On Mon, Jul 21, 2008 at 6:07 PM, Sven Lankes wrote:

On Mon, Jul 21, 2008 at 11:53:01PM +0200, Luc Verhaegen wrote:

...

...

This looks like TMDSA load detection not working. Luc, can you look into this when you find the time?

...

You will find that, when you run rhd_conntest, the monitor only at certain times puts loads on the digital lines. We have seen the same problem with some other monitors...

This is a Dell 24" - pretty common I'd say.

...

So it is a problem with not the TMDSA load detection, but with the monitor attached to it, and a suitable strategy still needs to be proposed here.

Pardon my ignorance but how does the radeon-driver handle this?

radeon uses ddc to detect the presence of a monitor, and then (for DACs at least) falls back to load detection if ddc fails. Alex -- To unsubscribe, e-mail: radeonhd+unsubscribe@opensuse.org For additional commands, e-mail: radeonhd+help@opensuse.org

On Tue, Jul 22, 2008 at 3:55 AM, Luc Verhaegen wrote:

On Mon, Jul 21, 2008 at 06:20:17PM -0400, Alex Deucher wrote:

...

On Mon, Jul 21, 2008 at 6:07 PM, Sven Lankes wrote:

...

On Mon, Jul 21, 2008 at 11:53:01PM +0200, Luc Verhaegen wrote:

...

...

This looks like TMDSA load detection not working. Luc, can you look into this when you find the time?

...

You will find that, when you run rhd_conntest, the monitor only at certain times puts loads on the digital lines. We have seen the same problem with some other monitors...

This is a Dell 24" - pretty common I'd say.

...

So it is a problem with not the TMDSA load detection, but with the monitor attached to it, and a suitable strategy still needs to be proposed here.

Pardon my ignorance but how does the radeon-driver handle this?

radeon uses ddc to detect the presence of a monitor, and then (for DACs at least) falls back to load detection if ddc fails.

Alex

No, this is not the explanation. The real explanation is that radeon doesn't even try to implement load detection for TMDSA, even though the r5xx/r6xx hardware supports it nicely.

We don't use load detection for TMDS and if we did it would be a fall back if ddc failed. I agree that the hw supports it, (even pre-avivo chips supported it to a certain extent), but the problem, as this thread shows, is that it's not reliable. I'm not saying ddc is reliable either, but it seems to generally be more reliable than load detection in my experience. Alex -- To unsubscribe, e-mail: radeonhd+unsubscribe@opensuse.org For additional commands, e-mail: radeonhd+help@opensuse.org

On Tue, Jul 22, 2008 at 4:03 AM, Luc Verhaegen wrote:

On Mon, Jul 21, 2008 at 06:20:17PM -0400, Alex Deucher wrote:

...

On Mon, Jul 21, 2008 at 6:07 PM, Sven Lankes wrote:

...

On Mon, Jul 21, 2008 at 11:53:01PM +0200, Luc Verhaegen wrote:

...

...

This looks like TMDSA load detection not working. Luc, can you look into this when you find the time?

...

You will find that, when you run rhd_conntest, the monitor only at certain times puts loads on the digital lines. We have seen the same problem with some other monitors...

This is a Dell 24" - pretty common I'd say.

...

So it is a problem with not the TMDSA load detection, but with the monitor attached to it, and a suitable strategy still needs to be proposed here.

Pardon my ignorance but how does the radeon-driver handle this?

radeon uses ddc to detect the presence of a monitor, and then (for DACs at least) falls back to load detection if ddc fails.

Alex

How do you distinguish between the DAC and TMDS parts on DVI connectors then?

We check the digital bit in the edid if the connector is DVI-I. if it's digitial, it's TMDS, otherwise, it's DAC.

Also, hpd, with some quirks for the cards that, even though the hardware was there and is solid, didn't provide any HPD information through atombios (R5xx), tends to be rather solid.

Right, at least on r5xx cards, the hpd information in the bios is not always reliable, but the ddc i2c info is.

DDC has proven to be a gamble as well, as monitor manufacturers aren't as great as ATI is at making hardware.

I realize ddc isn't always reliable, but it seems to work more reliably than load detection in my experience. Alex -- To unsubscribe, e-mail: radeonhd+unsubscribe@opensuse.org For additional commands, e-mail: radeonhd+help@opensuse.org