AMD cannot seem to make up their minds about their Mantle graphics API and Linux support, but thanks to new information it seems it may happen after-all. Having Mantkle on Linux would be useful I'm sure if it can help with some developers fear of OpenGL, but it's usefulness will still be limited to AMD cards until they properly open up the API.
It's not exactly likely we will see Mantle picked up by either Intel or Nvidia anytime soon either, so again the usefulness is limited even if it does come to Linux.
PC World spoke to Richard Huddy of AMD about it:
At least they are thinking about it right now.
It's nice of AMD to say that, but without them committing to supporting Linux it is just words of course and it still isn't confirmation they actually will.
I imagine right now AMD are waiting to see what happens with Steam Machines to see if they want to put the effort of porting Mantle to Linux.
I seriously hope they do bring it to Linux just to have OpenGL have competition, as competition is healthy for all of us when it comes to getting the best performance.
Source Article taken from GamingOnLinux.com.
It's not exactly likely we will see Mantle picked up by either Intel or Nvidia anytime soon either, so again the usefulness is limited even if it does come to Linux.
PC World spoke to Richard Huddy of AMD about it:
Richard HuddyIt could provide some advantages on Steam boxes. We are getting requests to deliver this high-performance layer.
At least they are thinking about it right now.
PC WorldAround 50 games have been developed or will be released soon using Mantle tools, with more in the works. Beyond Windows, AMD sees a vast opportunity to “reveal the goodness of gaming” in Linux, said Richard Huddy, gaming scientist at AMD.
It's nice of AMD to say that, but without them committing to supporting Linux it is just words of course and it still isn't confirmation they actually will.
I imagine right now AMD are waiting to see what happens with Steam Machines to see if they want to put the effort of porting Mantle to Linux.
I seriously hope they do bring it to Linux just to have OpenGL have competition, as competition is healthy for all of us when it comes to getting the best performance.
Source Article taken from GamingOnLinux.com.
0 Likes
Some you may have missed, popular articles from the last month:
The comments on this article are closed.
All posts need to follow our rules. For users logged in: please hit the Report Flag icon on any post that breaks the rules or contains illegal / harmful content. Guest readers can email us for any issues.
11 comments
Though I still remain critical about AMD and linux, it's good to hear they put some work in to improve linux gaming.
One more quote:
"Windows currently offers a better gaming experience than Linux, but Steam could change that."
It will.
One more quote:
"Windows currently offers a better gaming experience than Linux, but Steam could change that."
It will.
0 Likes
If they add it to the Catalyst Only AMD can p### off If AMD add's it to the Opensource Drivers and OpenGL 5 all be fine with that
2 Likes, Who?
FutureSuture Jun 18, 2014
Just stick with OpenGL and focus all your efforts on the open source driver, AMD. Stop wasting time, your customers have waited long enough. Sounds like tomfoolery to improve the open source driver and then switch APIs.
3 Likes, Who?
Typo on the second line. Mantle is spelled incorrectly.
Yes, I really want an open API, but AMD has not (AFAIK) released any documentation (yet).
Having an open-source implementation would be the best. Whatever they choose to do, I'm done with catalyst anyway.
While I agree with you on your first point, Mantle should bring some improvements to the drivers side : the whole point is talking more directly to the hardware, therefore eliminating driver bugs due to an incorrect implementation, or a not well-tested function. Reducing the complexity of the driver is a good thing on this side. The bad thing is, of course, that we're going lower-level, and we become more hardware dependent. Hence the GCN-only implementation : future cards can be mantle-compatible, but they should be designed for it. They said this about multi-vendor compatibility.
I can only imagine Mantle requires a specific set of instruction, and maybe some function pipeline that can't be implemented (or this is not trivial) on earlier cards.
Yes, I really want an open API, but AMD has not (AFAIK) released any documentation (yet).
Having an open-source implementation would be the best. Whatever they choose to do, I'm done with catalyst anyway.
it does not solve the problem of their sh**ty OpenGL drivers
While I agree with you on your first point, Mantle should bring some improvements to the drivers side : the whole point is talking more directly to the hardware, therefore eliminating driver bugs due to an incorrect implementation, or a not well-tested function. Reducing the complexity of the driver is a good thing on this side. The bad thing is, of course, that we're going lower-level, and we become more hardware dependent. Hence the GCN-only implementation : future cards can be mantle-compatible, but they should be designed for it. They said this about multi-vendor compatibility.
I can only imagine Mantle requires a specific set of instruction, and maybe some function pipeline that can't be implemented (or this is not trivial) on earlier cards.
