Current support and roadmap for discrete graphics card hot switching
Hi all, What is the current support and roadmap for discrete graphics card hot switching in Xorg? See: https://bugs.launchpad.net/ubuntu/+source/xorg/+bug/312756 http://forum.notebookreview.com/showthread.php?t=258304 Hybrids with the Ability to turn off the 3d chip: AMD/ATI calls it PowerXpress and Nvidia HybridPower. It uses 2 graphics cards, one energy efficient with little 3d power and the other for gaming (fast and uses more/a lot of power), the user can choose which chip to use. This is not a new concept. Sony has build in 2 graphics chips into some of their laptops for years. In the past a reboot was required to switch between the chips. With the new generation it is possible to change between chips on the fly, the screen will flicker but no need to reboot. At least in Windows Vista (XP, Linux not supported) the user can switch freely between the chips or set up a profile to do so automatically (eg when on battery use low power chip and when plugged in use the more powerful chip). The graphic card hybrid not only works with two Nvidia or AMD cards but the low power Intel graphics solutions (mostly shard memory) can also be combined with 3d chips from AMD or Nvidia. This solution is ideal for users who want maximum battery life and be able to play current games. The most likely combination is Intel shard memory graphics card for battery life and some low to mid level 3d chip. This will not give great 3d performance but enable you to play some games. Limitations are the drivers. Special drivers are needed depending on which graphic chips are combined in the hybrid. This will most likely make you depended on the Notebook manufactures driver support. It is uncertain if 3rd party drivers (such as laptopvideo2go) will be usable. One of the models is the Sony Vaio Z series. Right now, both cards are visible under Linux, but there is no way to hot-switching-off (if that is a word...) the Nvidia card. For a summary of users' experimentation with this laptop and Linux. There are currently ~40 users of Sony Vaio Z series using Linux that would like this feature to be implemented. See: https://launchpad.net/~sony-vaio-z-series<https://launchpad.net/%7Esony-vaio-z-series> Also, see: http://forum.notebookreview.com/showthread.php?t=325616&page=1 http://forum.notebookreview.com/showthread.php?t=325616&page=2 http://forum.notebookreview.com/showthread.php?t=325616&page=3 http://forum.notebookreview.com/showthread.php?t=325616&page=4 http://forum.notebookreview.com/showthread.php?t=325616&page=5 Thanks, Albert. ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On Thu, 15 Jan 2009 10:21:53 + Albert Vilella wrote: > Hi all, > > What is the current support and roadmap for discrete graphics card hot > switching in Xorg? OLPC does automatic switching of display controller for power management. > There are currently ~40 users of Sony Vaio Z series using Linux that would > like this feature to be implemented. See: > > https://launchpad.net/~sony-vaio-z-series<https://launchpad.net/%7Esony-vaio-z-series> Perhaps they can all extract the documentation from Nvidia 8) ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
> > > What is the current support and roadmap for discrete graphics card hot > > switching in Xorg? > > OLPC does automatic switching of display controller for power management. Interesting. So the OLPC also has a discrete and an integrated graphics card? Are these Intel or what brand? > There are currently ~40 users of Sony Vaio Z series using Linux that would > like this feature to be implemented. See: > > > https://launchpad.net/~sony-vaio-z-series<https://launchpad.net/%7Esony-vaio-z-series> > <https://launchpad.net/%7Esony-vaio-z-series> > > Perhaps they can all extract the documentation from Nvidia 8) With the bad spell they have had lately, hopefully Nvidia will disappear as a company soon and let Intel and ATI carry on supporting Linux :-p No, seriously, I am guessing Nvidia will add this to the list of things to do for their binary blob, but ATI/AMD should be in a better position to support or help in supporting this feature in Linux, right? ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On Thu, Jan 15, 2009 at 11:21, Albert Vilella wrote: > Hi all, > > What is the current support and roadmap for discrete graphics card hot > switching in Xorg? There is no support, and AFAIK no roadmap either. There are many technical reasons why this is not possible today. In short, I wouldn't suggest getting a dual GPU laptop with the purpose of using it under linux, as one of the GPUs will probably stay unused. Stephane ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
