Weekly Edition
Return to the Press pageRecent Features
| |||||||||||||
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch)
Linux-Watch looks at AMD's
acquisition of ATI. "With AMD at the helm, I can envision ATI
finally open-sourcing the code to its proprietary drivers. That should
quickly result in much better performance for Linux users, and better
performance for the Windows and Mac platforms as well."
(Log in to post comments)
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 25, 2006 15:40 UTC (Tue) by beoba (guest, #16942) [Link] Does AMD have a known track record with such things, or is the linked article entirely speculative? I'm not finding any reasoning beyond "I think it'd be a good idea".
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 25, 2006 15:45 UTC (Tue) by dmantione (guest, #4640) [Link] AMD will only do it if we put the pressure on it.
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 25, 2006 15:54 UTC (Tue) by dmarti (subscriber, #11625) [Link] "We" as in Intel, and "pressure" as in "decent, non-power-hogging 3D with out-of-the-box support"?
So, let me get this straight...
MSFT Vista requires 3D, so generic PCs must have decent 3D.
Intel can't afford not to be the differentiating "inside" product in the PC.
Intel has to make 3D hardware.
Intel doesn't want to join the driver-of-the-month club or pull..."edgy"...legal stunts to do proprietary drivers. And Intel has enough of a patent portfolio that it can scoff at the ATI/NVIDIA patent standoff anyway.
Therefore, Intel does open-source drivers, and MSFT Vista saves Linux from the 3D driver mess.
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 25, 2006 17:02 UTC (Tue) by mjr (guest, #6979) [Link] Indeed the article seems purely conjecture; now would be the time for the more prominent members of the Free Software community to approach AMD (as I'm sure they pretty much realize). I'd guess grassroots support wouldn't hurt, but keep it polite, people. And remember, while getting their drivers released as free software would be nice, just getting them to open up with the specs would be a big win as well.
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 25, 2006 15:54 UTC (Tue) by pbardet (guest, #22762) [Link] The word envision seems speculative enough to me.
What we can wonder though is: does AMD provide kernel coders some documentation about the new features of their processors, so that optimizations can be made instead of relying on "standard" i686 instruction set ?
If so, then there may be a slight chance they would apply the same methods to their graphics chipsets, otherwise, I can't see a change for us.
I guess we'll have to wait to see.
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 25, 2006 16:02 UTC (Tue) by dmantione (guest, #4640) [Link] Yes, they do. AMD documentation documents every detail of the processorand is available for free download. Big difference between ATi and AMD is that ATi does a lot of digital restrictions management (Macrovision analog, HDCP, ....) and AMD doesn't. Documenting the chips in detail means also documenting how to switch off Macrovision and HDCP.
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 25, 2006 16:04 UTC (Tue) by corbet (editor, #1) [Link] AMD has been quite Linux-friendly with its processors - that's why Linux had great x86-64 support well ahead of those other operating systems. I think they see that this strategy paid off for them. One can really hope that it will carry through to graphics drivers, once the acquisition is complete. Until we actually see something from the company, though, hope is all that it will be.
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 25, 2006 16:30 UTC (Tue) by vondo (guest, #256) [Link] Yes, but the market for cutting edge 3D graphics on linux for people who are opposed to binary drivers is small.
The market for high-powered 64-bit processors for workstations and (super)computer clusters is enormous by comparison.
I'm not going to be holding my breath.
Circular reasoning Posted Jul 25, 2006 19:03 UTC (Tue) by man_ls (subscriber, #15091) [Link] It is a bit of a chicken-and-egg problem: one of the reasons why desktop Linux has not caught on is graphic cards, and therefore graphic card makers do not see much need for open drivers. It is, in fact, not very different from what happened to 64 bits -- Intel did not see any possibilities in it beyond Itanium, and Itanium was a big flop. So there was little market for 64-bit processors until AMD started selling them. Then it exploded and Intel had to jump aboard, but meanwhile AMD has had a jumpstart of some months.Similarly, opening the graphics drivers could mean paving the way for desktop Linux, which in turn would put ATI (and AMD) ahead of the game with their libre, supported, high-performance drivers. Now imagine if Linux became a gamer's dream operating system.
