AMD Mantle API (Low level hardware access)

Discussion in 'Videocards - AMD Radeon Drivers Section' started by shadow_craft, Sep 26, 2013.

  1. BaNsHiE

    BaNsHiE Active Member

    Messages:
    56
    Likes Received:
    0
    GPU:
    Saphire 7970 TriFire
    Sigh, The only people that wont embrace Mantle are the hackers. Picture the hacker playing BF4 in DX11 and he cant pickup the players running Mantle on a tri-fire R9-290x system running at light speed, makes me cry thinking about it.

    NOT:banana:

    LOL
    BaNsHiE
     
  2. temo

    temo Member

    Messages:
    23
    Likes Received:
    0
    GPU:
    GTX 1070
    The performance will be improved but that's not really the interesting thing here.

    The question is just how close Mantle is to the existing consoles apis. Anand believes it's a straight rip from the Xbone and Digital Foundaries is saying it's suspiciously similar to the PS4.

    How much time and money will developers put into OGL/D3D ports for PC if there's something cheaper and easier?
     
  3. Redemption80

    Redemption80 Ancient Guru

    Messages:
    18,335
    Likes Received:
    185
    GPU:
    GALAX 970/ASUS 970
    That just got me thinking, will Mantle even have Crossfire support?
     
  4. -Tj-

    -Tj- Ancient Guru

    Messages:
    16,413
    Likes Received:
    1,498
    GPU:
    Zotac GTX980Ti OC
    ^
    If its close to metal like they claim, then crossfiring it wont have any issues, only more raw power.
    Ok yes 100k over even 150k is the best scenario, but even if its half of that ie 40-70K calls, its still A LOT!!


    Anyway, I can't wait to see this in action and how much more world detail will it bring :)
     

  5. The Mac

    The Mac Ancient Guru

    Messages:
    4,408
    Likes Received:
    0
    GPU:
    Sapphire R9-290 Vapor-X
    its an api, some douche hacker will find a way to make an injector.

    It still all boils down to money.

    The developers/publishers with the deep pockets (EA) MAY code for both render paths, but the smaller ones dont have the cash. Its just not possible for them.

    Throw in the fact that a lot of development costs are spent on assests, and they will either have to spend a lot more for mantle due to a much higher render target, or waste all that extra performance by reusing the same assets from DX.
     
  6. Death_Lord

    Death_Lord Master Guru

    Messages:
    711
    Likes Received:
    7
    GPU:
    Aorus 1080 GTX 8GB
    Seriously, I cant believe how ignorant some people is, really? what you think it costs to port DirectX to OpenGL? how many AAA games are out there running on Macs and macs only have OpenGL? what about linux? or consoles or tablets or phones?, they alll run on OpenGL, its just about translating the code from DirectX to OpenGL, and thats the easy part, porting to mantle will not be harder, and even if a game developer doesn't want to put money on making better graphics that will only run on Mantle, what about being able to run the same graphics at 90 FPS instead of 30 FPS?

    I remember back then when the first Unreal was released, it supported 5 different api's if I remember well, did they had to remade the engine 5 times???
    NO, they simply had to support the other api's, so seriously, we are not talking about PhysX or Cuda here, we are talking about a new api that will help developers get a lot more of performance out of AMD GPUs and maybe get console ports that doesnt need to be tweaked to run fine on OpenGL or DirectX.

    I dont see were all the hate comes, maybe unwinder doesnt want to redo Rivatuner Statics server to run on Mantle :p, or some people doesnt want better performance? (several GCN owners on this thread doesnt seem very excited about mantle...) but well, each on their own.
    :deadhorse:
     
  7. The Mac

    The Mac Ancient Guru

    Messages:
    4,408
    Likes Received:
    0
    GPU:
    Sapphire R9-290 Vapor-X
    I think its great, but ive been in the industry since its inception in the early 90's. Ive seen proprietary apis come and go.

    Its honestly a crap shoot.

    It will be either be embraced, or determined too expensive from a development standpoint.

    My take at this moment is that developers wont be willing to spend money on the secondary render path, and increased asset expenses.

    I hope that will change.

    time will tell.

    btw, QFT: you shouldnt make this stement and call people ignorant: "Seriously, I cant believe how ignorant some people is".
     