Twas brillig at 11:40:00 15.01.2009 UTC+01 when marche...@icps.u-strasbg.fr did gyre and gimble: SM> In short, I wouldn't suggest getting a dual GPU laptop with the SM> purpose of using it under linux, as one of the GPUs will probably SM> stay unused. Well, it should be possible to run some number-crunching on unused discrete GPU :) -- pgph1DXHB2QqS.pgp Description: PGP signature ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On Thu, 15 Jan 2009, Stephane Marchesin wrote: > On Thu, Jan 15, 2009 at 11:21, Albert Vilella wrote: >> Hi all, >> >> What is the current support and roadmap for discrete graphics card hot >> switching in Xorg? > > There is no support, and AFAIK no roadmap either. There are many > technical reasons why this is not possible today. In short, I wouldn't > suggest getting a dual GPU laptop with the purpose of using it under > linux, as one of the GPUs will probably stay unused. or fail to start X without setting the BusID, since the server can't decide which one to use: http://bugs.freedesktop.org/show_bug.cgi?id=18321 been there since 1.5, and unfortunately hasn't gathered much activity.. t ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On Thu, Jan 15, 2009 at 10:33:09AM +, Alan Cox wrote: > On Thu, 15 Jan 2009 10:21:53 + > Albert Vilella wrote: > > What is the current support and roadmap for discrete graphics card hot > > switching in Xorg? > > OLPC does automatic switching of display controller for power management. That's not even remotely comparable. (FWIW, every Nokia internet tablet, including 2004/2005's 770, right up to 2008's N810, has done this.) Cheers, Daniel signature.asc Description: Digital signature ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On Thu, Jan 15, 2009 at 12:22, Albert Vilella wrote: >>> There is no support, and AFAIK no roadmap either. There are many >>> technical reasons why this is not possible today. In short, I wouldn't >>> suggest getting a dual GPU laptop with the purpose of using it under >>> linux, as one of the GPUs will probably stay unused. > > Just to clarify the current situation: in some laptops, like Sony Vaio > models (SZ-series, Z-series), this feature is "partially" working: > > One can do a cold reboot: using the hardware stamina/speed switch in the > laptop to switch off the discrete > graphics card (Nvidia) at BIOS time. But the latest models (Z-series) allow > for a hot switch, right now only in Windows Vista. > If one installs Linux on these, both the Nvidia and the Intel will appear in > lspci, but xorg will not be able to handle both, > and the Nvidia hardware will be wasting battery and not being used. > Some people has managed to revert back to the cold reboot feature by > installing Windows XP on the laptop, > then switching on/off the discrete graphics card at BIOS time. This is not what you'd call support. This is just the bios exposing a single graphics card. As far as X.Org is concerned, there is only one graphics card at a time. > > So the next step is the hot switch. My hunch is that Windows Vista does some > sort of "gdm restart" equivalent, > by the looks of this video on computer.tv: > > Jump to 4:10 for the switching bit: > > http://www.youtube.com/watch?v=Qcvu2Aluy7g > > This machine is the successor of the SZ premium series, and has a Dynamic > Hybrid Graphics system that will enable/disable the nvidia graphics card > using a software "hot" switch instead of a hardware "cold" switch (SZ > series). > > http://vaio-online.sony.com/prod_info/series1/z/interview_Z/index_05.html > > Can I ask someone who is expert enough in xorg to give a list of > blockers/things to try for this to happen, so that people can play with? > For example, people has been investigating BiosBase on the Nvidia side of > things: > > http://avilella.googlepages.com/vaioz (look for BiosBase) > http://bugs.freedesktop.org/show_bug.cgi?id=2597#c37 > Basically, we lack : - documentation on how to switch GPUs at the laptop level (i.e. do what the bios does at boot when you choose the card in the bios) - documentation on cold booting the nvidia GPU - driver support on both sides implementing proper GPU power up/shut down (we're talking about something big here) and if you want to keep your session in between, we lack - X.Org infrastructure to hand a session from a graphics driver to another (there are a million of possible problems here) - drivers supporting said infrastructure (we're talking about something real huge here) IMO all this is not very likely to happen. When you buy a laptop on which you want to run linux, I really suggest you check hardware compatibility. This is no different than unsupported wifi chips. Stephane ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
