BIOS does not know when to select onboard/offboard graphics

[reikred]

I just had occasion to experiment with a new Gigabyte GA-MA78GM-S2HP
(AMD 780G/ AMD SB700) motherboard.

To my surprise the board video would not come up with a VGA plugged
into the onboard/internal video, and I had to insert an offboard/
external video card (ge7200gs) in order to get into BIOS and switch
the setting so that the onboard video was used.

This appears really strange to me. What is going on here? Can the
board not tell that a VGA plug is connected to it, and activate the
interface accordingly? I do not have a DVI-based monitor, but would
that make any difference?

I should mention that I have a KVM switch between the computer/vga and
the monitor. It is an older (ca. 2000?) Belkin Omnicube 4-port, and
the monitor is also a vintage 2003 LCD screen.

I'm starting to wonder whether there is something going on related to
the onboard VGA not being able to detect the LCD or the KVM because it
is an older model. In the Wikipedia article one VGA there is
discussion of classic VGA versus newer DDC2 that includes an i2c bus.
Is it possible that the onboard vga checks for i2c, finds nothing, an
therefore cannot tell whether a monitor is attached or not ????

 

[Paul]

I just had occasion to experiment with a new Gigabyte GA-MA78GM-S2HP
(AMD 780G/ AMD SB700) motherboard.

To my surprise the board video would not come up with a VGA plugged
into the onboard/internal video, and I had to insert an offboard/
external video card (ge7200gs) in order to get into BIOS and switch
the setting so that the onboard video was used.

This appears really strange to me. What is going on here? Can the
board not tell that a VGA plug is connected to it, and activate the
interface accordingly? I do not have a DVI-based monitor, but would
that make any difference?

I should mention that I have a KVM switch between the computer/vga and
the monitor. It is an older (ca. 2000?) Belkin Omnicube 4-port, and
the monitor is also a vintage 2003 LCD screen.

I'm starting to wonder whether there is something going on related to
the onboard VGA not being able to detect the LCD or the KVM because it
is an older model. In the Wikipedia article one VGA there is
discussion of classic VGA versus newer DDC2 that includes an i2c bus.
Is it possible that the onboard vga checks for i2c, finds nothing, an
therefore cannot tell whether a monitor is attached or not ????

I thought one technique that was used, was impedance sensing.
I've done limited testing of that here, by making myself a
VGA plug with a 75 ohm resistor to ground, on the R, G, and B
signals. My computer seems to sense such a "dongle" as a
monitor. The fact that my dongle does not have I2C and deliver
an EDID, doesn't seem to matter as far as turning on the
interface is concerned. Since that plug is a "secondary"
monitor, the driver right now is not in a rush to enable
it. But with the dongle in place, if I try to turn on both
monitor outputs, the setting "sticks" and works. With the
dongle unplugged, and attempt to turn on the second interface
fails.

The VGA pinout is here.

http://en.wikipedia.org/wiki/VGA_connector

The 75 ohms is between pin 1 and pin 6 (red),
between pin 2 and pin 7 (green), and
between pin 3 and pin 8 (blue). I don't know
if it is valid to capacitively couple the
interface or not. If your monitor interface
used coupling capacitors on the lines, then
an ohmmeter test may not be conclusive, due
to the charging of the capacitor.

On some TV sets (S-video interface), the impedance
sensing technique can fail, which is why the
display control panel has the notion of "force"
for that interface. I'm not aware of a "force"
option for VGA - implying it is supposed to
read an honest 75 ohms on each color signal, and be
detectable reliably via that resistance.

My dongle has nothing connected to H and V,
the sync signals.

Have you tried the monitor directly on the
motherboard, without the KVM there ? Does
that work ?

Paul

 

[reikred]

I thought one technique that was used, was impedance sensing.
I've done limited testing of that here, by making myself a
VGA plug with a 75 ohm resistor to ground, on the R, G, and B
signals. My computer seems to sense such a "dongle" as a
monitor. The fact that my dongle does not have I2C and deliver
an EDID, doesn't seem to matter as far as turning on the
interface is concerned. Since that plug is a "secondary"
monitor, the driver right now is not in a rush to enable
it. But with the dongle in place, if I try to turn on both
monitor outputs, the setting "sticks" and works. With the
dongle unplugged, and attempt to turn on the second interface
fails.