    Last edited: Sep 27, 2013
  8. draegunstrife

    draegunstrife Active Member

    Messages:
    66
    Likes Received:
    0
    GPU:
    EVGA 780 Classified
    I believe this is what AMD meant when they were talking about the Microsoft and Direct x.
     
  9. 007.SirBond

    007.SirBond Master Guru

    Messages:
    295
    Likes Received:
    0
    GPU:
    NVIDIA GTX Titan 6GB
    Even tho I have bought only Nvidia products for the longest time. I do realize that AMD/ATI has always been more supportive for developers and open source things.

    http://www.youtube.com/watch?v=_36yNWw_07g

    If PhysX was open sourced, we would of seen a lot more games using it, it would of helped the industry itself. Proprietary business hurts more than helps. It's stupid, the average consumer doesn't even give a **** about the specific features of things if they are hardly being used.
     
  10. Espionage724

    Espionage724 Guest

    When I first heard about Mantle, I was pretty excited :)

    I would love to see Mantle widely-used and quickly adopted, especially if it'll be cross-platform. Would also like to see Intel and NVIDIA quickly work with it to make it usable on their own hardware.

    I however, just don't know if that will happen. We've had OpenGL and DirectX for a while now, but look at what API most AAA games chose? We have OpenGL, a pretty decent API. And then DirectX, a proprietary, Windows-only API. Feature-wise, both OGL and DX can do practically the same thing, and yet, DirectX is what most developers choose to go with.

    These same developers also do OS X versions of their games too. Almost makes you wonder why they choose to double their work instead of just primarily using OpenGL, but who knows. I'm not a developer, but if I was, I'd use OpenGL for all of my clients (Windows, OS X, Linux), and also throw in Mantle on the side. Depending on how widely Mantle is accepted and compatibility, I'd probably just use Mantle primarily at some point too.

    As for Intel and NVIDIA, I doubt they'd jump on board too. Intel already publicly dropped all support for Mir on Linux, even though they were working on it. They seem pretty intent on Wayland, but this is a whole different discussion :p Intel can barely handle their own OpenGL support. Graphics-wise, I don't see Intel going for it.

    NVIDIA is pretty much only into their proprietary tech. I've never heard of NVIDIA using someone else's open tech. NVIDIA GPUs do support OpenCL and DirectCompute, but really how many owners really know this? It's all about CUDA, and NVIDIA is pretty well-intent on advertising and using it way before OpenCL and DirectCompute. Even on the Linux side, they've been pretty lock-and-key with their hardware specifications towards helping nouveau, but only recently, decided to publish a document detailing (what I understand) already-known information. I predict NVIDIA to make their own API similar to Mantle, and restrict it purely to their hardware (also throw in a bunch of buzzwords to make it sound better than Mantle too of course). Either that, or they'll just keep to what they're doing, and stick with DirectX and OpenGL.

    Simply put, the idea of Mantle is pretty awesome and I know I myself (if I were a developer) would start using it immediately. I don't see it being widely used or adopted though sadly, but I would love to be wrong about this.
     
    Last edited by a moderator: Sep 27, 2013

  11. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,755
    Likes Received:
    2,203
    GPU:
    5700XT+AW@240Hz
    They would be crazy to try to counter Mantle with their own solution of same grade, because they can have this solution on Shield only as console support.
    And while I like Tegra 4 performance, I do not like shield.

    Would be like fighting Phosphor Fire with candle light. Because for developers price wise Mantle Starts on PS4, maybe XBox1. And once used there, cost to get to PC Mantle will be small.
    So PS4 developer can Either nor use mantle and stay purely PS4 without spending additional cost on PC or use Mantle on PS4 and then cheaply and quickly get to 25% market on PC of AMD compatible chips getting same visual experience and having cross platform multiplayer.
     
  12. Unwinder

    Unwinder Moderator Staff Member

    Messages:
    14,441
    Likes Received:
    1,464
    The most stupid fanboy comment from entire thread, sorry.
     
  13. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,755
    Likes Received:
    2,203
    GPU:
    5700XT+AW@240Hz
    But pointing in direction I can agree with: "Hell, RadeonPro will not work with this at 1st."

    Otherwise that comment was pointless.
     