> > There is no support, and AFAIK no roadmap either. There are many >> technical reasons why this is not possible today. In short, I wouldn't >> suggest getting a dual GPU laptop with the purpose of using it under >> linux, as one of the GPUs will probably stay unused. > > Just to clarify the current situation: in some laptops, like Sony Vaio models (SZ-series, Z-series), this feature is "partially" working: One can do a cold reboot: using the hardware stamina/speed switch in the laptop to switch off the discrete graphics card (Nvidia) at BIOS time. But the latest models (Z-series) allow for a hot switch, right now only in Windows Vista. If one installs Linux on these, both the Nvidia and the Intel will appear in lspci, but xorg will not be able to handle both, and the Nvidia hardware will be wasting battery and not being used. Some people has managed to revert back to the cold reboot feature by installing Windows XP on the laptop, then switching on/off the discrete graphics card at BIOS time. So the next step is the hot switch. My hunch is that Windows Vista does some sort of "gdm restart" equivalent, by the looks of this video on computer.tv: Jump to 4:10 for the switching bit: http://www.youtube.com/watch?v=Qcvu2Aluy7g This machine is the successor of the SZ premium series, and has a Dynamic Hybrid Graphics system that will enable/disable the nvidia graphics card using a software "hot" switch instead of a hardware "cold" switch (SZ series). http://vaio-online.sony.com/prod_info/series1/z/interview_Z/index_05.html Can I ask someone who is expert enough in xorg to give a list of blockers/things to try for this to happen, so that people can play with? For example, people has been investigating BiosBase on the Nvidia side of things: http://avilella.googlepages.com/vaioz (look for BiosBase) http://bugs.freedesktop.org/show_bug.cgi?id=2597#c37 Thanks, Albert. ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On Thu, Jan 15, 2009 at 13:30, Colin Guthrie wrote: > 'Twas brillig, and Stephane Marchesin at 15/01/09 10:40 did gyre and gimble: >> On Thu, Jan 15, 2009 at 11:21, Albert Vilella wrote: >>> Hi all, >>> >>> What is the current support and roadmap for discrete graphics card hot >>> switching in Xorg? >> >> There is no support, and AFAIK no roadmap either. There are many >> technical reasons why this is not possible today. In short, I wouldn't >> suggest getting a dual GPU laptop with the purpose of using it under >> linux, as one of the GPUs will probably stay unused. > > Erm, forgive me if I'm wrong, but I thought this was something that Adam > Jackson's Shatter work would go part way to resolving? > > http://www.ziobudda.net/node/103982 > Sure, however : - it's not done yet - even then, the rest of the points I raised are still not covered - it relies on EXA, which is not implemented by the nvidia binary driver. So even if it was done, you could switch between "intel" and "nv" (or "intel" and "nouveau" but still there is no finished 3D). Stephane ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
'Twas brillig, and Stephane Marchesin at 15/01/09 10:40 did gyre and gimble: > On Thu, Jan 15, 2009 at 11:21, Albert Vilella wrote: >> Hi all, >> >> What is the current support and roadmap for discrete graphics card hot >> switching in Xorg? > > There is no support, and AFAIK no roadmap either. There are many > technical reasons why this is not possible today. In short, I wouldn't > suggest getting a dual GPU laptop with the purpose of using it under > linux, as one of the GPUs will probably stay unused. Erm, forgive me if I'm wrong, but I thought this was something that Adam Jackson's Shatter work would go part way to resolving? http://www.ziobudda.net/node/103982 Col -- Colin Guthrie gmane(at)colin.guthr.ie http://colin.guthr.ie/ Day Job: Tribalogic Limited [http://www.tribalogic.net/] Open Source: Mandriva Linux Contributor [http://www.mandriva.com/] PulseAudio Hacker [http://www.pulseaudio.org/] Trac Hacker [http://trac.edgewall.org/] ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On Thu, 15 Jan 2009 10:39:44 + Albert Vilella wrote: > > > > > What is the current support and roadmap for discrete graphics card hot > > > switching in Xorg? > > > > OLPC does automatic switching of display controller for power management. > > > Interesting. So the OLPC also has a discrete and an integrated graphics > card? > Are these Intel or what brand? It has a dumb frame buffer that can run with the main video card turned off, so the image can be updated on the dumb fb and the main video powered down as much as possible. None of it is Intel, Intel are the folks who compete with OLPC ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On Thu, Jan 15, 2009 at 01:05:10PM +, Alan Cox wrote: > On Thu, 15 Jan 2009 10:39:44 + > Albert Vilella wrote: > > > > What is the current support and roadmap for discrete graphics card hot > > > > switching in Xorg? > > > > > > OLPC does automatic switching of display controller for power management. > > > > Interesting. So the OLPC also has a discrete and an integrated graphics > > card? > > Are these Intel or what brand? > > It has a dumb frame buffer that can run with the main video card turned > off, so the image can be updated on the dumb fb and the main video > powered down as much as possible. Right, which reduces it to a simple power management issue akin to powering down the 3D core on any modern chipset when you're not doing any rendering. Adding different devices with separate drivers is another matter altogether. Cheers, Daniel signature.asc Description: Digital signature ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
> Right, which reduces it to a simple power management issue akin to > powering down the 3D core on any modern chipset when you're not doing > any rendering. > > Adding different devices with separate drivers is another matter > altogether. Isn't dual driver support logically equivalent to xrandr mirrored to both with either one or the other currently a 'switched away' vt ? ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On Thu, Jan 15, 2009 at 15:13, Alan Cox wrote: >> Right, which reduces it to a simple power management issue akin to >> powering down the 3D core on any modern chipset when you're not doing >> any rendering. >> >> Adding different devices with separate drivers is another matter >> altogether. > > Isn't dual driver support logically equivalent to xrandr mirrored to both > with either one or the other currently a 'switched away' vt ? Yeah, in a perfect world it is. But : - cross-card xrandr doesn't exist - not all drivers support xrandr + EXA which shatter will require to achieve cross-card xrandr - that also assumes the drivers power down the chip while switched away Stephane ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On Thu, Jan 15, 2009 at 02:13:45PM +, Alan Cox wrote: > > Right, which reduces it to a simple power management issue akin to > > powering down the 3D core on any modern chipset when you're not doing > > any rendering. > > > > Adding different devices with separate drivers is another matter > > altogether. > > Isn't dual driver support logically equivalent to xrandr mirrored to both > with either one or the other currently a 'switched away' vt ? Yes, which we don't really handle well now. signature.asc Description: Digital signature ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
now the question is: leaving Nvidia and the downstream problems aside, how difficult would it be to convince ATI/AMD to provide such kind of documentation? Anyone insider here that can answer? On Thu, Jan 15, 2009 at 2:34 PM, Daniel Stone wrote: > On Thu, Jan 15, 2009 at 02:13:45PM +, Alan Cox wrote: > > > Right, which reduces it to a simple power management issue akin to > > > powering down the 3D core on any modern chipset when you're not doing > > > any rendering. > > > > > > Adding different devices with separate drivers is another matter > > > altogether. > > > > Isn't dual driver support logically equivalent to xrandr mirrored to both > > with either one or the other currently a 'switched away' vt ? > > Yes, which we don't really handle well now. > > -BEGIN PGP SIGNATURE- > Version: GnuPG v1.4.9 (GNU/Linux) > > iEYEARECAAYFAklvSV4ACgkQUVYB1rKAgJR3YQCeItsIf1FxuSjmGzyH4Djebh9H > /Y8AoJxc3u2K+TrG3Sv7yG+tR5pdWppq > =Kk8B > -END PGP SIGNATURE- > > ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On Jan 15, 09 16:20:32 +, Albert Vilella wrote: > leaving Nvidia and the downstream problems aside, how difficult would it be > to convince ATI/AMD to provide such kind of documentation? > Anyone insider here that can answer? In the current (approximate) list of - 3D documentation (huge) - General(!) Powermanagement support - r8xx Documentation (which will be out then) - Enhanced 3D documentation (even larger) - Displayport support this comes probably last. Add a year per line, and you'll get your docs in about 5 years time. Pessimistic from the time it takes point of view, optimistic from the "we can get the information" point of view. Matthias -- Matthias Hopf ____ __ Maxfeldstr. 5 / 90409 Nuernberg (_ | | (_ |__ m...@mshopf.de Phone +49-911-74053-715 __) |_| __) |__ R & D www.mshopf.de ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On Thu, Jan 15, 2009 at 11:20 AM, Albert Vilella wrote: > now the question is: > > leaving Nvidia and the downstream problems aside, how difficult would it be > to convince ATI/AMD to provide such kind of documentation? > Anyone insider here that can answer? We can definitely look into it, the problem is we already have a backlog of stuff with higher priority (finishing 3D, newer power management bits, investigating IDCT/UVD, etc.) to work through at the moment, so I cannot say when we'd get to hybrid graphics. The other problem is that since many of these hybrid solutions are multi-vendor, we may not have the rights release certain IP. Even if would could release some information, as has been stated previously, the driver stack needs significant work to support something like this. Alex ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