The VGA pinout is here.

http://en.wikipedia.org/wiki/VGA_connector

The 75 ohms is between pin 1 and pin 6 (red),
between pin 2 and pin 7 (green), and
between pin 3 and pin 8 (blue). I don't know
if it is valid to capacitively couple the
interface or not. If your monitor interface
used coupling capacitors on the lines, then
an ohmmeter test may not be conclusive, due
to the charging of the capacitor.

On some TV sets (S-video interface), the impedance
sensing technique can fail, which is why the
display control panel has the notion of "force"
for that interface. I'm not aware of a "force"
option for VGA - implying it is supposed to
read an honest 75 ohms on each color signal, and be
detectable reliably via that resistance.

My dongle has nothing connected to H and V,
the sync signals.

Have you tried the monitor directly on the
motherboard, without the KVM there ? Does
that work ?

    Paul

Good stuff, Paul (as usual!). One of my first experiments will be to
plug the monitor directly into the computer and see what happens.

 

[reikred]

Good stuff, Paul (as usual!). One of my first experiments will be to
plug the monitor directly into the computer and see what happens.

Paul, first of all, I found online a manual for my LCD monitor and it
indicates that it has i2c, so it must be a DDC2-enabled monitor.
As for the KVM, I could not find out one way or the other.

Next, I tried plugging the monitor directly into the computer (i.e.
bypass the KVM). I tried with the monitor plugged into either onboard
or offboard VGA. The system behaved exactly as before. That is, I had
to select in BIOS (Award) InitDisplayFirst = OnChipVGA to get a video
signal on the onboard VGA connector. The other choices were {PEG, PCI-
slot} where PEG was shorthand for PCIE Graphics.

Either onboard or offboard worked as long as the setting was right,
but the Mobo still was not able to autodetect and autoselect the
correct adapter.

Upshot: No solution yet. What I really don't understand is why a swarm
of people are not having the same problem. Lots of people still must
be using VGA rather than DVI or HDMI, if nothing else because KVM -
switches still seem to be mostly VGA-based.

 

[Paul]

Paul, first of all, I found online a manual for my LCD monitor and it
indicates that it has i2c, so it must be a DDC2-enabled monitor.
As for the KVM, I could not find out one way or the other.

Next, I tried plugging the monitor directly into the computer (i.e.
bypass the KVM). I tried with the monitor plugged into either onboard
or offboard VGA. The system behaved exactly as before. That is, I had
to select in BIOS (Award) InitDisplayFirst = OnChipVGA to get a video
signal on the onboard VGA connector. The other choices were {PEG, PCI-
slot} where PEG was shorthand for PCIE Graphics.

Either onboard or offboard worked as long as the setting was right,
but the Mobo still was not able to autodetect and autoselect the
correct adapter.

Upshot: No solution yet. What I really don't understand is why a swarm
of people are not having the same problem. Lots of people still must
be using VGA rather than DVI or HDMI, if nothing else because KVM -
switches still seem to be mostly VGA-based.

If no separate video card is present, it should default to the built-in
graphics. The BIOS knows, by scanning the PCI and PCI express busses,
whether a video card is present. When no card is present, then
the "first display" option should effectively only point to the
onboard. "First display" is for control of scenarios, where more
than one video device is present, and there is a need to identify
which one gets used for BIOS. (I think which is the primary monitor
could be changed in the OS, as far as display numbering goes once
the OS is running.)

You have a selector setting, to select between DVI or HDMI, but for
both of those settings, "DSUB" is also enabled. So the DSUB or VGA
connector, should always be a candidate. It could be more of an issue,
if you had an HDMI display, and expected it to work by default. The
manual doesn't give the impression that the BIOS is clever enough
to auto detect the port in that case. It could be that the DVI versus
HDMI feature is implemented outside the GPU.

