p5e3 deluxe, cpuz reports link width of x16, ram 1.5v, despite manualsettings..

  • Thread starter Thread starter markm75
  • Start date Start date
M

markm75

I have the p5e3 deluxe board, two independent ati 4870 pcie 2.0 x16
cards..

I also have memory set to 1.9v for the ocz3g16004gk 8GB of ram (native
1600, set to 1333)..

When i have just those 2 cards in the board.. cpuz reports that the
"link width" ix x4 while the max supported is x16..

What is meant by link width? Does this mean i'm not getting the full
bandwidth out of my cards and likewise 3dmarks etc? (3d mark is
around 15,000)?

CPUZ and likewise memset.. report the ram as 1.5volt.. is this an
actual report of the ram volts, if not, how can i verify it is
actually 1.9volt?

Thanks for any help here
 
markm75 said:
I have the p5e3 deluxe board, two independent ati 4870 pcie 2.0 x16
cards..

I also have memory set to 1.9v for the ocz3g16004gk 8GB of ram (native
1600, set to 1333)..

When i have just those 2 cards in the board.. cpuz reports that the
"link width" ix x4 while the max supported is x16..
Click to expand...

Are the video cards in the two blue slots ? (The black one is x4.)

They should both be x16 if placed in the blue slots.

A better tool for checking the video cards, might be GPUZ.
The menu at the bottom allows selecting a card. Current
version 0.2.8 (I have 0.2.7 on disk here).

http://www.techpowerup.com/gpuz/
What is meant by link width? Does this mean i'm not getting the full
bandwidth out of my cards and likewise 3dmarks etc? (3d mark is
around 15,000)?

CPUZ and likewise memset.. report the ram as 1.5volt.. is this an
actual report of the ram volts, if not, how can i verify it is
actually 1.9volt?

Thanks for any help here
Click to expand...

There is a picture of CPUZ for a P5E3 Premium here. Not the same
as your board, but illustrative. The SPD tab seems to be showing the
JEDEC standard for voltage, and I don't know if that is actually
stored in the SPD or not. I thought only the "non-standard"
standards (i.e. invented outside JEDEC) actually stored the
desired voltage setting in the SPD.

http://www.vr-zone.com/articles/Practical_Power:_SilverStone_Olympia_&_FSP_Everest/5702-7.html

With your board, the first issue would be, is Vdimm actually
measured ? Or is it only visible as a setting, like in the
BIOS ? It may not be connected to the hardware monitor, which
has a limited number of channels.

On a P5E3 non-deluxe, this is someone's idea of how to check
Vdimm. It is possible the center pin on that MOSFET, is at
the same potential as the tab soldered to the board. That
is how the leg can be cut off by the MOSFET maker, and still
work. You can see the tiny, square regulator chip, which the
author of this picture was not able to identify. And the
code quoted in the picture "CZ-9F", is not typically useful
for lookup (when surface mount components get small
enough, a custom code is used to ID the component -
you have to know who makes it, and download a doc, to
look it up in a table). A similar looking circuit is
present on your board, near the CPU fan header. The
layout is not done exactly the same way, so you'd have
to probe more than one MOSFET to find it.

http://img171.imageshack.us/img171/5707/vdimmcn9.jpg

The one time I successfully measured Vdimm on a board,
I ended up probing a pin on the DIMM slot. That was easier
than trying to guess which MOSFET was doing it.

Paul
 
Are the video cards in the two blue slots ? (The black one is x4.)

They should both be x16 if placed in the blue slots.

A better tool for checking the video cards, might be GPUZ.
The menu at the bottom allows selecting a card. Current
version 0.2.8 (I have 0.2.7 on disk here).

http://www.techpowerup.com/gpuz/






There is a picture of CPUZ for a P5E3 Premium here. Not the same
as your board, but illustrative. The SPD tab seems to be showing the
JEDEC standard for voltage, and I don't know if that is actually
stored in the SPD or not. I thought only the "non-standard"
standards (i.e. invented outside JEDEC) actually stored the
desired voltage setting in the SPD.

http://www.vr-zone.com/articles/Practical_Power:_SilverStone_Olympia_...

With your board, the first issue would be, is Vdimm actually
measured ? Or is it only visible as a setting, like in the
BIOS ? It may not be connected to the hardware monitor, which
has a limited number of channels.