Thanks Alex for your answer. It's great that you can look into it, and I can say that for one, I am optimistic about it :-p On Thu, Jan 15, 2009 at 4:56 PM, Alex Deucher wrote: > On Thu, Jan 15, 2009 at 11:20 AM, Albert Vilella > wrote: > > now the question is: > > > > leaving Nvidia and the downstream problems aside, how difficult would it > be > > to convince ATI/AMD to provide such kind of documentation? > > Anyone insider here that can answer? > > We can definitely look into it, the problem is we already have a > backlog of stuff with higher priority (finishing 3D, newer power > management bits, investigating IDCT/UVD, etc.) to work through at the > moment, so I cannot say when we'd get to hybrid graphics. The other > problem is that since many of these hybrid solutions are multi-vendor, > we may not have the rights release certain IP. Even if would could > release some information, as has been stated previously, the driver > stack needs significant work to support something like this. > > Alex > ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
Stephane Marchesin wrote: > and if you want to keep your session in between, we lack > - X.Org infrastructure to hand a session from a graphics driver to > another (there are a million of possible problems here) Right; like a million display parameters which a client can query, but for which there is no mechanism to request notification of changes, and thus are (implicitly) constant over the lifetime of the client. I know that the X developers don't consider incompatible changes to be completely out of the question, but if you're talking about a particular screen suddenly changing e.g. its glGet* values, I don't see that happening. And I don't think that it's realistic for the server to expose a single set of parameters for two very different graphics chips. It's more realistic to treat this as a traditional multiple-"Screen" setup, with the ability to enable and disable screens. Obviously, windows would have to either be opened on the appropriate screen (programs which need the 3D GPU on the screen which has one), or the application/toolkit would need to explicitly provide migration. -- Glynn Clements ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On Thursday 15 of January 2009, Alex Deucher wrote: > On Thu, Jan 15, 2009 at 11:20 AM, Albert Vilella wrote: > > now the question is: > > > > leaving Nvidia and the downstream problems aside, how difficult would it > > be to convince ATI/AMD to provide such kind of documentation? > > Anyone insider here that can answer? > > We can definitely look into it, the problem is we already have a > backlog of stuff with higher priority > management bits, investigating IDCT/UVD, etc.) to work through at the > moment, so I cannot say when we'd get to hybrid graphics. Ability to just activate desired graphic card from OS when bios set in hybrid-mode would be very good start. I hope it would allow to stop X, switch card, start X on second card. > Alex -- Arkadiusz MiĆkiewiczPLD/Linux Team arekm / maven.plhttp://ftp.pld-linux.org/ ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
How about a "gdm restart"? That is effectively an X server restart, right? Then it's only about switching on and off the hardware, right? On Thu, Jan 15, 2009 at 7:20 PM, Glynn Clements wrote: > > Stephane Marchesin wrote: > > > and if you want to keep your session in between, we lack > > - X.Org infrastructure to hand a session from a graphics driver to > > another (there are a million of possible problems here) > > Right; like a million display parameters which a client can query, but > for which there is no mechanism to request notification of changes, > and thus are (implicitly) constant over the lifetime of the client. > > I know that the X developers don't consider incompatible changes to be > completely out of the question, but if you're talking about a > particular screen suddenly changing e.g. its glGet* values, I don't > see that happening. > > And I don't think that it's realistic for the server to expose a > single set of parameters for two very different