  14. temo

    temo Member

    Messages:
    23
    Likes Received:
    0
    GPU:
    GTX 1070
    The entire point of Mantle is to largely get rid of D3D/OGL as the secondary render path. They're both performance hogs and the new consoles either require a wrapper for one or don't support the other. The benefit of Mantle is that it's a mutually compatible open api between PC, PS4 and Xbone. In theory you could borrow large chunks of rendering code straight from PC to either console and save big bucks. Although, as DICE carefully pointed out, theory does not always equal practice.

    You're correct in that any transition will likely take years of lead time and AMD will need to prove the savings (likely their motivation for paying off DICE) if they wish to accelerate adoption.
     
  15. Death_Lord

    Death_Lord Master Guru

    Messages:
    711
    Likes Received:
    7
    GPU:
    Aorus 1080 GTX 8GB
    On what company you actually work? or what exactly you do on the industry?
    Are you sure that companies like Epic, Crytek, Id Software, Activision, EA, Valve, will not pay to implement something like Mantle that promises huge amounts of extra performance (even if its just a 40% more it would be important)? did not Epic include PhysX on their unreal engine, even if it was GPU dedicated or not? and how many other small companies has implemented PhysX on their small little engines even if it only runs on one specific brand?

    I'm sure AMD will have a support system for people that wants to implement this new Api on their software and im sure implementing it can be worth it for many people.

    People talk about Glide and other Apis like if they were standalone piece of software floating around, but if you knew about the history of gpus, you would clearly know that those apis died because their companies died or got bought by other companies, just like what happen to 3dFX, nvidia could have continued with glide, but they chose to just eat 3dfx and bury the world around it, Glide was supported by almost every single game that had GPU support, and all those games also supported OpenGL or DirectX.

    Mantle will be open in code, so anyone can implement it, and its also going to be cross platform, all next gen consoles use AMD hardware and im pretty sure they will see cost effective to implement Mantle directly into the pc port, since today every single company focus first on consoles and later on PC.

    Im sorry if I offended anyone by saying that some people are ignorant but the truth is that some people are ignorant and they will talk about something without a single notion of knowledge

    Unwinder, You need to start relaxing you know that? I know you have a lot of weight in your shoulders with all the work you do but every time i see you on a thread is arguing with someone else, if i remember right you even left the forums some time ago because you where tired of all the crap, so just chill, I'm not a fanboy, i just think that having having big arguments over something that could be a benefit is not the right way to go and a waste of time.
     

  16. Unwinder

    Unwinder Moderator Staff Member

    Messages:
    14,441
    Likes Received:
    1,464
    Wrong example. PhysX is a different story, AEGIA initially introduced this hardware and API to provide totally NEW class of API allowing game developers to use hardware acceleration for something that was implemented purely via software before. It was never aimed to replace some existing standardized physics calculation API in favor of one hardware vendor. But what NVIDIA did subsequently to PhysX after purchasing AEGIA (marketing games with locking PhysX acceleration if the primary GPU is not NV) is a different side of coin and it deserves to die and leave the market in a form as it is now.

    If you knew history of GPUs a bit better you wouldn't post that as well. The APIs didn't die because the companies magically left the scene. Vice versa, the companies died because their low-level APIs lost the battle against standardized ones.
    You're mistaking on Glide as well, keyword is architecture. It would be total insane for NVIDIA (or anyone else who could purchase 3dfx instead of NVIDIA) to rely on Glide simply because it was low-level API oriented on 3fdx chips specifics. Wrapping totally different GPU architecture in this API is much less effective and drops the sense of having low-level API to something close to zero.

    Marketing fable for those who don't understand what is low-level API. I have zero hopes to see NV or Intel's support for Mantle. Just try to guess why AMD provides Mantle for GCN architecture ONLY and leave even their own older GPUs behind. The answer is the same: architecture.
    The BEST thing that I can expect from Mantle is separated market and seeing games working effectively on AMD cards only. But the WORST thing is seeing some similar answer from NVIDIA side in form of their own low-level API and their aggressive marketing politics ruining standardized 3D APIs. DirectX and PC industry 3D APIs went too long standardization and unification way, it would be simply sad to see it ruined because one vendor failed to compete in DirectX driver development and decided to cut corners with low-level APIs.

