AMD "Mantle" - Page 2

AMD "Mantle"

Post just about everything that isn't directly related to Spring here!

Moderator: Moderators

User avatar
Funkencool
Posts: 542
Joined: 02 Dec 2011, 22:31

Re: AMD "Mantle"

Post by Funkencool »

Wouldn't this just allow software to use mantle for increased efficiency when applicable, yet still allow using OpenGL etc.. ? So developers have more potential added for performance and none taken away?
luckywaldo7
Posts: 1398
Joined: 17 Sep 2008, 04:36

Re: AMD "Mantle"

Post by luckywaldo7 »

varikonniemi wrote:jK: Mantle works on all GCN hardware. It is next gen as in no support for older cards.
Image

Mantle is the api one now can use to code for the low level GCN (or xbox one), bypassing direct x/open gl. "Mantle driver" is what nvidia needs to provide to support the open mantle api (programs that run on mantle).

Tell me why im wrong?
'All Graphic Core Next GPUs' means a specific architecture of amd cards branded 'Graphics Core Next', and not generically next-generation GPUs.

http://www.amd.com/us/products/technolo ... cture.aspx
User avatar
very_bad_soldier
Posts: 1397
Joined: 20 Feb 2007, 01:10

Re: AMD "Mantle"

Post by very_bad_soldier »

jK wrote: As already said it is a low level `api`, it is designed to map commands 1:1 on hardware commands similar to assembler. So instead of giving you functions like `upload texture`, `draw here`, ... it has commands like `copy data into register A`, `copy data from AGP ram to video ram`, `write color C to memory location X`.
Obviously different hardware designs use _different commands_ and so Mantle cannot be compatible by logic!
The same logic applies to x86 instruction set assembly for CPUs, no? They are compatible between different generations and even between Intel and AMD.
User avatar
AF
AI Developer
Posts: 20687
Joined: 14 Sep 2004, 11:32

Re: AMD "Mantle"

Post by AF »

very_bad_soldier wrote:
jK wrote: As already said it is a low level `api`, it is designed to map commands 1:1 on hardware commands similar to assembler. So instead of giving you functions like `upload texture`, `draw here`, ... it has commands like `copy data into register A`, `copy data from AGP ram to video ram`, `write color C to memory location X`.
Obviously different hardware designs use _different commands_ and so Mantle cannot be compatible by logic!
The same logic applies to x86 instruction set assembly for CPUs, no? They are compatible between different generations and even between Intel and AMD.
We don't develop for a mythical Openx86 API though do we?

EA isn't going to build using Mantle alone, in the same way an engine can have OGL and DirectX renderers. Unless AMD is paying them a wad of cash to do this I doubt Mantle will be anything more than a toggle under settings that will probably make things buggy.

Good idea or not, it's a shame the ATI people were the ones to come up with it
User avatar
jK
Spring Developer
Posts: 2299
Joined: 28 Jun 2007, 07:30

Re: AMD "Mantle"

Post by jK »

very_bad_soldier wrote:
jK wrote: As already said it is a low level `api`, it is designed to map commands 1:1 on hardware commands similar to assembler. So instead of giving you functions like `upload texture`, `draw here`, ... it has commands like `copy data into register A`, `copy data from AGP ram to video ram`, `write color C to memory location X`.
Obviously different hardware designs use _different commands_ and so Mantle cannot be compatible by logic!
The same logic applies to x86 instruction set assembly for CPUs, no? They are compatible between different generations and even between Intel and AMD.
CPUs are totally different to GPUs in many ways. I.e. x86 got more and more commands and they always were compatible will _all_ previous generations, so you can still run DOS, OS/2,Win3.1,... on a modern 4GHz hexacore CPU. This is only possible cause x86 was designed like that from the start, it always was the idea to use many commands, and if the vendor wants to save the silicon for an rarely used command, they use so called microcodes, e.g. this ways AMD likely implements a lot SSE commands, they aren't really supported by the hardware but they are _mapped to FPU cmds_. A damn lot commands are implemented this way.
But not all CPUs are constructed this way, so many vendors/engineers hated this design (a lot cmds + microcodes), cause it needs a lot silicon & adds a lot extra latencies. So another design was created so called RISC, found in e.g. ARM & MIPS. When you strictly follow the design rules (no one does so) you cannot remove features from those like in x86, cause you couldn't emulate the commands in hardware. Only recompiling the software help cause the API/ABI changed.