graphics chips. > > It's more realistic to treat this as a traditional multiple-"Screen" > setup, with the ability to enable and disable screens. Obviously, > windows would have to either be opened on the appropriate screen > (programs which need the 3D GPU on the screen which has one), or the > application/toolkit would need to explicitly provide migration. > > -- > Glynn Clements > ___ > xorg mailing list > xorg@lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/xorg > ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On 15.01.2009 13:04, Stephane Marchesin wrote: > Basically, we lack : > - documentation on how to switch GPUs at the laptop level (i.e. do > what the bios does at boot when you choose the card in the bios) > - documentation on cold booting the nvidia GPU > - driver support on both sides implementing proper GPU power up/shut down > (we're talking about something big here) I always wondered how this works at the hardware level. How are the chips shut off? I guess for the external chip what you really want to do is cut power completely - is this somehow standardized, can you do that through ACPI? How are the outputs switched - some (external to the graphic chips) multiplexers or what, again controlled by ACPI bios? Isn't power up quite similar to just regular posting of the chip? > > and if you want to keep your session in between, we lack > - X.Org infrastructure to hand a session from a graphics driver to > another (there are a million of possible problems here) > - drivers supporting said infrastructure > (we're talking about something real huge here) > > IMO all this is not very likely to happen. Yes, keeping the session sounds very complicated (though that's probably what people want, for most it probably wouldn't make a difference between rebooting or just restarting X). For the manual switch, I'm not sure how complicated that really would be - of course lacking documentation would be a problem. > When you buy a laptop on > which you want to run linux, I really suggest you check hardware > compatibility. This is no different than unsupported wifi chips. Well, for WiFi chips there's always the hope it will be supported in the future :-). Roland ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
> > > IMO all this is not very likely to happen. > Yes, keeping the session sounds very complicated (though that's probably > what people want, for most it probably wouldn't make a difference > between rebooting or just restarting X). For the manual switch, I'm not > sure how complicated that really would be - of course lacking > documentation would be a problem. I think a user logout->login, which at least in Ubuntu corresponds to a gdm restart nowadays, is a much leaner option than a cold reboot of the system. You only lose the opened windows, but all services like connection to internet, etc, are kept alive, so it's better than a reboot. > > > > When you buy a laptop on > > which you want to run linux, I really suggest you check hardware > > compatibility. This is no different than unsupported wifi chips. > Well, for WiFi chips there's always the hope it will be supported in the > future :-). > > Roland > ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On Fri, Jan 16, 2009 at 8:45 AM, Albert Vilella wrote: > I think a user logout->login, which at least in Ubuntu corresponds to a gdm > restart nowadays, is a much > leaner option than a cold reboot of the system. You only lose the opened > windows, > but all services like connection to internet, etc, are kept alive, so it's > better than a reboot. Just thinking out loud here: If desktop session management were good enough, even open windows could be "persisted". Even better would be if there were a mechanism to transparently disconnect an app from one X session, wait for X to restart, and then attach it to the new session. Probably doable at the toolkit level, but that doesn't help with all the zillions of apps written against legacy toolkits. Random idea: There are already several special-purpose X servers that run on top of Xorg supporting special magic like hardware compositing. What if there were a server that could dynamically dispatch to/from different Xorg instances? It would notice when Xorg dies, and stop sending it events. When a new Xorg launches, it would send a series of "new window" commands, and attach all of its clients to those windows. Right now I'm assuming that both cards would support equivalent resolutions and color depths. If not, then never mind. :-P Anyway, I agree that restarting the server is less painful than a full reboot. -- William Tracy afishion...@gmail.com -- wtr...@calpoly.edu Vice President, Cal Poly Linux Users' Group http://www.cplug.org "I disapprove of what you say, but I will defend to the death your right to say it." -- Evelyn Beatrice Hall, frequently mis-attributed to Voltaire ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
