why does fglrx suck so hard?

[Panix]

Time matters more than manpower the R600 opensource driver starts 5 years to late!

yes R600-R800 will have only basic support for opensource because of the 5 year late starting dev.

but R900 will have full opensource dev power from the beginning!

in the end AMD will do it right on the R900 because if they wana sell CPUs they need a good Grafic card to because people buy both in a round or they buy intel/nvidia......

I disagree, sir. If they don't have the manpower or resources, then the time co-relates. They will run OUT of time. What do you mean about the R900? What is that? What is 'full opensource dev power?' I was talking about fglrx drivers. Is ATI/AMD going to release full functionality via OSS drivers? I doubt it. It would be good but probably not realistic.

I don't understand ATI's/AMD's strategy. The most expensive cards are being bought by gamers. That's where the money is, isn't it? But, fglrx drivers sound like they're a mess. There is a great opportunity now with Nvidia blowing it with overpowered, high heat producing new cards that seem to be way less efficient compared to ATI's highest performance cards. I'm talking about Fermi and going by what I've read. I probably shouldn't comment like this but even at a basic level, AMD/ATI have a good opportunity right now in both Windows and especially Linux. Whether hard core Linux users like it or not, Linux devs are dumming down their distros (hey, I won't complain too much) and making it easier all the time for basic computer users. The Desktop is expanding into way more GUI styles and offer so many various DEs to use, Gnome, KDE, xfce, LXDE, OpenBox, Fluxbox etc. etc. Linux experts have called some of the changes 'more bloat' but really, if the growing pains can be dealt with, Linux could appeal more to the general user if the progressions continue. There's an actual threat to Microsoft despite the apparent success of Windows 7.

With gamers buying up ATI cards, some of those will start dual booting Linux but if problems are experienced because of poor ATI drivers, that experiment will fail. Other Linux users, whether they dual boot or not will want ATI cards when upgrading because the newest hardware is superior to Nvidia alternatives. So, poor ATI drivers will dissaude those Linux users as well. OSS drivers are good but to continually focus on these until fglrx drivers are 100% functional and on a level with Nvidia and/or ATI drivers in Windows is shortsighted, imho. OSS drivers should be an ultimate goal or ideal but to to have both types of drivers in an incomplete and subpar state ALL THE TIME is a big mistake that ATI/AMD might be underestimating the fallout. Imho, they are missing an opportunity to fix this problem. But, what do I know, I guess....

[Panix]

That's where the money is, isn't it? But, fglrx drivers sound like they're a mess.

Btw, I didn't mean to imply it's more money than professional workstation cards but to suggest that both strategies should be to optimize one of the drivers and ensure everything is working and up to date even with the changes in Linux. If that is the fglrx driver, then so be it. But, maybe experts of these things should comment. I'm just suggesting...

[droidhacker]

I don't understand ATI's/AMD's strategy. The most expensive cards are being bought by gamers. That's where the money is, isn't it?

I believe that this assumption is incorrect.

Tell me if you can; what proportion of graphics card owners are GAMERS?
1%? 0.5%? Very low!

Assuming 1%, then the value of having top performing and fully-featured open source linux drivers right away on launch day is worth 1% of 1% of the total card sales! And this isn't even considering the fact that ***MOST GAMERS*** can't even AFFORD the top-of-the-line card! They will therefore go with something a bit faster than middle-of-the-road! So this brings the open source support for the top of the top card even further from necessity.... say for example 1% of GAMERS can even afford the top of the line card... you are now down to 1% of (1% of 1%) of sales depend on immediate and complete linux support (open source or not!) -- that means ONE IN A MILLION!

The top end cards for linux gamers represent a NEGLIGIBLE revenue!

[Kano]

Most problematic are mainly laptops. You can not exchange the gfx part when you buy one. So if you would be a Fedora fan you have to stay far away from ATI. U users maybe a bit more happy, but compared to Nv all lost. Everytime i test a new app i found more errors with fglrx. Like when i began using XBMC then the Auto/GLSL setting was broken with fglrx and still no fix is in sight. That's really bad when you see no picture by default, only when you know that you can set it to ARB rendering method...

[bridgman]

What do you mean about the R900? What is that? What is 'full opensource dev power?'

R900 is presumably a reference to our next generation of GPUs after Evergreen (which some people call R800). I think "Full open source dev power" means that by the time the next generation of GPU comes out the developers will already have open source drivers for the previous generation with all the core functionality and will be able to concentrate on adding support for the new parts rather than having to split time between adding features for older parts and working on new parts (as is happening now with Evergreen).

I was talking about fglrx drivers. Is ATI/AMD going to release full functionality via OSS drivers? I doubt it. It would be good but probably not realistic.

Depends on what you mean by "release", I guess. I expect we will release enough information and sample code to write fully functional OSS drivers, with the exception of drm-sensitive areas like UVD and patent-encumbered areas where nobody has found a good way to license those patents for use in an open source stack, but remember that we aren't writing the OSS drivers ourselves, just helping the community write them.

I don't understand ATI's/AMD's strategy. The most expensive cards are being bought by gamers. That's where the money is, isn't it?

Nope, not even close. I think that's the main disconnect here. Most of the Linux money is in workstation, and most of the consumer business is low end to low-midrange cards. There are definitely some users doing high end gaming on Linux and we are trying to support them, but not because there is significant money there. The good news is that there is a pretty good overlap between high end gaming requirements and workstation requirements, with the caveat that workstation users tend to run their Windows apps on Windows, not on Wine+Linux.

The other "money" areas of the Linux market - servers and business clients - both have a strong preference for open source drivers and one-stop support from distro vendors.

