AMD is strongly considering open-sourcing at least a functional subset of ATI’s graphics drivers. It’s time for X Window System, OpenGL, and client virtualization for which ATI binary drivers aren’t available to escape the ghetto of the 1980s-era framebuffer. And what a boon for PR. If AMD’s graphics cards were the only ones with open device drivers, it might affect a buying decision or two.
I’ll wait till I hear an announcement from AMD themselves before I get excited.
Even though I personally prefer nVidia products, ATI opening their drivers could lead to nVidia opening their drivers as well. Gotta keep up with the competition right?
I think we can drown almost any hope right away. As ATI (and nVidia as well) always said, there is patented and licensed code from other parties in the drivers. I suppose that for example antialias and texture filtering algorithms are patented and thus will not be possible to be opensourced withot the consent of the patent owner, which is not necessarily ATI itself.
I cannot see what AMD could possibly do to overcome these legal issues.
But if hell freezes over it sure would be great for the OSS community.
I suppose that for example antialias and texture filtering algorithms are patented and thus will not be possible to be opensourced withot the consent of the patent owner, which is not necessarily ATI itself.
That never stopped Intel or Xgi.
An integrated graphics chipset doesn’t come close to the latest Radeon or Quatro as regards specialist techniques. Remember the patent dispute over Jhon Carmack’s self-shading process in Doom 3?
An integrated graphics chipset doesn’t come close to the latest Radeon or Quatro as regards specialist techniques.
Xgi.
Indeed. The opengl performance on most integrated cards are pretty bad. Many of Intel’s integrated cards aren’t too bad. Even the older onces. But they just don’t cut it compaired to a good GPU from ATI or Nvidia.
The patent dispute of Carmacks shading was truly disapointing. From what I could tell, ID didn’t fight it becuase they felt that a judge just wouldn’t be able to tell the difference between what their shader did and what Creative had patented (which were pretty similar).
Well, and that patent shouldn’t have been granted.
Anyway, if AMD opens the Ati drivers, that will certainly push me towards getting an ATI card (or two) when I next upgrade. Common AMD, don’t dissapoint!
Patents have NOTHING to do with the code; patents are merely a set of proceedures to get that done, it is up to the said software company to implement it in a way of their choosing.
There are opensource things that are patented, and yet, there hasn’t been any problems; the problem occurs when an opensource projects implements a feature which is patented and it hasn’t been authorised.
As for the patent holder of these said things from Ati; the patent holder doesn’t give a shit, as long as their organisation is getting cash, quite frankly, thats all that matters.
The real reason for Ati not opensourcing their drivers? its politics, thats all it is; the same shit that is occuring in Sun over Java; I’ll put money on it, Ati is split into two areas; the hardware engineers who are quite happy about the idea of the specifications being made available and drivers opensourced, and the other side, the opposing party, the software developers would be pissed off that some of them will lose their jobs in the process; I mean, when you’ve got volunteers, OEM’s and the likes fixing bugs in the drivers, you’ll no longer need the huge number of software people, so the amount will shrink.
Quite frankly, AMD/ATI should see opensourcing the drivers as a cost cutting measure to boost the bottom line rather than it being a fictional ‘lose of competitive edge’ as Ati liked to bullshit about.
I suppose that for example antialias and texture filtering algorithms are patented and thus will not be possible to be opensourced withot the consent of the patent owner, which is not necessarily ATI itself.
I cannot see what AMD could possibly do to overcome these legal issues.
Are these algorithms implemented in hardware or in the software driver? I would have guessed that it would be hardware. Hopefully we might find out soon.
I agree on the announcement thing. The rumour mill has been working overtime ever since the announcement of a possible AMD/ATI merger.
I reaqlly like both nVidia and ATI products. I have been favouring nVidia recently because their closed Linux drivers are better than ATI’s (and both are better than the Open Source ones simply because they can talk to the hardware properly). However, if AMD/ATi release their linux drivers as Open Source (and developers start improving them, which will probably start happening less than a day, if not less than an hour after the source is released), I’ll definitely switch over, at least until Open Source nVidia drivers are on par or better than the Open Source ATI ones.
Wow, take that nVidia!
If this actually happened it would be great for the entire open source movement, en maybe especially for OpenBSD, FreeBSD and alike.
If AMD/Ati open sourced their drivers i would see no other solution that nVidia having to open source them too, especially with the no closed source drivers policy that most of the big distros have now.
It suddenly makes PPC Linux much more viable. If this happens, one of my pet peeves with PPC linux is removed. Good going, AMD/ATI!