You should be able to verify the EDID is coming across the interface,
with the monitor program from entechtaiwan.com .

http://www.entechtaiwan.com/util/moninfo.shtm

There are some comments about video here, but not identifying the
exact same problem you've got. With integrated video, it seems
the testing standards are a bit more lax than for video cards,
because you will see reports of video failures. For example, some
of the Nvidia 6100 boards have had bad video (defective display
output) right out of the box. So the Northbridge chips don't see
to be tested as well as a video card's chip. Either that, or they're
being damaged during manufacturing.

http://www.newegg.com/Product/ProductReview.aspx?Item=N82E16813128360

I'd suggest flash updating the BIOS, but the release notes don't indicate
any improvements in the BIOS, from one release to the next. Sort of
like how Asus does it now ("support new CPU" is the release note for
each release). Laziness in the industry abounds.

Paul

 

[~misfit~]

If no separate video card is present, it should default to the
built-in graphics. The BIOS knows, by scanning the PCI and PCI
express busses, whether a video card is present. When no card is
present, then the "first display" option should effectively only point to
the
onboard. "First display" is for control of scenarios, where more
than one video device is present, and there is a need to identify
which one gets used for BIOS. (I think which is the primary monitor
could be changed in the OS, as far as display numbering goes once
the OS is running.)

You have a selector setting, to select between DVI or HDMI, but for
both of those settings, "DSUB" is also enabled. So the DSUB or VGA
connector, should always be a candidate. It could be more of an issue,
if you had an HDMI display, and expected it to work by default. The
manual doesn't give the impression that the BIOS is clever enough
to auto detect the port in that case. It could be that the DVI versus
HDMI feature is implemented outside the GPU.

You should be able to verify the EDID is coming across the interface,
with the monitor program from entechtaiwan.com .

http://www.entechtaiwan.com/util/moninfo.shtm

There are some comments about video here, but not identifying the
exact same problem you've got. With integrated video, it seems
the testing standards are a bit more lax than for video cards,
because you will see reports of video failures. For example, some
of the Nvidia 6100 boards have had bad video (defective display
output) right out of the box. So the Northbridge chips don't see
to be tested as well as a video card's chip. Either that, or they're
being damaged during manufacturing.

http://www.newegg.com/Product/ProductReview.aspx?Item=N82E16813128360

I'd suggest flash updating the BIOS, but the release notes don't
indicate any improvements in the BIOS, from one release to the next.
Sort of like how Asus does it now ("support new CPU" is the release note
for
each release). Laziness in the industry abounds.

Hi Paul,

You said in a more complete way (as usual <g>) what I would suggest; Try
updating the mobo BIOS.

Regards,

 

[reikred]

This is the root of the problem.  The bios should have had
the onboard video as the default.  Perhaps someone had
tested or used the board with a video card installed prior
to the attempt the OP made with integrated video, and left
the setting wrong, in which case clearing CMOS might have
solved the problem.

Otherwise I'd have to agree with you and misfit, eventually
a newer bios should be released that solves the problem, it
wouldn't be a matter of whether a monitor was detected at
this point.

It is amazing to me if this is really a bios bug. As an aside, I have
another mobo (asus m2npv-vm, chipset geforce6150) which exhibits
almost exactly the same symptoms in the same setup. Before that, I
have never had a mobo with integrated graphics, hence no other
experience.

I'll try and get a new BIOS soon (hate to disrupt a working system
right now), but the BIOS is already very new (2008-12-xx) and one
would think that AWARD would have their act together by now ??

Thanks for all the good suggestions.

 

[Paul]

It is amazing to me if this is really a bios bug. As an aside, I have
another mobo (asus m2npv-vm, chipset geforce6150) which exhibits
almost exactly the same symptoms in the same setup. Before that, I
have never had a mobo with integrated graphics, hence no other
experience.

I'll try and get a new BIOS soon (hate to disrupt a working system
right now), but the BIOS is already very new (2008-12-xx) and one
would think that AWARD would have their act together by now ??

