Increasing vcore voltage on semperon 3000+

I have a semperon 64bit 3000+ running mostly with vista home prem. 64x which
has been a bit unstable lately. (pc crashing with re-starts etc.)

On running sisoft sandra evaluation programme it says that the vcore should
be 1.4(1.44) but in fact is running too low at 1.36(1.38)

In the bios I find no setting to increase the vcore voltage. (As far as i
have been able to find out I have the newest bios that will work with this
mainboard.)

Mboard is a PCChips M860 v1

BIOS AMI v.080011smbios/dmi v.2.30 29.June.2005

I have tried changing the vcore voltage with the utility: crystal cpuid -
but when I go to the feature in the appl. to do this a fault window comes up
saying "stop multiplyer managament." O don't know what this actually means
or how to do it. Either way the vcore settings does not get changed.

I read that a low vcore voltage is a common cause for system instability but
have so far found no way to change this.

Not at all keen to get into "attacking" the hardware to alter this.

The sisoft sandra and crystal cpuid both say that the change vcore feature
is active/available.

Gratefull for any tips!

ron b.

"Ron Bartle" schrieb im Newsbeitrag
...
I have a semperon 64bit 3000+ running mostly with vista home prem. 64x
which has been a bit unstable lately. (pc crashing with re-starts etc.)

On running sisoft sandra evaluation programme it says that the vcore
should be 1.4(1.44) but in fact is running too low at 1.36(1.38)

In the bios I find no setting to increase the vcore voltage. (As far as i
have been able to find out I have the newest bios that will work with this
mainboard.)

Mboard is a PCChips M860 v1

BIOS AMI v.080011smbios/dmi v.2.30 29.June.2005


// Edit from orig. Poster: According to information I now have apparently
this budget(!) Motherboard/BIOS simply does not support changes to the
vcore. //


I have tried changing the vcore voltage with the utility: crystal cpuid -
but when I go to the feature in the appl. to do this a fault window comes
up saying "stop multiplyer managament." O don't know what this actually
means or how to do it. Either way the vcore settings does not get
changed.

I read that a low vcore voltage is a common cause for system instability
but have so far found no way to change this.

Not at all keen to get into "attacking" the hardware to alter this.

The sisoft sandra and crystal cpuid both say that the change vcore feature
is active/available.

Gratefull for any tips!

ron b.

Ron Bartle wrote:

"Ron Bartle" schrieb im
Newsbeitrag ...
I have a semperon 64bit 3000+ running mostly with vista home prem. 64x
which has been a bit unstable lately. (pc crashing with re-starts etc.)

On running sisoft sandra evaluation programme it says that the vcore
should be 1.4(1.44) but in fact is running too low at 1.36(1.38)

In the bios I find no setting to increase the vcore voltage. (As far
as i have been able to find out I have the newest bios that will work
with this mainboard.)

Mboard is a PCChips M860 v1

BIOS AMI v.080011smbios/dmi v.2.30 29.June.2005


// Edit from orig. Poster: According to information I now have
apparently this budget(!) Motherboard/BIOS simply does not support
changes to the vcore. //


I have tried changing the vcore voltage with the utility: crystal
cpuid - but when I go to the feature in the appl. to do this a fault
window comes up saying "stop multiplyer managament." O don't know what
this actually means or how to do it. Either way the vcore settings
does not get changed.

I read that a low vcore voltage is a common cause for system
instability but have so far found no way to change this.

Not at all keen to get into "attacking" the hardware to alter this.

The sisoft sandra and crystal cpuid both say that the change vcore
feature is active/available.

Gratefull for any tips!

ron b.


Well, to put your mind at rest, a slightly low measured value for
Vcore, does not guarantee trouble. Processors have what is known as
a "load line". That is a voltage versus current curve. When the processor
is running at 100% loading, the processor draws the maximum current.
On some processors, the expected voltage level on Vcore is allowed
to drop 0.150V compared to the nominal VID value. And in fact, if you
look at typical processor design datasheets, the voltage is never
supposed to be exactly equal to the VID value - it is alwsys
expected to be a little less. (There are exceptions to every rule,
and different motherboard makers have their own "trade mark" way
of dealing with the load line. For example, some allow less
droop then others, even though that may violate the processor
makers load line specification. They do that, because enthusiasts
believe that a "low droop" is a sign of quality, when in fact it
may violate the upper and lower bounds drawn on the processor
manufacturer's load line. Some manufacturers overvolt a tiny
bit, when the processor is idle.)

Try running the processor with a full computing load.