This sort of thing definitely affects my buying decisions. It open choices up as a machine progresses into its senior years ….
Same here… I’m an avid AMD supporter and will not be buying nVidia (although I’m an avid nVidia supporter too), but I have a feeling nothing will work as well as an all AMD machine
Well
My machine may be saying hello to AMD Graphics cards in the comming months/year then
With Novell recent decision to only use Open Source drivers this will be a necessary step to keep SLED, SLES, SUSE and OpenSUSE great.
AMD is a very good company and such and initial step right after the acquisition of ATI would be very telling.
Just like with Java, they are trying to save AND eat the cookie. “Functional subset” is a very ambiguous term. I believe it… when I see it!!!111sin(pi/2)111!!11 Until then; NO COOKIE!
According to OpenBSD developers, all that is truly needed (or wanted, really) is accurate and complete hardware documentation. With such documentation available, it would be possible to create fully functional free drivers without violating anyone’s code copyrights or (probably) patents.
In 30 years of OS development I’ve never once seen “accurate and complete” hardware documentation for any part more complex than a UART.
Well, except for the VHDL for the part.
ATi documentation is definately A++-quality. I have the hardware documentation for all chips up to the R200.
nVidia already tried that with their initial TNT2 drivers, but a couple of cease and desist letters from Microsoft and Intel [1,2] soon forced them to close the drivers.
-Unless- ATI did a clean room implantation of the OpenGL parts that ‘belong’ (…) to Microsoft and the PCI/AGP parts that ‘belong’ (yet again) to Intel, we are screwed…
Gilboa
[1] http://lists.suse.com/archive/suse-linux-e/2006-May/3373.html
[2] http://marc.theaimsgroup.com/?l=dri-devel&m=114981283225530&w=2
Great thread you pointed out! Thanks!
But most people here didn’t read it, judging from the comments.
You mean that there is a possibility, however so slight, that running fglrx drivers and then switching to a console might not actually crash my machine? Oh that would be nice.
From a Linux users perspective, dropping the name ATi is a good idea. Do a search for ati and linux and you will walk away with the (generally true) impression that their drivers are substandard and buggy. AMD, if they see the anti-ATi rants, would also see the “I’m only buying nVidia from now on” posts which normally follow them. I’m sure that they would prefer not to have a “AMD is junk, buy Intel/nVidia” resounding through messageboards.
Call me a pessimist, but I don’t really see it happening. While open source drivers would be nice, I don’t think it would be that beneficial to AMD. And while I’m sure they could find a way if they try, ATI never open sourced their drivers because they were using various algorithms that others had patents on and couldn’t be open sourced.
That and the author doesn’t give any definite source from AMD justifying that they are indeed trying to open source the drivers. Like the AMD ditching the ATI name, it seems likely that it is going to be just another rumor.
http://arstechnica.com/news.ars/post/20060808-7446.html
XGI, first of all, was bought out by ATI a few months ago.
Secondly, XGI’s 3D driver was a binary blob. Only the 2D X.org driver was open sourced:
http://www.thejemreport.com/mambo/content/view/173/81/
As the article points out, the other graphics manufacturers already release or support 2D driver code, so XGI didn’t really do anything special at all.
This would be great! Count me as one customer they would gain. I could finally have a nice card in all my boxes, until then integrated will do fine.
I work with and advise a few system builders in Spain which are going to put out Linux-based systems.
If AMD/ATI were to open source their drivers, they would be the default choice in thousands of systems. I hope AMD follows through with this as it would make it so easy for us to choose as our supplier.
Wait and see, I guess.
Edited 2006-08-09 22:06
…AMD and the power of the GPU? Can anyone forsee any possible integration of the GPU onboard AMD motherboards with chipsets utilising and sharing the power of the GPU for other uses other than only graphics processing? DOn’t you think that (depending) on what they apply the processing power to – will they have the advantage over Intel? Maybe Virtualisation/Database applications? The opensourcing of the drivers may be at the bottom of their list.
While we can spin our wheels a speculate about open source AMD/ATi graphics drivers, AMD has made it very clear that they intend to do exactly what you’re proposing. Starting in 2007, the will have a coherent HT bus for 3rd-party coprocessors. Expect this interface to replace PCIe for graphics on AMD systems during 2008.
