OSS Drivers Changing Your Next Purchase?

[Michael]

Intel's open-source drivers now support the 965G Express and it looks like they are working on improving their open-source drivers quite a bit. AMD/ATI has yet to publicly define what they may be opening up with their fglrx drivers once the merger is complete.

However, will this shake-up in drivers cause you to rethink your next upgrade/purchase due to the open-source status?

[Rob Williams]

No, because NVIDIA has me locked up in a cage. While I wish NVIDIA had OSS drivers, I enjoy the reliability and performance too much to go elsewhere.

[Patrick]

Until the performance on ATI/Linux can match NV/Linux I probably wouldn't switch, regardless of if it's open source. Although if ATI does open up part of their drivers will NVIDIA follow suit? Besides gaming though, Intel's open-source drivers are great for people that buy pre-built computers from large vendors.

[Svartalf]

The theoretical performance of ATI versus NVidia is actually VERY close to each other. ATI's being broken upon the wheel against NVidia by the differences in the drivers- this could be just software architechture problems (You can go about doing things needed in a 3D card in several different ways, but only a couple of them are ideal under a given architechture- ATI's answers for Windows may not mesh as nicely as NVidia's under Linux...) or a hardware architechture issue where the interface doesn't allow for a "nice" hook into the OS in question.

I suspect software problems, myself. Having seen all too many hardware vendors trying to take a 12# sledgehammer to the Square Peg of a Windows driver solution to the much smaller round hole of the Linux driver problem, some of it to avoid giving out details, some of it because they don't want "to duplicate work with different code paths", thinking that it'll save them time to just re-use the Windows answer no matter what- that I doubt it is anything other than that. Open Sourcing the drivers will allow someone to FIX that problem and bring the performance to where it should be. There is no good reason for the why my Express 200M laptop is roughly half the speed of Windows' drivers other than trying to wedge a Windows solution to a Linux problem in the drivers.

[Svartalf]

Now, as to my commentary on whether this will impact my purchasing decisions- YES. As it stands, I plan on getting an i965G based motherboard, something I would not normally do since I'm more of a fan of AMD based solutions. Part of it is for testing reasons- I'm dead sure there'll be a lot of these machines on the market shortly; if the performance is in the same general ballpark as the mid-end ATI and NVidia integrated offerings currently are, the machine will be able to play many if not most of the modern games at least in a mediocre fashion. With it being largely open source, I'm also pretty sure that unless it's a total stinker, a lot of Linux users and developers will buy it just for that reason.

That, folks, is my current customer base for the stuff I've been doing off and on for the past 3+ years now. ATI's drivers don't do this for me except at the top-end and the performance is positively lackluster right at the moment. NVidia's got my attention only because they work as well as their Windows counterparts do on the driver front. If I had a compelling (within 5-10% performance difference either way of NVidia's) vendor step up to the plate with Open Sourced drivers, they'd have my attention just like Intel has right at the moment.

[Andrew]

I'm waiting for a day when both the Nvidia and ATI are open source and comply with GPL, meaning they'll be included with your favourite distro rather than having to download it post install to get your games running right (and overclocking if that takes your fancy).

[drag]

Now, as to my commentary on whether this will impact my purchasing decisions- YES. As it stands, I plan on getting an i965G based motherboard, something I would not normally do since I'm more of a fan of AMD based solutions. Part of it is for testing reasons- I'm dead sure there'll be a lot of these machines on the market shortly; if the performance is in the same general ballpark as the mid-end ATI and NVidia integrated offerings currently are, the machine will be able to play many if not most of the modern games at least in a mediocre fashion. With it being largely open source, I'm also pretty sure that unless it's a total stinker, a lot of Linux users and developers will buy it just for that reason.

I bought a Intel-based motherboard and Intel-based CPU for the first time _ever_. Before it was always AMD and Via..

But now all you find is Nvidia or ATI motherboards. New Nvidia motherboards suck because hardware support is crap and what you do have is in the form of these horrific binary drivers. WTF would I want a binary-only driver for ethernet or sound? There are about a billion other devices with great open source support. I went on the Internet and saw people in forums all over struggling to get Linux working well on these things.