Download Prime95 from mersenne.org and run the "Torture Test" option.
The test allocates a large block of memory, and runs FFTs
in it. It is a sensitive test of both memory integrity,
correct processor operation, and also would exercise the Northbridge
(on an Intel board). It has the added plus, of running the
CPU at 100% load, making the CPU as hot as possible, and
depressing the Vcore as much as possible as well. (There are
other programs for doing this, Orthos is one, and one of the
other enthusiast sites has their own program, that is even better
at the task of loading the processor. So there are other options.)

In any case, run Prime95 and select the "Torture Test".

Does the computer immediately crash when you run Prime95 ?
Then you might suspect Vcore.

Does it take a while and eventually crash ?

Does it crash only if you make Prime95 stop ?

Do you have a temperature monitoring software, such as Speedfan
from almico.com ? What happens to the monitored temperatures on
the board ?

Are all fans operating ?

If problems do not correlate with processor loading (i.e. crashes
happen most when the computer is idle), then perhaps there is
something else going on.

If Prime95 immediately halts and reports an error, but the
computer does not crash, then something is still not stable.

With regard to setting the VID, check your "Cool N' Quiet" setting.
I would have though that the existence of "Cool N' Quiet" would
imply the passage of an ACPI object from the BIOS to the OS. The
OS should then have the ability to change P-States. The "Crystal"
program should be able to do something in that case - that would
be my guess.

In the AMD 26094.pdf document, they mention that it is the job of
the BIOS, to decide whether the motherboard supports VID changes.
I hadn't really thought about it too much, but I suppose it is
possible for a board to be so cheap, that the Vcore regulator
doesn't meet overshoot/undershoot specs on VID changes. The implication
of that would be, the motherboard manufacturer would be well advised
in that case, to code the BIOS so that the info it passes to the
OS says "please do not change the voltage". But somehow I really
doubt any motherboard can be that bad. I just don't see the
manufacturer saving any money by doing something like that.
(By the time you pay for a multi-phase regulator, you get all
the well behaved properties for free. And your motherboard does have
a three phase Vcore. If they were "shaving nickels of cost",
they could have tried building it with a two phase Vcore.
Your board looks no different than many others. And if they
really wanted to save money, they would have made it microATX
and removed a couple PCI slots, and uses a smaller PCB.)

Paul

Hi Paul,
thanks for your detailed and informative answer.
There really does not seem to be any way to affect the vcore voltage with
this motherboard.
Nowhere in the bios does one find any reference in this and in a review of
the board I found on the net there are indications that exactly this is one
of the down-side arguments against this (35US$-) M-Board.

Given the above I don't see much point in stressing the system to test its'
integrity when I have no way or altering much in any case.

Onother major point is that for whatever reason the pc has now been running
stable again for a couple of days.

Once again - thanks for your interesting response.

ron b.


"Paul" schrieb im Newsbeitrag
...
Ron Bartle wrote:

"Ron Bartle" schrieb im
Newsbeitrag ...
I have a semperon 64bit 3000+ running mostly with vista home prem. 64x
which has been a bit unstable lately. (pc crashing with re-starts etc.)

On running sisoft sandra evaluation programme it says that the vcore
should be 1.4(1.44) but in fact is running too low at 1.36(1.38)

In the bios I find no setting to increase the vcore voltage. (As far as
i have been able to find out I have the newest bios that will work with
this mainboard.)

Mboard is a PCChips M860 v1

BIOS AMI v.080011smbios/dmi v.2.30 29.June.2005


// Edit from orig. Poster: According to information I now have apparently
this budget(!) Motherboard/BIOS simply does not support changes to the
vcore. //


I have tried changing the vcore voltage with the utility: crystal
cpuid - but when I go to the feature in the appl. to do this a fault
window comes up saying "stop multiplyer managament." O don't know what
this actually means or how to do it. Either way the vcore settings does
not get changed.

I read that a low vcore voltage is a common cause for system instability
but have so far found no way to change this.

Not at all keen to get into "attacking" the hardware to alter this.

The sisoft sandra and crystal cpuid both say that the change vcore
feature is active/available.

Gratefull for any tips!

ron b.


Well, to put your mind at rest, a slightly low measured value for
Vcore, does not guarantee trouble. Processors have what is known as
a "load line". That is a voltage versus current curve. When the processor
is running at 100% loading, the processor draws the maximum current.
On some processors, the expected voltage level on Vcore is allowed
to drop 0.150V compared to the nominal VID value. And in fact, if you
look at typical processor design datasheets, the voltage is never
supposed to be exactly equal to the VID value - it is alwsys
expected to be a little less. (There are exceptions to every rule,
and different motherboard makers have their own "trade mark" way
of dealing with the load line. For example, some allow less
droop then others, even though that may violate the processor
makers load line specification. They do that, because enthusiasts
believe that a "low droop" is a sign of quality, when in fact it
may violate the upper and lower bounds drawn on the processor
manufacturer's load line. Some manufacturers overvolt a tiny
bit, when the processor is idle.)