William Tracy wrote: > > I think a user logout->login, which at least in Ubuntu corresponds to a gdm > > restart nowadays, is a much > > leaner option than a cold reboot of the system. You only lose the opened > > windows, > > but all services like connection to internet, etc, are kept alive, so it's > > better than a reboot. > > Just thinking out loud here: If desktop session management were good > enough, even open windows could be "persisted". > > Even better would be if there were a mechanism to transparently > disconnect an app from one X session, wait for X to restart, and then > attach it to the new session. Probably doable at the toolkit level, > but that doesn't help with all the zillions of apps written against > legacy toolkits. > > Random idea: There are already several special-purpose X servers that > run on top of Xorg supporting special magic like hardware compositing. > What if there were a server that could dynamically dispatch to/from > different Xorg instances? It would notice when Xorg dies, and stop > sending it events. When a new Xorg launches, it would send a series of > "new window" commands, and attach all of its clients to those windows. > > Right now I'm assuming that both cards would support equivalent > resolutions and color depths. If not, then never mind. :-P The problem is that, in order to operate without explicit support from application code, both cards would need to support equivalent *everything*. A solution which relies upon the toolkit to do everything will only work for applications where ... the toolkit does everything, i.e. those whose interaction with X is limited to creating stock widgets using parameters which don't depend upon any hardware-dependent parameters. If the application queries the bit depth, or the pixel format, or the maximum size of a texture, or the maximum depth of the matrix stack, or any of a zillion other hardware-related parameters, you need to either: 1. add a mechanism to indicate that the parameter has changed, and ensure that applications allow for such changes, or 2. expose an interface which either the toolkit or the X server can emulate in its entirety atop a dumb framebuffer, and eat the (potentially huge) performance hit when it has to do so. -- Glynn Clements ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
Hi Alex, Any official or officious news from the ATI/AMD front about this? Thanks and apologies for my impatience :-p Albert. On Thu, Jan 15, 2009 at 4:56 PM, Alex Deucher wrote: > On Thu, Jan 15, 2009 at 11:20 AM, Albert Vilella > wrote: > > now the question is: > > > > leaving Nvidia and the downstream problems aside, how difficult would it > be > > to convince ATI/AMD to provide such kind of documentation? > > Anyone insider here that can answer? > > We can definitely look into it, the problem is we already have a > backlog of stuff with higher priority (finishing 3D, newer power > management bits, investigating IDCT/UVD, etc.) to work through at the > moment, so I cannot say when we'd get to hybrid graphics. The other > problem is that since many of these hybrid solutions are multi-vendor, > we may not have the rights release certain IP. Even if would could > release some information, as has been stated previously, the driver > stack needs significant work to support something like this. > > Alex > ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Current support and roadmap for discrete graphics card hot switching
On Tue, Jan 20, 2009 at 11:37 AM, Albert Vilella wrote: > Hi Alex, > > Any official or officious news from the ATI/AMD front about this? > > Thanks and apologies for my impatience :-p > Sorry haven't had a chance to look into this yet. Anything more than a cursory glance will probably be a ways off. As I said we have higher priority stuff to work through at the moment. Alex > Albert. > > On Thu, Jan 15, 2009 at 4:56 PM, Alex Deucher wrote: >> >> On Thu, Jan 15, 2009 at 11:20 AM, Albert Vilella >> wrote: >> > now the question is: >> > >> > leaving Nvidia and the downstream problems aside, how difficult would it >> > be >> > to convince ATI/AMD to provide such kind of documentation? >> > Anyone insider here that can answer? >> >> We can definitely look into it, the problem is we already have a >> backlog of stuff with higher priority (finishing 3D, newer power >> management bits, investigating IDCT/UVD, etc.) to work through at the >> moment, so I cannot say when we'd get to hybrid graphics. The other >> problem is that since many of these hybrid solutions are multi-vendor, >> we may not have the rights release certain IP. Even if would could >> release some information, as has been stated previously, the driver >> stack needs significant work to support something like this. >> >> Alex > > ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Getting there Re: Current support and roadmap for discrete graphics card hot switching