0 Likes
What open source API is even supposed to mean ?
AMD over time will dedicate resources to the task, Huddy said. AMD hasn’t provided a time frame for when Mantle-based games for Linux will become availableNot started yet ? We are not going to see this before a while ...
1 Likes, Who?
^Yeah, that's the flipside to a low-level API, if they don't have high-level functions too. But then developers have been asking for the low-level access so they could optimize. You could do that in OGL currently, even if some people think it's a mess right now =P They should just improve OGL with a 'redesign' in 5.x and retain legacy (2.x, 3.x, 4.x) compatibility.
0 Likes
Yes, I think OpenGL 5 should drop some compatibility if they want their implementation to be less varying from one vendor to an other. Of course, the drivers could still allow earlier versions to be executed, or even calling OGL5 functions while using earlier sets.
This can of course be abstracted with libraries, while still allowing to do certain optimizations which are impossible to do nowadays. These libraries will probably be written by competent people. The code will just be as tested as the game is, and frankly, I do not want to play a game that has not been tested enough (TW2, anyone? :P).
Being lower level is sometimes a good thing : the advantage is that you can build yourself (eventually with existing tools) as many abstraction layers as you wish, while being able to do some very specific optimization at some times.
As an example, just take the cache hit problem in processors : variables that have been used are cached in L1/L2/L3 caches, to avoid latency from pulling these from RAM. Many algorithms exists (check wikipedia to replace the right variable (one that will not be accessed again, in the best case). But, as a programmer, you have no way (AFAIK) to tell usage frequency of those variables. You are not close enough to the hardware to do so, but it would certainly speed things up a lot if it were possible, especially with certain algorithms. This is an example of a situation where being lower level would be profitable in term of performance.
That said, I am not really into computer graphics (I may dive into the day I find a suitable API ;) ), so I can't tell if advantages of Mantle outweigh its drawbacks, but from an objective perspective, it seems reasonable.
But code that is no longer in the driver will now be in the applications. So instead of well tested code written by people who know what they are doing, it will be written many times differently by many users, be less tested, and probably less efficient. Seems like a waste to me.Of course, code duplication is a problem. But almost every game out there is using its very own code to handle, let's say, data compression.
This can of course be abstracted with libraries, while still allowing to do certain optimizations which are impossible to do nowadays. These libraries will probably be written by competent people. The code will just be as tested as the game is, and frankly, I do not want to play a game that has not been tested enough (TW2, anyone? :P).
Being lower level is sometimes a good thing : the advantage is that you can build yourself (eventually with existing tools) as many abstraction layers as you wish, while being able to do some very specific optimization at some times.
As an example, just take the cache hit problem in processors : variables that have been used are cached in L1/L2/L3 caches, to avoid latency from pulling these from RAM. Many algorithms exists (check wikipedia to replace the right variable (one that will not be accessed again, in the best case). But, as a programmer, you have no way (AFAIK) to tell usage frequency of those variables. You are not close enough to the hardware to do so, but it would certainly speed things up a lot if it were possible, especially with certain algorithms. This is an example of a situation where being lower level would be profitable in term of performance.
That said, I am not really into computer graphics (I may dive into the day I find a suitable API ;) ), so I can't tell if advantages of Mantle outweigh its drawbacks, but from an objective perspective, it seems reasonable.
0 Likes
The full video of the interview, it's really interesting. And long. (Not really about Linux)
https://www.youtube.com/watch?v=8uoD8YKwtww
https://www.youtube.com/watch?v=8uoD8YKwtww
0 Likes
I found an interesting article on how features are added to OpenGL (http://www.openglsuperbible.com/2014/02/28/opengl-origins/) and it seems quite a slow process and I wonder if OpenGL is really a good choice for gaming considering that Khronos seems focused a lot more on other technologies (WebGL/CL, OpenGL ES...) and that AMD and Apple (both promoter members of Khronos) could start boycotting/damaging OpenGL to help Mantle and Metal gaining traction...
Hopefully there will be some news at SIGGRAPH in August, after all the articles about how OpenGL sucks and the new APIs from Khronos members I think they really should try to reassure developers that OpenGL has a future.
Hopefully there will be some news at SIGGRAPH in August, after all the articles about how OpenGL sucks and the new APIs from Khronos members I think they really should try to reassure developers that OpenGL has a future.
0 Likes
The full video of the interview, it's really interesting. And long. (Not really about Linux)I was searching for that video, very informative.
Thank you
0 Likes
This is going to be 3dfx all over again if Mantle takes off.
0 Likes
Patreon
PayPal
See more from me