Now GPUs are neither CISC (x86) nor RISC, but they must be FAST, damn FAST. Being to complicated with too many unused special cmds like in x86 this cannot be done, so they are a combination of a reduced & fast instruction set. But with time ppl demand more and so T&L, shaders, ... were added and the GPUs needed to become more flexible. So a GPU is a Hybrid of being flexible & being fast. Like all Hybrids their nature is not 50:50, some are more flexible some less etc.
So there are a lot totally different approaches: Nvidia is different from ATi, and those are different from PowerVR ... and vendors even switch theirs with time (e.g. Fermi GPUs are totally different to the previous ones)

So GPUs don't share the same instruction or feature set!
Only the driver adds a layer that makes them map features to their own instruction set and allows so to run the same software on different hardware.
AF wrote:EA isn't going to build using Mantle alone, in the same way an engine can have OGL and DirectX renderers. Unless AMD is paying them a wad of cash to do this I doubt Mantle will be anything more than a toggle under settings that will probably make things buggy.
For EA it's no problem atm, they just port their PS4/XBOX1 renderer with a few minor adjustments. But this doesn't save them any work (in contrast to what ATi tries to promote in their slides), cause they still need to write a DirectX renderer (maybe even 2 for DX9 & 10/11) when they want to support non-GCN gpus and nvidia (duno if they support intel).
For them it's just a PR lowcost trick to sell their game: a lot ati ppl will buy the game with their new gpu just to see the speedup effect.
varikonniemi
Posts: 451
Joined: 03 Jul 2011, 11:54

Re: AMD "Mantle"

Post by varikonniemi »

Can you post any source that states you have to code low level codepaths specifically for each GPU? I still believe you code for the mantle API (similar to DX or OGL API) and that the driver translates this into low level instructions on the supported GPUs :)
zerver
Spring Developer
Posts: 1358
Joined: 16 Dec 2006, 20:59

Re: AMD "Mantle"

Post by zerver »

I always liked low level shit, so for the time being I'm giving AMD thumbs up. The .NET framework currently fills up one quarter of my SSD, this is a prime example of what happens when these higher level APIs become increasingly bloated.
User avatar
AF
AI Developer
Posts: 20687
Joined: 14 Sep 2004, 11:32

Re: AMD "Mantle"

Post by AF »

zerver wrote:I always liked low level shit, so for the time being I'm giving AMD thumbs up. The .NET framework currently fills up one quarter of my SSD, this is a prime example of what happens when these higher level APIs become increasingly bloated.
Which .Net framework?
User avatar
very_bad_soldier
Posts: 1397
Joined: 20 Feb 2007, 01:10

Re: AMD "Mantle"

Post by very_bad_soldier »

All the people that I know who are complaining about .NET are the ones who never seriously coded with it. No joke.
As long as file size of the runtime is the major complain against .NET I think MS did a great job.
Btw .NET filling 1/4 of your SSD? Ever considered to upgrade your 8GB SSD? 8)

And no, I am not an MS fanboi. I am coding 98% of the time in C++ with gcc (which is great for its purposes also) but I am happy about every chance to use something more advanced (language and library wise).
zerver
Spring Developer
Posts: 1358
Joined: 16 Dec 2006, 20:59

Re: AMD "Mantle"

Post by zerver »

I didn't say it was bad. C# code looks better than C++.