OSS drivers are good but to continually focus on these until fglrx drivers are 100% functional and on a level with Nvidia and/or ATI drivers in Windows is shortsighted, imho.

I don't understand this statement at all. Most of our development effort goes into fglrx, not the open source drivers. When we recommend open source drivers today it's because they are a better fit for the specific requirements of a specific user.

OSS drivers should be an ultimate goal or ideal but to to have both types of drivers in an incomplete and subpar state ALL THE TIME is a big mistake that ATI/AMD might be underestimating the fallout.

Again, I don't get this part. Cancelling our open source efforts and shifting all the resources to fglrx would not make a substantial difference in fglrx progress, and moving all of our fglrx developers to work on the open source drivers would not allow those drivers to replace fglrx in the workstation market. I understand that nobody likes to wait for development to happen but between starting and finishing something you have to pass through some "part way there" states which can't be avoided unless you have easy access to time travel (which has its own problems, see http://xkcd.com/716/ ).

Imho, they are missing an opportunity to fix this problem. But, what do I know, I guess....

I guess I don't understand exactly what you are suggesting. There's a big heap of work to do and we are working through it, and are already investing quite a bit more than the current Linux market can justify. In the meantime we have open source drivers which meet the needs of many consumer users, proprietary drivers which meet the needs of many *other* users, and an ever-diminishing gap between the two.

What exactly are you proposing that we do ?

[PsynoKhi0]

From a potential customer position though the hardware/software combo still has to deliver.

Just curious, does anybody else than me add "company ethics" to the equation?

[jrch2k8]

Just curious, does anybody else than me add "company ethics" to the equation?

well if you think about it is obvious the lack of interest of AMD to provide any sort of better support to that useless crappy piece of junk called fglrx, but they have a reason, is extremely hard to compete with nVidia in the workstation side. let see why

nVidia quadro
+ big power processing
+ use anywhere everywhere (windows, linux, bsd, solaris)
+ Tesla ass kicker enabled
+ suport for the latest technologies except maybe DX11
+ Cuda have become basically the standard of gpgpu software out there cuz amd slow development in that area
+ purevideo widely available (VDPAU)
+ standard reference for QA, try to debug or support any sort last tech without an nvidia card i dare you!
- a bit expensive

AMD firegl
+ big power processing
+ pretty decent in windows, suck everywhere else lol even maya for linux in rh5 eld sux
+ very stalled in the software side to support latest tech
+ poor or none gpgpu support
+ UVD well in some configuration i heard it works, cuz my windvd in windws 7 performe worst than vlc with bluray so ....
+ ovbiously lol ofc
- really expensive in some models some dont

so fglrx will never get more attention cuz they motivation is enterprise, and let say clearly AMD cant compete with nvidia here either in any time soon, so the few maret share they still retain in unixes oses is fading probably very fast against their counterpart except maybe in some niche sectors. so their focus is windows with the big market share and thats it.

so why they keep releasing fglrx? simpy for support on those remaining here and there in the enterprise and as PR thing, either way when you founded out what fglrx really is, well erm you already buyed the card, so as far as the marketing team cares they sold a card .

so the OSS is the only way for now cuz they can save money in human resources and eventually when the driver get good enough they can market that, and attack hard the unixes market to get more sales (remember after all normal ppl looking to migrate to linux or any other unix like os is gonna ask us, you know, and lets face it a pure OSS driver is like the ozzfest for geeks ).

the only good thing here is that in the meantime we are getting the first all OSS driver for a big card and that is good enough for me, especially cuz is getting really good, in some areas kick even the nvidia driver already is some other need more work.

so you wanna a fully amd driver, get yourself a distro with easy access to latet OSS code like ubuntu, test it, use it, embrace it and post all the bug you found like a mad men in mailing list, phoronix, irc and if you can code try to contribute too. in the OSS side AMD guys with the community are doing a nice job so kudos for you all.

just a thought for bridgman , since fglrx is not like is gonna get any better any time soon, talk to your boss to assign you more time to the oss driver that have more chances to spread like an infection when is ready

[mugginz]

Just curious, does anybody else than me add "company ethics" to the equation?

I do. I however absolutely do not even for one moment equate the production of a binary blob driver with being unethical.

nVidia have come through in spades with the graphical goods as no one else has for the Linux desktop from the perspective of functionality. Just look at this announcement for a recent demonstration of that.

The microsecond that there's an OSS solution for graphics that delivers in a similar way I'll have my wallet out. I only buy networking hardware with open drivers, and both of my laptops have had Intel graphics. For their role the performance was completely acceptable and also very capable with a composited desktop.

At the moment if I want to run with ATI graphics hardware I'll need to run Microsoft Windows which would diminish the point somewhat. I've been waiting since last November to buy a graphics card and have only been putting it off in the hope that I can buy an ATI card, but the drivers have to work properly before I can do that.

Many people try to describe the ATI drivers as viable, but they can only be considered that if you're prepared to go without certain aspects of a Linux desktop. Their drivers are suitable for certain restricted environments but mine isn't one of those.

[mugginz]

*crosses fingers for 10.04*

https://launchpad.net/ubuntu/lucid/+...8.721-0ubuntu1

[bridgman]

just a thought for bridgman , since fglrx is not like is gonna get any better any time soon, talk to your boss to assign you more time to the oss driver that have more chances to spread like an infection when is ready

The "Linux" part of my job has just been managing the open source graphics project, since mid-2007 when the project started. I try to answer questions about fglrx and the Linux market when I can, but that's more of a hobby than a job. The rest of my time goes to non-Linux work, not fglrx.

Your comparison table might be missing DX11. We did launch an entire line of DX11 cards with Linux support when you thought we weren't doing anything