

Iraq Veterans Against the War and Related:
http://ivaw.org
http://couragetoresist.org
http://antiwar.com

Send spam to the FTC at

Thanks, robots.

Ed Light wrote:
> Paul,
>
> I don't understand it as well as you.
>
> But I downloaded RM Clock.
>
> The Intel pdf for my E5200 gives a VID Voltage Range of 0.85V – 1.3625V.
>
> RM Clock says 1.1 to 1.275.
>
> The BIOS says that it belongs at 1.275. But it sets it at 1.25,
> according to Gigabyte Easy Tune, and slightly lower according to cpu-z.

In the RMClock graphs, does the graph show 1.275 or 1.25 ? Maybe
RMClock can verify the setting for you (the current value shown
in the graph).

CPUZ is showing the measured value, and in my tests, the measured
value seemed to be ~0.040V lower than the set value. So at least
my new Asus board, doesn't "overshoot" in the traditional Asus way,
when my new board is idle.

>
> I can select a Turbo mode in the BIOS, which gives about 1.35 - but - it
> overvolts my memory past spec by .1 to 2.1, and I can't turn it down
> manually (only up).

Some of the named modes such as Turbo, do all sorts of weird stuff.
In past years for example, Turbo set memory CAS to CAS2. Multiple
settings may be modified by such choices (because they're "doing
the overclocking for you").

In terms of researching how much voltage a memory chip can take,
you can get an "Absolute Max" from the datasheet. In this example,
Table 5 on PDF page 23, says the absolute max is 2.3 volts for
this Micron chip. Not all memory chips, have useful information
printed on them, so this method of inquiry doesn't always work out.
Removing a "heat spreader" from a DIMM, can void the warranty.
And one reason for sticking the heat spreader on there in the
first place, is just to cover up the chips. The technology that
really needed heat spreaders for heating reasons, was RDRAM/RIMMs.

http://download.micron.com/pdf/datas...r2/1GbDDR2.pdf

Paul

On 5/21/2010 11:51 PM, Paul wrote:

> In the RMClock graphs, does the graph show 1.275 or 1.25 ?

1.275

> In terms of researching how much voltage a memory chip can take,
> you can get an "Absolute Max" from the datasheet.

Not in my module's pdf. I'll have to contact support.

Thanks!

--
Ed Light

Better World News TV Channel:
http://realnews.com

Iraq Veterans Against the War and Related:
http://ivaw.org
http://couragetoresist.org
http://antiwar.com

Send spam to the FTC at

Thanks, robots.

Gigabyte says that there's nothing wrong with the board, and I have to
disable EES in Windows. Couldn't find where to do that -- wrote them
back asking how.
--
Ed Light

Better World News TV Channel:
http://realnews.com

Iraq Veterans Against the War and Related:
http://ivaw.org
http://couragetoresist.org
http://antiwar.com

Send spam to the FTC at

Thanks, robots.

Ed Light wrote:
> Gigabyte says that there's nothing wrong with the board, and I have to
> disable EES in Windows. Couldn't find where to do that -- wrote them
> back asking how.

Looking in the downloadable PDF user manual, I found a reference to

"Easy Energy Saver"

which is part of their installed software ("chopware"). It is
part of EasyTune 5 Pro (Section 4-3 in the manual). That is what
they want you to disable.

*******

You should have been able to use non-Turbo mode, set the voltage
in the BIOS, do a Save and Exit, re-enter the BIOS, go to the
"PC Health Status" screen in the BIOS (section 2-8 in the manual),
and verify your Vcore is properly overvolting. That would avoid
any interference from chopware. Remembering that the measured
value, at idle, is a bit less than the "applied" value, by perhaps
0.040V or so.

Paul

On 5/25/2010 1:07 PM, Paul wrote:

> You should have been able to use non-Turbo mode, set the voltage
> in the BIOS, do a Save and Exit, re-enter the BIOS, go to the
> "PC Health Status" screen in the BIOS (section 2-8 in the manual),
> and verify your Vcore is properly overvolting. That would avoid
> any interference from chopware. Remembering that the measured
> value, at idle, is a bit less than the "applied" value, by perhaps
> 0.040V or so.

Yes, I can do that. But in Windows it always goes back down.

Well, they have asked me to RMA it, so I may do that in July, when I'll
be out of town, not using it.

They say they couldn't duplicate the problem; I'm hoping they really
booted into Windows with it.

They also asked me to try a different cpu, but I haven't got one. The
E5200 does lack some atributes that the Core 2 Duos have.

--
Ed Light

Better World News TV Channel:
http://realnews.com

Iraq Veterans Against the War and Related:
http://ivaw.org
http://couragetoresist.org
http://antiwar.com

Send spam to the FTC at

Thanks, robots.