Originally posted by Hamish Wilson
View Post
Announcement
Collapse
No announcement yet.
AMD Has Open-Source Driver For HD 8000 Series
Collapse
X
-
Since nobody explicitly meantioned it already:
AMD team is on good track. They publish their drivers for new hardware, earlier and earlier with each generation. (7xxx was victim of new architecture and 8xxx which could be served in time!)
So for 9xxx there will be good support BEFORE release. That mean many thing:
1) AMD folks will have more time, since they will not have to start work on next gen hw, right away.
2) AMD folks will have more time, since they will not have to deal with hw bugs on their own, since Catalyst team will be also interested in resolving them.
3) AMD folks will be "more productive", since they will release code in smaller packages, but for longer time spans. (Now its big pushes for each hw)
4) AMD folks will be "more productive", since they priority number one "get pre-release support done!" will take less time, leaving them more time on pushing new OpenGL stuff.
5) AMD folks will be "more productive", since they will be able to molest their legal department more often for permission to release various bits and pieces.
Hmm. Its time to see how my 5720M is doing with r600g
PS Do not school Michael for his messy understanding of the state of radeonSI. Grab PTS and upload some test results to OpenBenchmarking.org. And send him links :P (Or share on MESA mailing list, since he have some automated spiders tracking such stuff )
Comment
-
-
I do not question AMD mesa/gallium developer abilities, I don't think anybody here can say they do not do stelar work. R600g is already a very good driver, probably as good as or even better than intel's, IP issues aside.
The problem lies exactly in the IP issues. I question AMD management's commitment to the open drivers and linux support in general (specially in the consumer space), and how far are they prepared to go to provide proper consumer ready, feature complete bug free linux support of their products. Currently, my understanding is that it is pretty much up to the mesa/gallium devs to write some magical code that is able to perform as desired without revealing any compromising IP. Maybe they will be able to do that, maybe not. Maybe this can only be properly resolved by some other route, such renegociating the license contract with whatever third party IP provider they have, buying said IP, buying the IP holder or release it and risk going to court. But that is outside of the scope of the devs.
Some of these options are probably completely absurd. But the worse possible alternative is that all of them are, and that only leaves avoid using such IP in future generations, but that will not help current cards. The most we can expect is fomr them is to honestly confirm that features X, Y and Z canot be supported for generations xxx-xxx due to such IP licensing issues. That has been pretty much done for UVD for example.
What astonishes me is that according to brigdman AMD sees mesa/gallium as the way to go to support consumer GPUs, so fglrx is only a temporary solution, but at the same time, they cannot honestly say that we will be able to see a mesa/gallium feature complete driver in the next 5 yearss. fglrx shouldn't be expected to properly work also (also not feature complete). So what we conclude from that is that for AMD's management, consumer GPU linux support is not even an afterthought. They are pretty much betting all their coins on windows (in the consumer space). That has been a safe bet for the last 25 years, but I personaly do not see a future where windows continues to be only consumer operating system with a decent marketshare. To me, the mobile revolution is much more than simply smartphones and tablets, it is about these quirky, goodlooking and userfriendly OSes that are light, portable and flexible enough to be in every device, providing you with as much information and as much configurability as possible to extract the most from our devices. This resembles a lot AMD's vision of "surround computing", but it is hardly compatible with MS business. Android is what it is today because people are releasing consoles, watches, robots, toaters, fridges, god knows what running the OS, MS cannot compete with that. Will they have some space in next 5-10 years? Sure! Will windows still be the dominant OS? well, hard to say...
Comment
-
Originally posted by fa5hion View PostI wouldn't be so sure about that. Seems that the 8xxx cards are just rebrands of the 7xxx ones for the oem market. Can bridgman comment on driver status regarding actual next gen hardware?
Originally posted by Figueiredo View PostThe problem lies exactly in the IP issues. I question AMD management's commitment to the open drivers and linux support in general (specially in the consumer space), and how far are they prepared to go to provide proper consumer ready, feature complete bug free linux support of their products. Currently, my understanding is that it is pretty much up to the mesa/gallium devs to write some magical code that is able to perform as desired without revealing any compromising IP. Maybe they will be able to do that, maybe not. Maybe this can only be properly resolved by some other route, such renegociating the license contract with whatever third party IP provider they have, buying said IP, buying the IP holder or release it and risk going to court. But that is outside of the scope of the devs.
Originally posted by Figueiredo View PostSome of these options are probably completely absurd. But the worse possible alternative is that all of them are, and that only leaves avoid using such IP in future generations, but that will not help current cards. The most we can expect is fomr them is to honestly confirm that features X, Y and Z canot be supported for generations xxx-xxx due to such IP licensing issues. That has been pretty much done for UVD for example.
Originally posted by Figueiredo View PostWhat astonishes me is that according to brigdman AMD sees mesa/gallium as the way to go to support consumer GPUs, so fglrx is only a temporary solution,
Um... no. What I said was that I saw fglrx as the way to go for 3D workstation, and that I expected a mix of driver usage for consumer.
Originally posted by Figueiredo View Postbut at the same time, they cannot honestly say that we will be able to see a mesa/gallium feature complete driver in the next 5 yearss.Test signature
Comment
-
Originally posted by Figueiredo View PostThe problem lies exactly in the IP issues. I question AMD management's commitment to the open drivers and linux support in general (specially in the consumer space), and how far are they prepared to go to provide proper consumer ready, feature complete bug free linux support of their products.
Originally posted by bridgman View PostOne more time -- our commitment is to support community development of open source drivers, and we are going further than that by hiring some really good developers to work on the drivers full time.
Comment
-
Originally posted by Asariati View PostJust curious: Are you looking to increase the number of driver developers?
Comment
-
Originally posted by bridgman View PostUm... no. What I said was that I saw fglrx as the way to go for 3D workstation, and that I expected a mix of driver usage for consumer.
Comment
-
Originally posted by curaga View Post@Bridgman Does 2d on glamor still tear?
Originally posted by ChrisXY View PostIs "3D workstation" only the firegl "professional" graphics cards? Can a notebook with enduro be a workstation too?Last edited by bridgman; 06 February 2013, 08:09 PM.Test signature
Comment
Comment