Thanks for all the good suggestions.

To make further progress, it might take a second monitor, as well
as the first, so you can do some testing. The idea would be,
to make the known good monitor the primary graphics solution,
and then attempt to add the (failing) monitor as a second
display. Then, perhaps, if impedance sensing is not
working, attempts to add the (failing) monitor will fail,
indicating there is a problem with that monitor.

I'm not sure how much more progress you can make with just
one monitor, in terms of testing.

Another thing I'm not sure of (and have no way to test here),
is regarding what connectors are enabled when nothing is
connected. Video connectors have not always had impedance
sensing. There was a time, when all the connectors "just
worked". But someone had the bright idea, that radio
interference (emissions) could be reduced, if unused
video connectors were disabled. The question is, are all
video connectors disabled when nothing is connected ? Or
is one connector enabled anyway ? I don't know the answer
to that. Basically, the test cases I can run, are connecting
a working monitor first, and then doing experiments with
the second video connector. If my dongle is not connected,
I cannot span the display across two connectors. When my
three 75 ohm resistors on the dongle are plugged in,
then I'm allowed to span the display to that second
(bogus) dongle.

Paul

 

[reikred]

To make further progress, it might take a second monitor, as well
as the first, so you can do some testing. The idea would be,
to make the known good monitor the primary graphics solution,
and then attempt to add the (failing) monitor as a second
display. Then, perhaps, if impedance sensing is not
working, attempts to add the (failing) monitor will fail,
indicating there is a problem with that monitor.

I'm not sure how much more progress you can make with just
one monitor, in terms of testing.

Another thing I'm not sure of (and have no way to test here),
is regarding what connectors are enabled when nothing is
connected. Video connectors have not always had impedance
sensing. There was a time, when all the connectors "just
worked". But someone had the bright idea, that radio
interference (emissions) could be reduced, if unused
video connectors were disabled. The question is, are all
video connectors disabled when nothing is connected ? Or
is one connector enabled anyway ? I don't know the answer
to that. Basically, the test cases I can run, are connecting
a working monitor first, and then doing experiments with
the second video connector. If my dongle is not connected,
I cannot span the display across two connectors. When my
three 75 ohm resistors on the dongle are plugged in,
then I'm allowed to span the display to that second
(bogus) dongle.

    Paul

Another tidbit: I hooked up my laptop external VGA through the KVM and
it had no trouble downloading all the monitor parameters, using the
moninfo program described above, and presumably through the i2c bus.
This should mean that at least the KVM is not the problem.

(I cannot run Windows on the motherboard, it is a linux machine)

 

[reikred]

Another tidbit: I hooked up my laptop external VGA through the KVM and
it had no trouble downloading all the monitor parameters, using the
moninfo program described above, and presumably through the i2c bus.
This should mean that at least the KVM is not the problem.

(I cannot run Windows on the motherboard, it is a linux machine)

I upgraded the BIOS to version F4, but bios still cannot figure out
where the monitor is connected. If I remove the external video card,
it will switch back to the onboard, though. I don't think it did that
before.

 

[reikred]

Then it seems to be working properly now, the bios is not
supposed to figure out where the monitor is connected.  It
is supposed to default to onboard video, then either keep
using onboard video until you change a bios setting to
making an add-on video card the primary/default video
device, or automatically make the add-on video card the
primary/default when it detects a card is installed.

If it required detection of a monitor this could be a
problem for anyone wishing to run a system without a monitor
attached, though I suppose it might be possible to have a
bios setting for that as well, or have it default to
whatever it finds first as a video device if no monitor were
detected... but more of a pain to have to connect a monitor
to change that setting if it defaulted to detecting a
monitor then the CMOS settings were lost for whatever reason
(like a low battery).

My experience is that with no monitor connected, the system never gets
past POST and into the boot process. If it did, the system woul be
ping-able shortly thereafter, but I do not think it has been. I could
be wrong, but this is what I recall seeing.