And ATI is worse.

Via seems to have fallen behind and I couldn't realy find a motherboard they used that I liked. They weren't realy offering any thing I realy liked.

So I bought a Intel motherboard with onboard gigabit ethernet, sound, and video and installed Debian testing on it. Sound worked out of the box. Ethernet worked out of the box. 2D acceleration worked out of the box. 3d acceleration worked out of the box (and with tweaking it worked well).

Although I would much prefer AMD and think that Intel isn't realy all that great I won't be going back unless that ATI/AMD merger produces devices with good open source driver support.

[1c3d0g]

Good luck with that one. Knowing ATI's crappy driver support, I wouldn't touch them with a ten foot pole.

[niniendowarrior]

Well, opensource drivers don't normally influence me of my next purchase. Competent drivers do influence my decisions. So if the Intel 965G can outpace my X800 Pro 256... then I'd gladly buy one mobo from Intel.

Just a note... has anyone actually tried to Opensource drivers for ATi cards pre-9200? 3D hardware acceleration is working out of the box, I believe.

[Michael]

Just a note... has anyone actually tried to Opensource drivers for ATi cards pre-9200? 3D hardware acceleration is working out of the box, I believe.

For pre-9200 hardware, I believe David (of Phoronix) uses some Radeon 7500's and/or 8500's with the open-source drivers. Personally I haven't touched any pre-9200 stuff in a long while.

[drag]

For pre-9200 hardware, I believe David (of Phoronix) uses some Radeon 7500's and/or 8500's with the open-source drivers. Personally I haven't touched any pre-9200 stuff in a long while.

Out of the R200 era cards (8500-9250 I beleive) the 8500 should be the fastest. Faster then the 9200. It had higher clocks on memory and gpu if I remember correctly. Back in the day 'The Weather Channel' actually paid to help get 3d drivers for these things.

edit:
http://en.wikipedia.org/wiki/Radeon_R200
The 8500 had 275mhz clock on Memory and GPU. The 9200 had 240mhz clock and 200mhz ram. Also it had a more complex pipeline arrangement (4/2:2 for the 8500 vs 4/1:1 for the 9200.. whatever that means) and the 128meg version had it's memory interweaved which probably did a slight performance boost.)


With 7500 I donno. I expect their would be 3d acceleration from Free drivers, but it's probably been a long time since anybody tried to use them for gaming so I don't know of the stability.

ATI was somewhat helpfull with 3d drivers in Linux up until the time right around they won the contract to start developing a GPU for the Xbox 360. It is probably a coincidence, but maybe not. It's the same time frame anyways. Now a X developer has made 2D drivers for the r500 series cards and ATI has him under NDA because of what he does for a living and they are refusing to let him release code for those, even though they are very basic drivers.

It's probably just normal corporate BS. Everybody is too chicken to actually say 'yes' to anything.

Well, opensource drivers don't normally influence me of my next purchase. Competent drivers do influence my decisions. So if the Intel 965G can outpace my X800 Pro 256... then I'd gladly buy one mobo from Intel.

That would take magic, not extra special drivers. There is no way in hell that the 965G X3000 will come close to the ATI X800 with both having decent drivers.
(anyways.. If it was raw speed your after both your ATI card and the majority of your games would run much better under Windows, so why bother with Linux? (we both know it's more then just speed))

I am doubting that the X3000 will even been that much better then the GMA 950 at this point. It'll probably help in those games that take advantage of the extra hardware features that X3000, but otherwise I'd bet it's a incremental increase in performance, not a huge one.

[Michael]

ATI was somewhat helpfull with 3d drivers in Linux up until the time right around they won the contract to start developing a GPU for the Xbox 360. It is probably a coincidence, but maybe not. It's the same time frame anyways. Now a X developer has made 2D drivers for the r500 series cards and ATI has him under NDA because of what he does for a living and they are refusing to let him release code for those, even though they are very basic drivers.

David was under NDA with R200/300/400 generation GPUs -- not R500. By running the program that he did on an R500 component, he technically violated his NDA. Thus since he was not authorized to do so, ATI will not permit the release of his code at this time.