Circular reasoning Posted Jul 25, 2006 20:02 UTC (Tue) by NAR (subscriber, #1313) [Link] imagine if Linux became a gamer's dream operating system.I think it's not just good 3D which is needed - gamers expect that they buy the DVD, put it into the Windows computer or a console and it "just works". I'm not sure that it's possible to the same in the Linux world - will the same binary work well on the last few versions of SuSE, Fedora and Ubuntu? Not to mention that the support of a desktop Linux system tends be lower than the windows systems: I mean WinXP was released 5 years ago and I can run the latest games on it - could you run a current binary program on a 5 years old Linux distribution (e.g. Debian potato)?
Circular reasoning Posted Jul 26, 2006 2:58 UTC (Wed) by xanni (subscriber, #361) [Link] Why would you want your game to run on multiple Linux distros? Surely it's easier to make a "Live-CD" with whatever kernel and libraries the game needs on it and run the game as process 1.
Circular reasoning Posted Jul 26, 2006 5:31 UTC (Wed) by beoba (guest, #16942) [Link] What happens when the game gets patched?
Circular reasoning Posted Jul 26, 2006 5:32 UTC (Wed) by beoba (guest, #16942) [Link] Woops, forgot to mention this too:
Say you have a computer which uses the successor to PCIe, and your game CD has a kernel that does not have support for this successor.
Saved games Posted Jul 26, 2006 8:44 UTC (Wed) by dark (subscriber, #8483) [Link] I've been thinking along these lines, but it leaves the problem of where to store saved games. The best I can think of is to take a leaf from Sony and require people to buy a USB stick for the purpose.
Then there's the speed issue: one reason many games install to disk is to speed up access to the game files.
If a live-CD tried to copy its files to disk it has a significant risk of breaking the installed system.
Circular reasoning Posted Jul 26, 2006 5:53 UTC (Wed) by tao (subscriber, #17563) [Link] Seems to have worked well enough with the games from Loki...
Circular reasoning Posted Jul 26, 2006 9:24 UTC (Wed) by NAR (subscriber, #1313) [Link] I'm not that sure, Loki went backrupt, didn't it?
Circular reasoning Posted Jul 26, 2006 10:41 UTC (Wed) by drag (subscriber, #31333) [Link] "worked out well for games from loki" != "worked out well for loki"
Shit happens.
Circular reasoning Posted Jul 26, 2006 15:28 UTC (Wed) by man_ls (subscriber, #15091) [Link] I don't think gamers would have much problem in updating their operating system every few years. Heavy Windows users are used to reinstall the OS every few weeks anyway, otherwise performance suffers.As to binaries, it can be done -- the first example that comes to mind is the Java VM -- but I'm not sure that binaries are the best way to deliver games anyway. A different model could be explored maybe: distribute the engine as libre software, for free; charge for the levels with a different license. In the same vein as what Id did with the original Doom (a demo version for free, a pay version with full access to the levels), but a step further.
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 25, 2006 23:39 UTC (Tue) by jeroen (subscriber, #12372) [Link] The market for not so cutting edge Linux machines where you simply don't have all the trouble you get with proprietary drivers seems to be bigger. And at the moment, it's mostly the intel graphics chips which serve this market. And those chips come only with an intel CPU. So if AMD+ATI wants to compete in that market, they have to release specifications just like Intel is doing. Not only will they lose market share in the graphic chip market (even if it's only a tiny part at the moment, it might grow bigger in the next few years), but also in the CPU market.
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 26, 2006 0:12 UTC (Wed) by ewan (subscriber, #5533) [Link] I think you're neglecting the influence of the resellers - a lot of those compute clusters need visualisation machines with decent 3D, and the same companies (HP, IBM, SGI et al) sell both systems together.A large part of their proposition is support, and they can support neither the hardware nor the OS if it requires closed drivers. Buy a graphics workstation with (say) RHEL and you can have your graphics or your enterprise support. At the moment there's little choice, but with ATI and AMD one and the same these companies can use their processor spending as leverage to get open drivers to end the support problems.