Try running the processor with a full computing load.

Download Prime95 from mersenne.org and run the "Torture Test" option.
The test allocates a large block of memory, and runs FFTs
in it. It is a sensitive test of both memory integrity,
correct processor operation, and also would exercise the Northbridge
(on an Intel board). It has the added plus, of running the
CPU at 100% load, making the CPU as hot as possible, and
depressing the Vcore as much as possible as well. (There are
other programs for doing this, Orthos is one, and one of the
other enthusiast sites has their own program, that is even better
at the task of loading the processor. So there are other options.)

In any case, run Prime95 and select the "Torture Test".

Does the computer immediately crash when you run Prime95 ?
Then you might suspect Vcore.

Does it take a while and eventually crash ?

Does it crash only if you make Prime95 stop ?

Do you have a temperature monitoring software, such as Speedfan
from almico.com ? What happens to the monitored temperatures on
the board ?

Are all fans operating ?

If problems do not correlate with processor loading (i.e. crashes
happen most when the computer is idle), then perhaps there is
something else going on.

If Prime95 immediately halts and reports an error, but the
computer does not crash, then something is still not stable.

With regard to setting the VID, check your "Cool N' Quiet" setting.
I would have though that the existence of "Cool N' Quiet" would
imply the passage of an ACPI object from the BIOS to the OS. The
OS should then have the ability to change P-States. The "Crystal"
program should be able to do something in that case - that would
be my guess.

In the AMD 26094.pdf document, they mention that it is the job of
the BIOS, to decide whether the motherboard supports VID changes.
I hadn't really thought about it too much, but I suppose it is
possible for a board to be so cheap, that the Vcore regulator
doesn't meet overshoot/undershoot specs on VID changes. The implication
of that would be, the motherboard manufacturer would be well advised
in that case, to code the BIOS so that the info it passes to the
OS says "please do not change the voltage". But somehow I really
doubt any motherboard can be that bad. I just don't see the
manufacturer saving any money by doing something like that.
(By the time you pay for a multi-phase regulator, you get all
the well behaved properties for free. And your motherboard does have
a three phase Vcore. If they were "shaving nickels of cost",
they could have tried building it with a two phase Vcore.
Your board looks no different than many others. And if they
really wanted to save money, they would have made it microATX
and removed a couple PCI slots, and uses a smaller PCB.)

Paul

Ron Bartle wrote:
Hi Paul,
thanks for your detailed and informative answer.
There really does not seem to be any way to affect the vcore voltage
with this motherboard.
Nowhere in the bios does one find any reference in this and in a review
of the board I found on the net there are indications that exactly this
is one of the down-side arguments against this (35US$-) M-Board.

Given the above I don't see much point in stressing the system to test
its' integrity when I have no way or altering much in any case.

Onother major point is that for whatever reason the pc has now been
running stable again for a couple of days.

Once again - thanks for your interesting response.

ron b.


I was curious as to what Vcore regulator chip they are using, and
while there are a few really high resolution photos of the board,
none of them cover the Vcore chip. In the Newegg picture here,
as near as I can tell, the Vcore chip is to the right of the
6 capacitors, and down a little bit.

http://images10.newegg.com/NeweggIma...185-050-02.jpg

Some motherboard makers, hide their Vcore chip underneath the plastic
retention bracket that surrounds the CPU socket. I thought I had
located the chip, up near the I/O connectors, but that turned out
to be a GD75232, which is an RS-232 interface chip. So that leaves
the chip which is to the right of the CPU socket as the most likely
candidate. But I cannot find a closeup, so I can read the numbers and
letters off the top. One thing that does not translate well, is the
product logo. For example, on the GD75232, the symbol is recognizable
as the Texas Instruments brand (their logo is a map of Texas).

The reason I was curious about what chip was used, was to follow up
on the theory that their chip choice is the limitation.

Another explanation, is they were too dumb to put in proper BIOS
support for FID/VID, but they must have done that for some of
their other products, so the oversight still seems mighty
strange.

And everybody likes a stable board, so don't say another word :-)
You'll put the jinx on it.

Paul

On Thu, 03 May 2007 10:25:50 +0200, Ron Bartle wrote:

There really does not seem to be any way to affect the vcore voltage with
this motherboard.

If it's a 754 board, this can do it.

http://home.wi.rr.com/zippyzap/computer/754vcore/

--
Want the ultimate in free OTA SD/HDTV Recorder? http://mythtv.org
http://mysettopbox.tv/knoppmyth.html Usenet alt.video.ptv.mythtv
My server http://wesnewell.no-ip.com/cpu.php
HD Tivo S3 compared http://wesnewell.no-ip.com/mythtivo.htm