That there is the basic answer.

[drag]

David was under NDA with R200/300/400 generation GPUs -- not R500. By running the program that he did on an R500 component, he technically violated his NDA. Thus since he was not authorized to do so, ATI will not permit the release of his code at this time.

That there is the basic answer.

It's a possible answer, but he still went through ATI to get permission. It's very likely that if he broke his NDA there would be more ramifactions then ATI simply ignoring his request with no explaination. They didn't tell him no or yes or slap him on the hand or anything. Doesn't sound like it to me that he did anything wrong. He is just being ignored.

If he violated the NDA I would expect _something_ to happen.

Anyways there is only 2 groups of people who would know about weither or not he violated his NDA and that would be either ATI or Airle. Did you talk to either one of them about it?
;)

[Michael]

It's a possible answer, but he still went through ATI to get permission. It's very likely that if he broke his NDA there would be more ramifactions then ATI simply ignoring his request with no explaination. They didn't tell him no or yes or slap him on the hand or anything. Doesn't sound like it to me that he did anything wrong. He is just being ignored.

If he violated the NDA I would expect _something_ to happen.

Anyways there is only 2 groups of people who would know about weither or not he violated his NDA and that would be either ATI or Airle. Did you talk to either one of them about it?
;)

Yes, actually I had briefly discussed the situation with a representative from ATI after David's blog post.

[drag]

That's too bad. I was hoping that someday I'd be able to purchase a r500 card.

But unless ATI cleans up it's act, it's just not going to happen. The only reason I have the ATI cards I have is because I can get 3d going without binary-only modules. Ultimately if I am forced to choose between to bads then Nvidia + propriatory drivers is far far superior then ATI + propriatory drivers. There is just no contest.

If AMD-ATI thing doesn't work out well and Intel doesn't produce a decent card.. then it's Nvidia or bust. Too bad, ATI stuff is generally pretty good.

[halfmanhalfamazing]

Just a note... has anyone actually tried to Opensource drivers for ATi cards pre-9200? 3D hardware acceleration is working out of the box, I believe.

Yes, I currently own an ati 9000 128MB pro, a 9200 128mb, and in my main computer I have a firegl 8800 128MB. I have these cards specifically because of the OSS r200 drivers.

Yes, I know they are written based off of partial specs. They work very well IMHO. And being as I'm using an FGL 8800, it's pretty fast. And it's great that everything works "out of the box".

Having OSS drivers influences my purchase, I'll probably have intel next time around.

It'd be nice if Mike would pick something like the FGL8800 to use as a reference due to it's OSS drivers and run with it in all benches.(for comparison purposes) That is, until something new comes out with OSS drivers. Then that would become the reference for comparison. But even if he doesn't, I still really like this website.

[drag]

Just so you know cards that do have open source drivers that are usable for games and stuff like GL desktops.

For the R200 DRI drivers...
ATI r200 series
http://en.wikipedia.org/wiki/Radeon_R200
8500-9250 ATI cards
For the consumer cards the 'ATI 8500' with 128 megs of RAM would be the fastest.
These are probably the best supported ATI cards. Maybe better supported then the Intel stuff.
Supported well out of the box.


For the R300 DRI drivers...
ATI r300 series
http://en.wikipedia.org/wiki/R300
and
ATI r400 (or r420) series
http://en.wikipedia.org/wiki/X800
Probably would want to aim for the X800
They provide decent speed, work well with AIGLX, but feature support isn't all there.
I have to use CVS versions of the DRM (Linux kernel portion), DRI, and Mesa for best stability and performance.
http://dri.freedesktop.org/wiki/Building


The 915 DRI drivers...
Intel GMA 900 and GMA 950 IGP.
http://en.wikipedia.org/wiki/Intel_GMA
They are featurefull (for the limited hardware), stable, and supported out of the box on recent distros. (anything using Xorg 7.1 especially).

There are a few variations. The GMA 900 comes with 915 chipsets and comes in a veriaty of core speeds from 166 for the very low voltage chipsets to 333mhz.