Proving a negative Posted Jul 27, 2006 13:23 UTC (Thu) by gdt (subscriber, #6284) [Link] Yes, but the market for cutting edge 3D graphics on linux for people who are opposed to binary drivers is small. That's a rather definate statement of a negative and I'm wondering how you've come to that conclusion. As one simple example, two unrelated groups I know of looked at using small PCs with high end graphics cards to implement a video wall of large size and high resolution (think of the walls you see in James Bond movies) to be driven from a supercomputer. About 1,000 cards in each proposal. Both proposals foundered because they needed small changes to the graphics drivers. Proceeding with closed source drivers and hoping for the best was too high a risk for the project sponsors.
AMD, LinuxBIOS, and X.org Posted Jul 25, 2006 16:41 UTC (Tue) by brugolsky (subscriber, #28) [Link] AMD's open specs have been a great boon to LinuxBIOS; nearly all recent AMD-based motherboards are working or nearly working. Intel is always a mixed bag: they've released plenty of documentation for their NICs, ICH* SATA controllers, graphics chips, etc. But lots of simple northbridge/southbridge documentation is not freely available. And then there's the dreaded SMM gunk.
Currently, we've got this strange quandary where AMD-based systems have the best-documented innards on the parts that they own, but most PCIe motherboards have Nvidia motherboard chipsets; meanwhile Intel is producing the best-documented 3D graphics chips. :-(
If some of the ATI motherboard and graphics chipset documentation can be pried open, we can have very high-performance hardware running an entirely free software stack from initial power on.
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 25, 2006 16:11 UTC (Tue) by vmole (guest, #111) [Link] Yes, they do provide the CPU specific stuff. So does Intel, of course. None-the-less, this article is mostly wishful thinking. The CPU group and the graphics group will continue to be completely different groups for a *long* time. AMD is not aquiring ATI for the graphics technology, they're acquiring them for the motherboard chipsets technology - remember, Intel makes its own chipsets, while AMD does not (well, did not). The graphics stuff is just coming along for the ride. I'd be happy to be proved wrong, of course.
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 25, 2006 17:53 UTC (Tue) by brouhaha (subscriber, #1698) [Link] I think you're wrong about AMD not acquiring ATI for the graphics technology. That's where I think most of the value is. AMD has the ability to produce reasonably chipsets inhouse, and has done that in the past. It's true that they have fallen behind on that (due to reduced need with multiple third-party chipset vendors), so they'll certainly get some benefit from the ATI chipset operation. But chipsets are a fairly low-margin operation, so the acquisition can't be justified on that as the primary basis.Going forward, integrated graphics will be the big win for them. Initially that will be in the integrated chipsets for value and mobile systems, but over time all but the super-high-end graphics will get absorbed into the chipsets, and the value to medium-high-end graphics will migrate onto the processor die (especially for the mobile segment), along with most of the chipset functionality (USB ports, etc.)
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 25, 2006 18:05 UTC (Tue) by corbet (editor, #1) [Link] Along these lines, the Inquirer has an interesting article on what they think AMD is up to. It has everything to do with integrating the graphics engine into the processor core. If that is truly the way of it, AMD will be hard put to keep information on graphics proprietary while simultaneously keeping its processors open.
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 25, 2006 18:34 UTC (Tue) by brouhaha (subscriber, #1698) [Link] I suspect that they'll just bring it on-die. There's no point to actually integrating it into the processor core; just giving it a port to the crossbar switch is sufficient.
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 25, 2006 21:56 UTC (Tue) by drag (subscriber, #31333) [Link] Yep it would just be another core on a multicore chip.
Think about it.. A graphics chip utilizing the same manufacturing proccess and having the similar clock settings (even running at 50% speed would be fast) and such being used in opteron/amd64 chips.