Around this time, AMD hopes that a modular chip design will allow their partners to develop on-die coprocessors for high volume parts. This was announced a few weeks ahead of the ATi acquisition, so when they said “partner,” they meant “the company we’re about to buy.” This timeline suggests a general availability of an AMD MPU (multimedia processing unit, or CPU + GPU) in 2009, just in time for inclusion in the next cycle of gaming consoles.
It was easy to see this coming. What kind of 3rd-party coprocessor has enough volume to justify a new lithographic mask and everything? The only kind that already ships will every single computer and console in the world, the GPU.
hum
Maybe Nvidia and ATI(AMD) will take note!
http://news.com.com/Intel+aims+for+open-source+graphics+advantage/2…
Intel’s always been pretty good about keeping their drivers open.
It does state that SOME of the code may be opensource, this doesn’t imply the whole kitten kaboodle. But it could be enough to boost the base performance in the kernel tree.
We just had an argument over binary drivers in the core tree, and i think the biggest concern when people talk about closed drivers is the ATI+NVidia ones.. Sure there are come wifi drivers and maybe some modem drivers that arent OSS, but the graphics drivers are the big showstoppers.
If any part of this turns out to be true AND they actually follow through with it. I think NVidia would be put in a very sticky situation. I for one have always been a nvidia supporter mainly because their linux drivers were off such better quality. The opening of any part of the code can do nothing but benefit AMD in the long haul.
Give me FGLRX quality driver in the core tree, yup my next card would probably be a ATI, oh escuse me AMD 😉
When I was choosing my laptop, I wanted to have fully working, open source drivers for my hardware. So I bought one with Intel 915 graphics card. But AMD’s chipsets are faster and if only I had opportunity to have drivers for it running out of the box on every linux distro I can imagine, I would by one without a doubt. In fact s I _will_ buy AMD/ATI graphics card if they open source their drivers. The market will choose… but _we_ are the part of the market – remember that AMD!
I have fully working, open source drivers for the graphics chip in my laptop, and it’s an ATI. OK, so it’s not the most up to date (Mobility 9000), but it’s good enough for 3d games like Torcs.
When I purchase, or build computers, I have gone with Intel for the low end stuff, and Nvidia when I want extra performance.
As a Linux user, I haven’t purchased ATI in four years. I have two very old ATI machines left. They run with Xorg’s Radeon driver. The Intels run with the Xorg i810 driver. The Nvidias require Nvidia’s proprietary driver. Each time I change the kernel, I rebuild the Nvidia kernel module that hooks the driver into the kernel. It’s not hard, but it is a hassle.
I have chosen Nvidia over ATI, because of on-line reputation. Nvidia has a good one; ATI has… Well, let’s just say it hasn’t.
If ATI open sources key elements of its drivers, that would change my purchasing equation. Like others have mentioned, open source drivers are important to me. I use them when I can.
If done correctly, I hope that community input would fix many of the bugs that continue to afflict the ATI drivers.
May this news be true!
It’s Christmas!
My hopes are high, my believes are more realistic. A good OpenGL implementation for XGL would do the trick for me.
And if AMD does it I will select AMD for computers of people asking my assistance when buying a new computer.
Six points about why open sourcing of linux driver would be beneficial to Linux:
1. Many new cards supported out of the box. Others as simply as typing apt-gwt upgrade. No more flirting with out of tree kernel and similar userspace binary blobs.
2.Integration into the existing DRM/DRI/Mesa. Although ATi drivers a based on DRI, they differ. This effort would bring together implementations of intel, ATI (and older cards) working on the same framework and same Mesa GL library (vendor specific extensions can still be possible).
3. New people working on it. AMD could hire some people to work with DRI folks on the mailing list. Quality and feature support, not only of the ATI driver, but of the whole DRI/DRM and MESA framework would substantially improve (e.g. OpenGL 2.0 support, scheduling and memory management, etc).
4. Accelerated X development – certainly the whole Xorg project is currently held back by lack of proper documentation for some important chips. AIGLX, XGL, XvMC, etc. might improve and become supported out-of-box on each ATI chip. More people will be attracted to work in X.org if manufacturer is willing to cooperate.
5. Nvidia might follow. Probably not by open sourcing the whole framework, but maybe at least the kernel part (except if the whole damn thing is in kernel). They coould also use existing libGL with dispatching support instead of overwriting their own.
6. One big step closer to a Linux as mainstream gaming platform (but not even near). Still, critical mass of users is not reached to begin converting those loyal to Microsoft and DirectX (as DX10 is coming, it will be even harder).
This is not just a Linux thing – all non Windows platforms will benefit.
And no – nVidia will never in a million years follow.