On a P5E3 non-deluxe, this is someone's idea of how to check
Vdimm. It is possible the center pin on that MOSFET, is at
the same potential as the tab soldered to the board. That
is how the leg can be cut off by the MOSFET maker, and still
work. You can see the tiny, square regulator chip, which the
author of this picture was not able to identify. And the
code quoted in the picture "CZ-9F", is not typically useful
for lookup (when surface mount components get small
enough, a custom code is used to ID the component -
you have to know who makes it, and download a doc, to
look it up in a table). A similar looking circuit is
present on your board, near the CPU fan header. The
layout is not done exactly the same way, so you'd have
to probe more than one MOSFET to find it.

http://img171.imageshack.us/img171/5707/vdimmcn9.jpg

The one time I successfully measured Vdimm on a board,
I ended up probing a pin on the DIMM slot. That was easier
than trying to guess which MOSFET was doing it.

    Paul
Click to expand...

Thanks for that info.. i figured the voltage was just a reference in
cpuz..

as far as the pciex16 thing.. they are both in x16 slots (blue)..

I did come across one person who mentioned they had their bios pcie
rate at 100, instead of auto.. they said if they set it to 105, then
the pciex4 became x16; once i can get a vista system going again, i
will try this change and test it, as well as with the gpuz tool..

Thanks again
 
Thanks for that info.. i figured the voltage was just a reference in
cpuz..

as far as the pciex16 thing.. they are both in x16 slots (blue)..

I did come across one person who mentioned they had their bios pcie
rate at 100, instead of auto.. they said if they set it to 105, then
the pciex4 became x16;   once i can get a vista system going again, i
will try this change and test it, as well as with the gpuz tool..

Thanks again
Click to expand...

Changing the value to 105, didnt change the cpuz reading of x4..
 
Try GPUZ. See what it says.

    Paul
Click to expand...

What value should i be looking for in GPUz.. per say... (btw.. i put
the wrong title on this thread :( should say reports x4 instead of
x16)..
 
What value should i be looking for in GPUz.. per say... (btw.. i put
the wrong title on this thread :(  should say reports x4 instead of
x16)..
Click to expand...

In GPUz.. for bus interface i see.. "PCI-E 2.0 x16@ x4 2.0"
 
In GPUz.. for bus interface i see.. "PCI-E 2.0 x16@ x4 2.0"
Click to expand...

I should add, if i change the card selection at the bottom, it changes
to PCI-E 2.0 x 16@ x16 2.0

Both blue slots should be x16, i dont understand this..
 
markm75 said:
I should add, if i change the card selection at the bottom, it changes
to PCI-E 2.0 x 16@ x16 2.0

Both blue slots should be x16, i dont understand this..
Click to expand...

So one slot is operating at x16 and the other at x4, would be
my guess. Now you've confirmed it with a second utility.

When a motherboard is manufactured, the motherboard may be washed
to remove soldering residues. Sometimes the excess paste has to
be removed, and in the past, dry cleaning type solvents would have
been used. Modern solder systems use aqueous wash, and the paste is water
soluble. If not enough washing is done, residue dries on connectors.
I've had an Ethernet connector that was dirty from the factory, and
required plugging and unplugging the cable a number of times, before
it made contact.

PCI Express detects as many contiguous lanes as possible, before
declaring the width. For example, you are allowed to plug a
x16 card into a socket which only has x4 lane wiring (such as
your black slot). The PCI Express negotiates, and in that case,
only the bottom four lanes are present, so that is what they
decide to use.

Say I have 16 lanes, and I bust lane 9. The bottom eight (leftmost)
are available, and so that is what gets used. The seven on the
right are "above" the missing lane, and are not used. In this
case, GPUZ would report "x16, at x8".

X X X X X X X X _ X X X X X X X

If I bust a lane further down, such as the second lane, the slot
gets reduced to running in x1 mode. The other fourteen perfectly
good lanes, are ignored. GPUZ would report "x16, at x1".

X _ X X X X X X X X X X X X X X

I cannot say I place much trust in the "dirt" theory in this
case, but figured I'd pass that on. It is also possible for
a driver I/O on the Northbridge to be damaged, a coupling
capacitor (see next to the PCI Express slot) to be missing,
or for a driver I/O on the video card GPU to be bad. But the
same principle applies, of the "lowest contiguous power_of_two
number of lanes" to be used at any one time.