The best performing currently is the GMA 950 IGP chipset which comes in two variations.. The regular 400mhz and then a 200mhz one for very low voltage chipsets.


Then there is the GMA X3000 and the GMA 3000 which are a unknown quantity right now.




For 'real' gaming, obviously the best choice is going to go with Nvidia software and propriatory drivers. Best performance and best compatability with products like Cedega and propriatory games.

I'd be willing to buy a Nvidia card if I was still realy into FPS games, but now not so much..

But I will NEVER EVER buy a motherboard or audio card or network card or drive controller or ANYTHING with propriatory drivers. With Video cards you have little choice, You have to choose nvidia if you want best performance.. But for motherboards and everything else there are much much better selection of aviable hardware with some of the best stuff supported by Open source drivers. There is no point in having to put with crappy propriatory drivers if you don't have to.

[Michael]

Just so you know cards that do have open source drivers that are usable for games and stuff like GL desktops.

For the R200 DRI drivers...
ATI r200 series
http://en.wikipedia.org/wiki/Radeon_R200
8500-9250 ATI cards
For the consumer cards the 'ATI 8500' with 128 megs of RAM would be the fastest.
These are probably the best supported ATI cards. Maybe better supported then the Intel stuff.
Supported well out of the box.


For the R300 DRI drivers...
ATI r300 series
http://en.wikipedia.org/wiki/R300
and
ATI r400 (or r420) series
http://en.wikipedia.org/wiki/X800
Probably would want to aim for the X800
They provide decent speed, work well with AIGLX, but feature support isn't all there.
I have to use CVS versions of the DRM (Linux kernel portion), DRI, and Mesa for best stability and performance.
http://dri.freedesktop.org/wiki/Building


The 915 DRI drivers...
Intel GMA 900 and GMA 950 IGP.
http://en.wikipedia.org/wiki/Intel_GMA
They are featurefull (for the limited hardware), stable, and supported out of the box on recent distros. (anything using Xorg 7.1 especially).

There are a few variations. The GMA 900 comes with 915 chipsets and comes in a veriaty of core speeds from 166 for the very low voltage chipsets to 333mhz.

The best performing currently is the GMA 950 IGP chipset which comes in two variations.. The regular 400mhz and then a 200mhz one for very low voltage chipsets.


Then there is the GMA X3000 and the GMA 3000 which are a unknown quantity right now.




For 'real' gaming, obviously the best choice is going to go with Nvidia software and propriatory drivers. Best performance and best compatability with products like Cedega and propriatory games.

I'd be willing to buy a Nvidia card if I was still realy into FPS games, but now not so much..

But I will NEVER EVER buy a motherboard or audio card or network card or drive controller or ANYTHING with propriatory drivers. With Video cards you have little choice, You have to choose nvidia if you want best performance.. But for motherboards and everything else there are much much better selection of aviable hardware with some of the best stuff supported by Open source drivers. There is no point in having to put with crappy propriatory drivers if you don't have to.

New R300 Open / Closed driver tests were just completed by Phoronix. More information can be found @ http://www.phoronix.net/forums/showthread.php?p=1518#post1518 (Please discuss the information in the appropriate threads).

[Svartalf]

New R300 Open / Closed driver tests were just completed by Phoronix. More information can be found @ http://www.phoronix.net/forums/showthread.php?p=1518#post1518 (Please discuss the information in the appropriate threads).

Yeah, and the performance is dismal, for either the OpenSourced drivers or the ATI ones, though the ATI ones are much more performant. It remains to be seen how things play out, but I've got it on good authority that ATI's been listening- the silence has been for a reason on the drivers we DO have. I'm adopting a wait-and-see attitude (for several reasons, actually...) with them- but if they were to Open Source things right now or at least give us real technical data, I'd be convinced on the spot.

[Michael]

Yeah, and the performance is dismal, for either the OpenSourced drivers or the ATI ones, though the ATI ones are much more performant. It remains to be seen how things play out, but I've got it on good authority that ATI's been listening- the silence has been for a reason on the drivers we DO have. I'm adopting a wait-and-see attitude (for several reasons, actually...) with them- but if they were to Open Source things right now or at least give us real technical data, I'd be convinced on the spot.