I hear all these people talking about switching to AMD/ATI if they open-source their drivers. My question, why? Its great to preach the OSS mantra, but what does it really mean? All I care is that I can use all the features of the video card I buy without problems. Closed or open drivers, does it really matter if they work? I have used nvidia exclusively for 4 years with Linux and never had any issues. So why would I considering switching just because another company provides OSS drivers? Does this mean the drivers will be better? Probably not, but I’m sure some will argue errors can be fixed by more people, or mistakes spotted, blah blah blah. Sounds good but does it really change much. (Might for ATI users, but not for nvidia users)
Comparison to MS not opening specs is different as in their case they use their lock in formula to make sure you are forced to buy their other products. Video cards drivers are a different story.
“Closed or open drivers, does it really matter if they work? I have used nvidia exclusively for 4 years with Linux and never had any issues.”
Then consider yourself lucky. Many of us have. Meanwhile, we’ve had amazing results with bundled open drivers distruted with X, the kernel, and/or our distro of choice.
“Sounds good but does it really change much”
Yes, it does. nVidia has definitely had the best closed drivers, BTW, and I’ve even had trouble with them. RAID controllers that ‘support Linux’ w/o open drivers can be hell. Typically, you’ve not got hundreds of people working on them, but only one or two. But these one or two are talented, and actually fix problems people find. Also, it’s fairly easy to keep an open driver up to date w/ kernel changes.
Also techniques found in different drivers can be ‘merged’ to one, if all drivers uses the same routines to draw something on the screen. When developers has for example three drivers, they can make one universal driver, that works with lot’s of various devices. Opensourcing drivers means more supported hardware with less pain and bytes needed.
BTW that can make big changes in all Linux graphics system. By modifying drivers developers can fix them to meet new technologies or simply optimize for use with current ones. For now XGL and other good stuff is just in strange state. If XGL now would be finished and mature, it couuld’t be installed and started by default, because flawless operation will occur not with all hardware. And this is bit stopping such projects to evolve. Some developers may be even unsure, is all their work worth it.
Personally i hope for those open drivers, but there’s so much secrets and other marketing mambo-jumbo involved. Nvdidia for example uses the same chips and processors for “slower and cheaper” and “faster and more expensive” cards. So actually when you buy cheap card, you actually get that ‘more expensive’ one. All slowing down is done by drivers. There are patched drivers, that dramatically increases performance of sone Nvidia cards. I don’t know how ATI deals, but if Nvidia woud decide to go opensource, those drivers should be the ‘most’ slowest for all chips. Without that someone can say, hey, my linux games runs faster than Windows!
Edited 2006-08-10 03:32
IMOa nVidia/ATi aught to work to make their next gen open. So what if there’s hidden stuff? Stick a hardware layer of abstraction in there, and let devs not be able to see the protected parts.
I don’t care about patents and trade secrets–I care about it working. And if one compnay has open specs and there are good drivers, I’ll take it, and recommend it. I now only recommend Ralink-based wireless cards, FI; and spent most of last night trying to get a linksys one working, largely due to the laptop in question having only 1 USB port (I have a USB Ralink-based one, too).
“All slowing down is done by drivers. There are patched drivers, that dramatically increases performance of sone Nvidia cards.”
The performance differences are not dramatic, except in cases where pipelines have been turned off. There are many of those chips where that has been done in which those pipelines are bad. Some of them are to keep prices in check, just like Semprons and Celerons, and those are the ones you hope to get. But the drivers are not doing any extra slowing down. The software mods are enabling things turned off which could be potentially dangerous.
The above, of course, is null for Geforce v. Quadro, in which case it’s all about keeping prices high.
“Without that someone can say, hey, my linux games runs faster than Windows!
”
I say it can happen! It has definitely occured in the past for RAID controllers, and CPU use of sound cards is very low on Linux. While not too many people typically work on any driver, it seems those that do tend to be extremely talented and creative; and I am also of the opinion that the social environment of open source brings out excellence, if just because there are not middle managers getting in the way
. Also, knowing how the drivers work (open source and all that), games themselves could be tweaked for better interaction with X and the driver.
[quote]”Closed or open drivers, does it really matter if they work? I have used nvidia exclusively for 4 years with Linux and never had any issues.”
Then consider yourself lucky. Many of us have. Meanwhile, we’ve had amazing results with bundled open drivers distruted with X, the kernel, and/or our distro of choice.[/quote]
Nothing is perfect but for the majority of linux users the closed drivers from nvidia have been excellent and almost on par with windows for all features.