I think that it would effectively blur the line between software rendering and hardware rendering. Also if they keep it open enough it should be very usefull for running high speed encryption/decryption as well as media encoding.
What AMD's ATI acquisition means for Linux (and Macs) (Linux-Watch) Posted Jul 26, 2006 7:23 UTC (Wed) by Wol (guest, #4433) [Link] You miss the point. Graphics and CPU technology are converging (what is a GPU, if not a powerful, dedicated, CPU?).
AMD need to put GPU technology into the CPU (at which point the CPU will become a massively parallel processor, and GPUs will disappear...). It won't be "just another core".
Look at the Open Graphics Project. Although it is allegedly "just a video card", it's actually a pretty powerful DSP and by reprogramming the firmware it would be very easy to turn it into a parallel co-processor, a pc on a chip, a soundcard, or whatever. If AMD can't put that flexibility into its CPUs, it'll be a goner in the next few years.
Charlie wasn't saying "this is a convenient acquisition so AMD can compete with Intel on the graphics front". He said "this is a survival play so AMD can catch up with what Intel currently have in planning". AMD need ATI so they can match the stuff that Intel currently has on the back of a napkin, that will take five years to appear but will be nothing like what we have today ...
Cheers,
Forget F/OSS drivers or specs Posted Jul 25, 2006 17:03 UTC (Tue) by jmorris42 (subscriber, #2203) [Link] We won't be seeing any change on the 3D front. Because while we still don't know WHY both ATI & NVIDIA are so insistent on keeping all their stuff secret it is doubtful that AMD being the ultimate owner of ATI's tech will change it.
Almost every hardware component in a modern desktop PC has good Free drivers but no video vendor (excepting Intel) wants to release 3D specs. Seen lots of theories float, no idea which might be true.
Theory 1: Specs were hard to come by in the past, but ATI did help X in the past as did Nvidia at one point. But no company has released squat since the DirectX 9 level. So of course if one is a Microsoft conspiracy theorist the idea that MSFT imposed some sort of non-disclosure in the licensing to get on the DX9 bandwagon is a good one. But a non-disclosure so all encompasing that nobody can even admit it exists? Knowing that exposing it would almost certainly set off another round of DOJ action nobody has leaked anything?
Theory 2: All and sundry know they are violating the hell out of various patents and fear releasing specs or source would make that plain, releasing the dread legions of patent lawyers. This one has the advantage of partially explaining why Intel doesn't care as much, having cross licenses with darned near everyone and being big enough to unleash a bit of patent wrath of their own back on anyone foolish enough to tackle them. But they can somehow keep their violations secret just by keeping the Linux folk from seeing how it works? Nobody is capable of reverse engineering or corporate spying?
Theory 3: The 3D hardware isn't really all that interesting, just a lot of brute force that any Chinese fab could replicate in short order if there were free drivers available. Considering the continual problems with drivers ATI & Nvidia have this one is my fav. If most of the value proposition is in the software layer, allowing the Free Software community to create a generic 3D engine would be a death sentence.
Forget F/OSS drivers or specs Posted Jul 25, 2006 21:26 UTC (Tue) by drag (subscriber, #31333) [Link] This guy seems to have some insights:http://voodoopc.blogspot.com/
His conjecture is that the ATI and AMD merger will remove the 'cut throat' style compitition between Nvidia and ATI.
This sort of 'cut throat' activity leads to a lot of buggy hardware and paranoid execuatives (which directly lead to closing off and going ultra-propriatory).
Now AMD that controls ATI Nvidia will have to work very close with it's previously biggest competitor. Whatever they do will have to be mutually benificial. If AMD and Nvidia fights then it will mean certain doom for Nvidia and AMD will loose a valuable business partner that helps make them lots and lots of money in the server and desktop markets.
So if anything it should foster more openness and cooperation.
