Open ATI Driver To Receive PowerPlay Push?

[signor_rossi]

Hi signor_rossi, I think I remember you from Arch Linux forums. I am giving Arch a break right now since my internet connection is too slow for constant updating, and also due to non-working fglrx (my card is mobility X1600), I am not yet satisfied with the current state of power saving, my laptop is a both hotter and noisier than with my Debian Lenny's fglrx. And it would be the only thing needed for me to make the switch, I don't need any performance, if I need high-performance gaming I boot my Windows XP (or a gaming console).

Doesn't Debian Lenny get updates too once in a while?
I have had to stay for years with fglrx until I could switch to radeon, but you might be able to switch this year, who knows? Just guessing, mind you...

[Ghworg]

Debian stable (lenny is the current stable) only gets updates for security fixes and bugs. No new functionality is supposed to go in, that's why it's called stable.

[agd5f]

The UMS drivers have basic support for lowering the engine clock and PCIE lanes either statically, or dynamically during DPMS on/off. The current KMS patches are here:
http://people.freedesktop.org/~agd5f/pm/
All radeons should be supported.

[baghera]

The UMS drivers have basic support for lowering the engine clock and PCIE lanes either statically, or dynamically during DPMS on/off. The current KMS patches are here:
http://people.freedesktop.org/~agd5f/pm/
All radeons should be supported.

I was searching for them here and there, thank you for providing the link.
Does these set of patches apply to kernel 2.6.32 or 2.6.33?
Is there any chance to see them included upstream for kernel 2.6.33 or should we wait for 2.6.34 (or later)?

[Neuro]

I've got a Ati Mobility HD 3450.

The UMS drivers have basic support for lowering the engine clock and PCIE lanes either statically, or dynamically during DPMS on/off. The current KMS patches are here:
http://people.freedesktop.org/~agd5f/pm/
All radeons should be supported.

Right. UMS stands for userspace mode settings, I presume?
Will these patches be merged into mainline 2.6.33? Are they stable?
What does 'dynamically during DPMS on/off' mean? Does it mean that I need to turn off my monitor signal via DPMS to clock the card down or is there a separate control?

There's another issue I'd like to ask about, please bear with me on this one. I've got two cards in my notebook (Ati 3450 and Intel 4500MHD). The computer works in two modes:
- Discrete only - only the Ati card is available on the bus and everything works as usual
- Switchable mode - both cards are initially available on the bus. VGA display and internal LCD are accessible from the Intel card. Currently (no KMS in intel, fglrx for ATI) everything works ok. The framebuffer is initialized to the Intel card and I can disable the Ati card via a vendor-specific ACPI call (Lenovo U330) made from a custom-made kernel module, which is loaded before X starts. At the time X starts everything is fine and dandy: it can only see the Intel card and I can happily save A-LOT of working on the Intel card only (Ati+Intel on is 21W, Intel only is 13W).

Now I can ask my question: will using Ati KMS and Intel KMS somehow mess up my setting? I fear that at kernel load, the Intel KMS will try to initialize the display, but the Ati KMS would also be loaded. Will they conflict with each other? Will I be able to unload Ati modules later on in userspace and load my Ati-turn-off module?

Sorry if this is a bit off topic.

[agd5f]

I've got a Ati Mobility HD 3450.

Right. UMS stands for userspace mode settings, I presume?

Yes.

Will these patches be merged into mainline 2.6.33?

Not likely.

Are they stable?

They work, although some users report a flash on the screen when the clocks change.

What does 'dynamically during DPMS on/off' mean? Does it mean that I need to turn off my monitor signal via DPMS to clock the card down or is there a separate control?

It means the driver turns the clocks down when the displays go off and back up when they come back on. There are no manual controls.

There's another issue I'd like to ask about, please bear with me on this one. I've got two cards in my notebook (Ati 3450 and Intel 4500MHD). The computer works in two modes:
- Discrete only - only the Ati card is available on the bus and everything works as usual
- Switchable mode - both cards are initially available on the bus. VGA display and internal LCD are accessible from the Intel card. Currently (no KMS in intel, fglrx for ATI) everything works ok. The framebuffer is initialized to the Intel card and I can disable the Ati card via a vendor-specific ACPI call (Lenovo U330) made from a custom-made kernel module, which is loaded before X starts. At the time X starts everything is fine and dandy: it can only see the Intel card and I can happily save A-LOT of working on the Intel card only (Ati+Intel on is 21W, Intel only is 13W).

Now I can ask my question: will using Ati KMS and Intel KMS somehow mess up my setting? I fear that at kernel load, the Intel KMS will try to initialize the display, but the Ati KMS would also be loaded. Will they conflict with each other? Will I be able to unload Ati modules later on in userspace and load my Ati-turn-off module?

You can unload the radeon drm module if it loads or blacklist it so it doesn't load in the first place if you don't want to load it.

[signor_rossi]

Debian stable (lenny is the current stable) only gets updates for security fixes and bugs. No new functionality is supposed to go in, that's why it's called stable.

Ups, knew about the security fixes but forgot that you don't get new package versions, but I am also used completely to the rolling release scheme now after almost ten years with Gentoo and Arch. :-)
@adg5f
So in 2.6.34 there will be the same support for the X600 that radeon-git without KMS offers now? ;-)
Great work btw, I would have never thought I would get radeon working this good on this old laptop of mine after so many years. And no, this is not meant ironically.

[Neuro]

Not likely.

Why? Were they late? Or are they doing something hackish? Any chance of them getting into the mainline kernel in the next version (2.6.34 perhaps)?

They work, although some users report a flash on the screen when the clocks change.

But no reports of "OMG, it fried my laptop"?

You can unload the radeon drm module if it loads or blacklist it so it doesn't load in the first place if you don't want to load it.

If I can unload it and it'll revert all the changes, it would do However, I can't blacklist, since I'd like to use the radeon drm whenever I run in "Discrete only" mode.

Thanks for your replies, they were very informative.

[Loris]

Does these set of patches apply to kernel 2.6.32 or 2.6.33?

Compiling right now on linus-2.6.git.

[agd5f]

So in 2.6.34 there will be the same support for the X600 that radeon-git without KMS offers now? ;-)

Hopefully better than what UMS currently offers.