Svartalf:

What silence are you referring to from ATI?

[Svartalf]

Svartalf:

What silence are you referring to from ATI?

Well, there's things like Sideport initially being a functionality supported in the Linux drivers and then it's pulled, never seemingly to return. And then there's the absolutely dismal performance of these chips compared to the Windows drivers (never mind the comparison between the theoretical performance on the spec sheets versus the Linux speed...). There's a raftload more of these things, but they know about them all, and don't seem to be interested (or lack the resources, more like...) in fixing the problems in a timely manner (I've had issues since I owned the ATI chipset based Athlon64 laptop, nearly a year ago today- issues that don't seem to be getting resolved...). That would be relative silence in many's books (incl. mine...).

But, as it's been indicated in my previous posting, I've got info from a pretty reliable source that this is about to change for the better. No, can't say any more than that- it'd probably get me in trouble, it would... :D

[Michael]

Well, there's things like Sideport initially being a functionality supported in the Linux drivers and then it's pulled, never seemingly to return. And then there's the absolutely dismal performance of these chips compared to the Windows drivers (never mind the comparison between the theoretical performance on the spec sheets versus the Linux speed...). There's a raftload more of these things, but they know about them all, and don't seem to be interested (or lack the resources, more like...) in fixing the problems in a timely manner (I've had issues since I owned the ATI chipset based Athlon64 laptop, nearly a year ago today- issues that don't seem to be getting resolved...). That would be relative silence in many's books (incl. mine...).

The lack of silence from corporations isn't anything ground shattering. NVIDIA is the same way, if not worse, as there isn't even a public unofficial BugZilla. The release cycle for NVIDIA can also be debated. ATI is very interested in correcting bugs and improving the quality of their drivers, they simply have priorities and other objectives that they must accomplish first. If you will recall from the months past with the dynamic display management options and all of the other mobile-related improvements, Lenovo announced (http://www.phoronix.net/forums/showthread.php?t=184&highlight=Lenovo) their Linux support for the T60p right after that... Coincidence? :rolleyes:

[drag]

Oh, this is a nice one:
http://download2.rapid7.com/r7-0025/

Security vunerabilities in Nvidia binary drivers.

With active proof-of-concept exploit for Linux. This is not only a local vunerability with X + nvidia drivers, but it can be remotely exploitable.. for instance if your browsing to a malicious website it's possible for somebody to have your browser display something _which_could_give_them_root_access_.

This is a problem with how Nvidia accelerates rendering of text. This is a very very serious problem.

This bug has been around for years now. First reported in 2004 it took nearly 2 years for Nvidia to aknowledge the problem, which was in July 2006 and they still haven't fixed it.

To me this is headlines-style stuff.

[da.phreak]

First reported in 2004 it took nearly 2 years for Nvidia to aknowledge the problem, which was in July 2006 and they still haven't fixed it.

Fixed in 96xx Beta drivers :-)

[Svartalf]

If you will recall from the months past with the dynamic display management options and all of the other mobile-related improvements, Lenovo announced (http://www.phoronix.net/forums/showthread.php?t=184&highlight=Lenovo) their Linux support for the T60p right after that... Coincidence? :rolleyes:

Heh, I'd think not. As for priorities...You'd think they'd FIX something that is patently broken rather than adding new features...

[MU_Engineer]

The availability or lack thereof of OSS drivers does not really influence my purchase. I happen to have run both NVIDIA and ATi cards on open and blob drivers and they all run fine. My laptop's Radeon M9000 (RV250) runs nicely on the OSS drivers and my desktop's x1900GT runs nicely on the fglrx drivers. I used to use an NVIDIA 6200 and it ran fine with its blob drivers too.

Oh, and my motherboard is an NVIDIA Nforce4 chipset model. All of the devices on it work with OSS drivers:

SATA: nv_sata
IDE: generic NForce IDE drivers
Ethernet: forcedeth (better than the prop. nvnet module as forcedeth supports GbE.)
USB: generic OHCI/UHCI drivers
Sound: snd_intel-8x0
i2c: i2c_nforce2