Paul
 
So one slot is operating at x16 and the other at x4, would be
my guess. Now you've confirmed it with a second utility.

When a motherboard is manufactured, the motherboard may be washed
to remove soldering residues. Sometimes the excess paste has to
be removed, and in the past, dry cleaning type solvents would have
been used. Modern solder systems use aqueous wash, and the paste is water
soluble. If not enough washing is done, residue dries on connectors.
I've had an Ethernet connector that was dirty from the factory, and
required plugging and unplugging the cable a number of times, before
it made contact.

PCI Express detects as many contiguous lanes as possible, before
declaring the width. For example, you are allowed to plug a
x16 card into a socket which only has x4 lane wiring (such as
your black slot). The PCI Express negotiates, and in that case,
only the bottom four lanes are present, so that is what they
decide to use.

Say I have 16 lanes, and I bust lane 9. The bottom eight (leftmost)
are available, and so that is what gets used. The seven on the
right are "above" the missing lane, and are not used. In this
case, GPUZ would report "x16, at x8".

X X X X X X X X _ X X X X X X X

If I bust a lane further down, such as the second lane, the slot
gets reduced to running in x1 mode. The other fourteen perfectly
good lanes, are ignored. GPUZ would report "x16, at x1".

X _ X X X X X X X X X X X X X X

I cannot say I place much trust in the "dirt" theory in this
case, but figured I'd pass that on. It is also possible for
a driver I/O on the Northbridge to be damaged, a coupling
capacitor (see next to the PCI Express slot) to be missing,
or for a driver I/O on the video card GPU to be bad. But the
same principle applies, of the "lowest contiguous power_of_two
number of lanes" to be used at any one time.

    Paul
Click to expand...

This makes sense as well, ill take a look at that later.. ill also try
each card in each slot by itself to narrow things down..

I'm assuming my 850watt psu has enough power:
http://www.newegg.com/Product/Product.aspx?Item=N82E16817171017

In fact, one reviewer says he is using 2 ati's, but maybe he isnt
aware of the issue.
 
So one slot is operating at x16 and the other at x4, would be
my guess. Now you've confirmed it with a second utility.

When a motherboard is manufactured, the motherboard may be washed
to remove soldering residues. Sometimes the excess paste has to
be removed, and in the past, dry cleaning type solvents would have
been used. Modern solder systems use aqueous wash, and the paste is water
soluble. If not enough washing is done, residue dries on connectors.
I've had an Ethernet connector that was dirty from the factory, and
required plugging and unplugging the cable a number of times, before
it made contact.

PCI Express detects as many contiguous lanes as possible, before
declaring the width. For example, you are allowed to plug a
x16 card into a socket which only has x4 lane wiring (such as
your black slot). The PCI Express negotiates, and in that case,
only the bottom four lanes are present, so that is what they
decide to use.

Say I have 16 lanes, and I bust lane 9. The bottom eight (leftmost)
are available, and so that is what gets used. The seven on the
right are "above" the missing lane, and are not used. In this
case, GPUZ would report "x16, at x8".

X X X X X X X X _ X X X X X X X

If I bust a lane further down, such as the second lane, the slot
gets reduced to running in x1 mode. The other fourteen perfectly
good lanes, are ignored. GPUZ would report "x16, at x1".

X _ X X X X X X X X X X X X X X

I cannot say I place much trust in the "dirt" theory in this
case, but figured I'd pass that on. It is also possible for
a driver I/O on the Northbridge to be damaged, a coupling
capacitor (see next to the PCI Express slot) to be missing,
or for a driver I/O on the video card GPU to be bad. But the
same principle applies, of the "lowest contiguous power_of_two
number of lanes" to be used at any one time.

    Paul
Click to expand...


It did in fact turn out that this 2nd card had something wrong with
the connectors up a little ways on the card.. this was effectively
causing it to run in x4 mode..

I was able to run the other card in both slots at x16.. while the
other card showed up as x4 in both slots..

Thanks for the help and tips.

You were right on the money.. now i'm wondering if this card was
leading to instability all along (with my ddr3 1600mhz memory).. eek..
back to testing memory.
 
Back
Top