Hi all, Just to give a heads up to everyone interested in this thread, one of the Sony Vaio Z-series launchpad team members has found a way to turn on/off the discrete nvidia graphics card by investigating the DSDT tables. So we have at least 60 happy users that can enjoy this feature now. Details here: https://launchpad.net/~sony-vaio-z-series http://www.basyskom.org/~eva/log_installation_vaio_z21vnx.html - The mode can be switched through sysfs: echo stamina > /sys/devices/platform/sony-laptop/speed_stamina echo speed > /sys/devices/platform/sony-laptop/speed_stamina Make sure you have no X server running while doing this, as X will break. The drivers starts in stamina mode as default. This can be changed by a parameter, to startup in speed mode: insmod sony-laptop speed_stamina=1 It is unsure initialisation of the Nvidia card is complete and this has not been tested. Loading such a module at boot time might break your linux installation (or at least making it impossible to start up the X server). Unfortunately the PCI system doesn't really recognize the change and still shows the 00:02.0 Intel graphics controller and the 01:00.0 Nvidia graphics controller, but not the 00:02.1 Intel card. So if you use scripts to switch your X.org config, this will break. - I guess it's a matter of X.org picking up from here now, trying to get this to work better with a running X server somehow. Cheers, Albert. On Thu, Jan 15, 2009 at 10:21 AM, Albert Vilella wrote: > Hi all, > > What is the current support and roadmap for discrete graphics card hot > switching in Xorg? > > See: > > https://bugs.launchpad.net/ubuntu/+source/xorg/+bug/312756 > > http://forum.notebookreview.com/showthread.php?t=258304 > > Hybrids with the Ability to turn off the 3d chip: > > AMD/ATI calls it PowerXpress and Nvidia HybridPower. It uses 2 graphics > cards, one energy efficient with little 3d power and the other for gaming > (fast and uses more/a lot of power), the user can choose which chip to use. > This is not a new concept. Sony has build in 2 graphics chips into some of > their laptops for years. In the past a reboot was required to switch between > the chips. With the new generation it is possible to change between chips on > the fly, the screen will flicker but no need to reboot. At least in Windows > Vista (XP, Linux not supported) the user can switch freely between the chips > or set up a profile to do so automatically (eg when on battery use low power > chip and when plugged in use the more powerful chip). > > The graphic card hybrid not only works with two Nvidia or AMD cards but the > low power Intel graphics solutions (mostly shard memory) can also be > combined with 3d chips from AMD or Nvidia. This solution is ideal for users > who want maximum battery life and be able to play current games. The most > likely combination is Intel shard memory graphics card for battery life and > some low to mid level 3d chip. This will not give great 3d performance but > enable you to play some games. > > Limitations are the drivers. Special drivers are needed depending on which > graphic chips are combined in the hybrid. This will most likely make you > depended on the Notebook manufactures driver support. It is uncertain if 3rd > party drivers (such as laptopvideo2go) will be usable. > > One of the models is the Sony Vaio Z series. Right now, both cards are > visible under Linux, but there is no way to hot-switching-off (if that is a > word...) the Nvidia card. For a summary of users' experimentation with this > laptop and Linux. > > There are currently ~40 users of Sony Vaio Z series using Linux that would > like this feature to be implemented. See: > > https://launchpad.net/~sony-vaio-z-series > > Also, see: > > http://forum.notebookreview.com/showthread.php?t=325616&page=1 > http://forum.notebookreview.com/showthread.php?t=325616&page=2 > http://forum.notebookreview.com/showthread.php?t=325616&page=3 > http://forum.notebookreview.com/showthread.php?t=325616&page=4 > http://forum.notebookreview.com/showthread.php?t=325616&page=5 > > Thanks, > > Albert. > ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg
Re: Getting there Re: Current support and roadmap for discrete graphics card hot switching
_DSM? Nnngh. -- Matthew Garrett | mj...@srcf.ucam.org ___ xorg mailing list xorg@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/xorg