    I don't find it relaxing to comment nonsense related to my work. Especially when somebody talk about something without a single notion of knowledge ;)
     
  17. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,755
    Likes Received:
    2,203
    GPU:
    5700XT+AW@240Hz
    I agree with most, but I got feeling you see mantle as following: Application<>Mantle<>HW. But it has one more element...
    You should look at Mantle as what it is, it's layer cake. And each layer as what it stands for:
    Application - You know them pretty well
    API - OGL/DX7,8,9,10,11/Mantle API (this is main difference)
    Driver - Your ability to translate commands from API to HW calls (each manufacturers business and they are pretty equal now)
    Hardware - any GPU out there, here counts raw power unless limited by Driver/API/Application

    What Mantle wants to achieve:
    Application - Delivery of much greater/complex visuals
    API - part where we are hitting wall, therefore place where Mantle wants to do main change to increase possibilities for application developers
    Driver - no changes needed in approach
    Hardware - Use all that raw power to max.

    Deployment obstacles for each layer:
    Application - none, just another API, programmers learn dozens of programming languages and graphical APIs, One more, no problem for real programmer.
    API - Has to redo all effects structures which were developed already through 2 decades.
    Driver - Time consuming as API, optimizations have to be redone.
    Hardware - no obstacles, any modern HW can do any effect from any API, in worst case as emulated (see OGL methods)

    Main point is, that there are no big obstacle preventing deployment. API done by AMD/game developers maybe nVidia if it's willing.
    Driver is what it stands for and since nV know their GPUs they should have as much trouble as creating driver for win/linux/android (new for each architecture).
    Drivers are developed so often that it's only decision of supporting this or not supporting this.

    Now if you look at DX7,8,9,10,11... writing code for each was quite new experience they are basically new APIs with different approaches, elements from old ones + new feature sets where new often replaced old.
     
  18. Unwinder

    Unwinder Moderator Staff Member

    Messages:
    14,441
    Likes Received:
    1,464
    I've got a feeling I understand it quite well, including programming in Direct3D since ancient Direct3D7 and working with both vendor GPUs at low level. But anyway thanks for taking time on explaining.
    The only things Mantle really wants to achieve are:
    1) Gaining extra performance due to minimizing API overhead. = good thing
    2) Gaining extra performance due to taking GPU vendor abstraction totally out of context and forcing devs to code in favor of their own architecture ONLY. = bad, bad thing
     
  19. Fox2232

    Fox2232 Ancient Guru

    Messages:
    9,755
    Likes Received:
    2,203
    GPU:
    5700XT+AW@240Hz
    This is the risk of doing your own API, I remember intel C++ compiler ignoring AMD's CPU optimizations favoring own platform.
    On other hand I believe that if nV comes and say: "Hey AMD this method of DOF/Bokeh is crap on our platform, we want it this way." then AMD will react positively and adds platform based switching to do improvements for nV.
    On other hand that means either nV to work closely with AMD (and nV does not like to work closely with anyone) or they would be in lag with implementation here and there.
    - I remember game developer talking (think it was mafia2) about nV helping them to make something working bit better on nV side. Nv Team came, rewrote part of their code... "no questions", and left. It was then OK on nV but total crap on AMD and they had to revert that code and redo it by themselves anyway.)

    But honestly, no matter what, DX11 is hitting performance wall. It has it's methods which likely does not suit both nV/AMD. Which does not prefer either side. So it can be called fair to gpu vendors, but limiting at same time.

    I do not think Mantle will bring some incredible performance boost, but will allow adequately strong HW to be fully used.

    Looking at Frostbite, they came to microsoft to address some bad things. MS did few things but apparently ignored core problem they had.
    So they approached AMD instead.

    I was talking about wall which we are close to on DX, they did proper analysis of bottlenecks in BF3 and this is result.
    For me it does not matter if I have low/med/high details in BF3 because I am mostly limited by CPU. Yes, limited by intel Quad Core @ 4.5GHz.

    You can see this wall in all benchmarks when they are testing on lower resolutions where GPU does have much lower workload, but FPS does not go over certain threshold.
    60Hz gaming is safe today, but 120Hz+ gaming is already severely affected.
    We need the change and if someone want's to stay behind it's his own decision. But it will result in slower wide deployment for all and no deployment for their (read nV) customers.
     
  20. Final8ty

    Final8ty Master Guru

    Messages:
    315
    Likes Received:
    0
    GPU:
    2x5970 Toxic 4GB 900/1200
    http://www.neogaf.com/forum/showpost.php?p=83775657&postcount=397
    So are we saying 20000 draw calls with the performance hit of 3000 on the PC ?
     

Share This Page