But it is bloated and there are many versions and security patches that add up. Each installation takes a long time because it precompiles stuff.

AMD deciding to use .NET for its driver package is the biggest mistake ever. Something that should be a 2 meg download is instead 200 megs.
varikonniemi
Posts: 451
Joined: 03 Jul 2011, 11:54

Re: AMD "Mantle"

Post by varikonniemi »

jK wrote:2. OpenGL supports extensions for this very reason, so they could implement all their shiny commands with it, but no they want to have a `AMD only` feature (similar to PhysX for Nvidia just worse)
They are implementing everything they can from mantle as open gl extensions.

http://www.dsogaming.com/news/amd-aims- ... ottleneck/
User avatar
very_bad_soldier
Posts: 1397
Joined: 20 Feb 2007, 01:10

Re: AMD "Mantle"

Post by very_bad_soldier »

jK wrote:
varikonniemi wrote:Nvidia is almost forced to support mantle. They have no chance of pushing a competing product out since AMD produces the current and next gen consoles graphics and mantle can be used to code for both xbox one and pc.
You don't understand anything at all what Mantle is.
Mantle IS NEVER compatible between different hardware, it won't even be for 2 different AMD gpu generations (Mantle won't be compatible for HD5xxx & their upcoming GCN!!!).
It is low-level, means you have to code for each hardware separatly.

So Nvidia could write their own Mantle `driver`, but it could never be compatible with AMD's Mantle for current gen or GCN.
http://www.heise.de/newsticker/meldung/ ... 45398.html

Basically he says that Mantle is possible for every GPU that meets some prerequisites.

So never say never...
varikonniemi
Posts: 451
Joined: 03 Jul 2011, 11:54

Re: AMD "Mantle"

Post by varikonniemi »

Latest updates on Mantle:

AMD Wants Mantle On Linux, OS X, Mobile Devices
http://www.phoronix.com/scan.php?page=n ... px=MTUxMzY

Where it seems i was wrong was that XBO does not use Mantle as was speculated, but rather Mantle is derived from the xb360/xbo graphics API.
klapmongool
Posts: 843
Joined: 13 Aug 2007, 13:19

Re: AMD "Mantle"

Post by klapmongool »

http://community.amd.com/community/amd- ... for-gamers

Mantle has been released, will it work on Spring?
User avatar
Floris
Posts: 611
Joined: 04 Jan 2011, 20:00

Re: AMD "Mantle"

Post by Floris »

ofcource not

There is lots of other cpu optimizations that need to be done first. And GPU is not the bottleneck of spring. Esp in case of the mantle enabled cards out there.
And yes mantle could reduce the cpu load a bit too but well.... Optimizing just only for the already high end is just not viable for spring imho.

....unless adding support is just adding a libary, but I highly doubt that.
User avatar
Silentwings
Posts: 3720
Joined: 25 Oct 2008, 00:23

Re: AMD "Mantle"

Post by Silentwings »

Afaik no one has expressed any interest in doing it and since ATI/AMD has a long history of handling opengl poorly I can't really see why anyone would want to try. So, no!
User avatar
AntiAllez
Posts: 105
Joined: 06 Mar 2012, 18:22

Re: AMD "Mantle"

Post by AntiAllez »

sounds ignorant...
User avatar
Silentwings
Posts: 3720
Joined: 25 Oct 2008, 00:23

Re: AMD "Mantle"

Post by Silentwings »

code won't write itself...
Super Mario
Posts: 823
Joined: 21 Oct 2008, 02:54

Re: AMD "Mantle"

Post by Super Mario »

gajop wrote:*image*
Thanks for the laugh, I needed that.
User avatar
smoth
Posts: 22309
Joined: 13 Jan 2005, 00:46

Re: AMD "Mantle"

Post by smoth »

AntiAllez wrote:sounds ignorant...
doesn't stop you from posting things...
Post Reply

Return to “Off Topic Discussion”