You make the assumption that a change to OSS means there will be no more problems. Please. How many OSS projects, take Xorg/XFree86 for example, have had bugs for years that never got fixed? Although it sounds great and appears everything improves with open-sourcing, drivers could take a step back for a while. Depending on how they handle things you could have different distros including different patches/fixes to the drivers leaving a fractured set of users with everyone using a slightly different driver. This could cause many issues and don’t assume every fix someone comes up with will immediately or ever mean it will be universally accepted into the main trunk.
“You make the assumption that a change to OSS means there will be no more problems.”
No, I make the assumption that a change to OSS will mean few problems as time goes on.
“Please. How many OSS projects, take Xorg/XFree86 for example, have had bugs for years that never got fixed? Although it sounds great and appears everything improves with open-sourcing, drivers could take a step back for a while.”
I’m not disputing that. I spent hours two days ago trying to get a Linksys wireless card working, because the Broadcom driver loads but doesn’t work right, and screws up ndiswrapper!
However, it will get worked out. It has before, and it will again, if the specs are open, so that things can really be fixed, rather than shots in the dark. I don’t think temporary fragmenting is a problem, as long as it eventually goes back.
I doubt different drivers would exist for too long, but that would certainly be a short-term problem, possibly lasting as long as a year.
“This could cause many issues and don’t assume every fix someone comes up with will immediately or ever mean it will be universally accepted into the main trunk.”
That would be a problem.
This was not a pro-open source nor pro-linux post, therefore your negative score.
— The moderation system
I have always liked AMD’s symbiotic relationship with the open source community … that’s why I will always purchase AMD workstations & servers
AMD will open source what they legally can .. However, why couldnt they expose the interface to the closed source components, which can then be ‘filled in’ by open source developers?
I would buy an ATI product ever if it’s performance is lower, if it open sources it’s drivers. Take Samsung for example it gives it’s Linux driver in the installation cd it self.
This could be a HUGE win for the state of the FLOSS desktop. Call ATI and AMD to tell them this is serious and you care about open source. Do NOT be rude but politely tell them why you consider it important to open source video drivers and that you will buy their products. Sales people will listen to potential customers and that is you!
http://www.amd.com/us-en/Processors/TechnicalResources/0,,30_182_35…
http://www.ati.com/companyinfo/contact/consales.html
Great news, Imagine Cairo hardware accelerated desktop from the box! (pdf, htmp engines, ooo). xorg list everyday messages such as “patch to speed up Doom4/WoW/etc 20% benchmark here…” or doom9.net “xvid/x264 compression using GPU”. Go, AMD!
i like ATI and nVidia hardware equally.
however, i always buy nVidia because of the superior linux support.
this would change overnight if ATI decided to open-source large parts of their driver.
if degree of openness was enough to permit their inclusion in SUSE et-al then it would be a done deal, my nVidia buying days would be over.
until nVidia are forced to follow suit and then it would become a matter of driver quality again.
“suppose that for example antialias and texture filtering algorithms are patented and thus will not be possible to be opensourced withot the consent of the patent owner, which is not necessarily ATI itself.”
Since the idea of these card is hardware acceleration I’ll go out an a limb and expect such things to be implemented in, uh, the hardware and thus you wouldn’t need to know anything about the implementation to use the API.
Abstraction isn’t a new concept.
…make this happen! This move would be good for you! They say any publicity is good publicity, but surely good publicity is even better! (And of course it would be good for us!)
Another issue with the closed drivers is that they are x86 only, opening them up could lead to a port to power PC and other platforms which would be really nice.
I believe that this announcement is probably in response to Intel’s announcement yesterday (or perhaps Tuesday) that they will be providing open-source drivers for their current and upcoming graphics chipsets under the MIT and GPL licenses.
I’d be cautious about the ATi announcement. In part because the widely held understanding as to why their drivers were closed-source in the first place was that there may be evidence of patent or even copyright infringement in the code. A lot of people seem to think that graphics chipset/card vendors are ripping off bits of each others technology — an “everybody does it” sort of situation. If one vendor exposes all, but the rest don’t — well, guess who’s going to court.
If graphics card vendors could come up with a consortium that agrees to technology cross-licensing for some of the basics, then you’d see open-source drivers all over (frankly, the vendors would just assume contribute and not develop everything themselves because doing that is tedious and expensive).
Great news……Now eveyone should email AMD directly and let them know that the community supports them and it will inturn mean more money in their coffers if they do indeed make this decision