Also the idea of integrating GPU technology directly into the silicon shared with the CPU is interesting. If that happens I don't see ANY point in keeping 3d acceleration programming interfaces a secret. How is requiring binary-only drivers for your CPU going to work out? I don't see how it does and Linux is valuable to AMD also. This is aiming for 2008 for this technology AMD or at least it seems that way.
However I still don't see AMD automaticly openning up documentation for ATI video cards. There is a lot of tech in those cards that ATI doesn't own. For instance somebody stated above that macromedia protection BS. Also with OpenGL there are a decent number of software patents you have to deal with. In order to support open source drivers you would have to forgo some binary compatability and loose a certain amount of performance with games and applications that were previously programmed to use the propriatory drivers.
Forget F/OSS drivers or specs Posted Jul 26, 2006 2:16 UTC (Wed) by wrobbie (subscriber, #39320) [Link] "Almost every hardware component in a modern desktop PC has good Free drivers but no video vendor (excepting Intel) wants to release 3D specs. Seen lots of theories float, no idea which might be true."
Hmm, I keep hearing that these days.. So, where can I download the programmer's reference for say a 855 or 915 chipset? The Intel site only offers more global GMCH specs. Or are these available only under NDA to selected parties? For the old 815 a good set of free documents is available.
Cheers!
Rob
Forget F/OSS drivers or specs Posted Jul 26, 2006 6:06 UTC (Wed) by drag (subscriber, #31333) [Link] Ya Intel is paying Tungsten Graphics to create open source drivers. Documentation is provided under NDA.
They do release specs for most other stuff though.
AMD is much better at it then Intel is. Intel is so-so. They do a good job providing good support for Linux on most stuff, but then they do stupid things like that binary daemon for controlling their new wifi card and making their firmware a download-from-intel-only affair.
I think that Intel wants to have good support for Linux and agree with Greg KH that binary drivers are illegal. Also binary only drivers are nearly impossible to support under Linux, by design.
Work for me, though. Intel isn't doing this out of the kindness of their heart, and I didn't buy a Asus motherboard with their chipset and cpu out of the kindness of mine.
Forget F/OSS drivers or specs Posted Jul 27, 2006 19:34 UTC (Thu) by shane (subscriber, #3335) [Link] Hmm, I keep hearing that these days.. So, where can I download the programmer's reference for say a 855 or 915 chipset?Did you even try Google? A search for "intel 915 programmer specifications" brought me immediately to this URL: http://www.intel.com/design/chipsets/915g/documentation915g.htm There is a 426 page PDF that describes the registers, memory, operation modes, and so on in complete detail.
Forget F/OSS drivers or specs Posted Jul 28, 2006 1:06 UTC (Fri) by wrobbie (subscriber, #39320) [Link] "There is a 426 page PDF that describes the registers, memory, operation modes, and so on in complete detail."
That is not the programmer's reference manual, but a much less detailed datasheet. We were talking graphics drivers -- if you look at the relevant sections in that manual you'll see there is no information at all. The register descriptions you speak of are mostly PCI config space registers etc.
Compare that to the 815's programmer's reference manual.
Cheers,
Rob
Not holding my breath Posted Jul 25, 2006 17:36 UTC (Tue) by freggy (guest, #37477) [Link] Reading blog messages such as http://airlied.livejournal.com/31180.html , I won't hold my breath for this to happen...
Not holding my breath Posted Jul 27, 2006 14:33 UTC (Thu) by cventers (subscriber, #31465) [Link] Do you realize how long acquisitions take? The AMD/ATI merger was justformally announced. AMD's top priority is not to accelerate Dave's NDA code review.
It will be some time before AMD and ATI begin to really integrate. But
Don't forget media center PCs Posted Jul 26, 2006 15:01 UTC (Wed) by broom (guest, #2914) [Link]
We also need a decent open-source driver for media-center PCs (so you can do xvmc in the graphics card for instance) without needing a binary blob in your kernel.
I'd say there are lots of possibilities for innovation in the media center environment if that were to happen, and whoever releases / makes possible good 3D drivers first will have a major advantage (until the other side(s) follow suit).
|
|||||||||||||