Is AMD's new 'Mantle' API potentially as.. sieismic as it sounds..?


Garrett

Advanced Member
Joined
Aug 24, 2011
Messages
1,255
Location
Workington, Cumbria, UK
Can't claim to understand all that's being talked about, but the various tech sites all seem to think this is the most exciting thing to happen in PC gaming for a very long time & is quite literally a game changer (at least in the PC/SteamBox space)

I know there's folks here that understand far more than I could ever hope to.. what do you guys make of it..?

http://www.techspot.com/news/54134-amd-unveils-revolutionary-mantle-api-to-optimize-gpu-performance.html

a more in depth Eurogamer article (& they've nicked my tagline)
 
Last edited by a moderator:
This is as revolutionnary as a setback imho.

This is the best way for AMD to try to keep the gaming for themself.

Imagine a world where a game is none only designed to run only on windows, but only for ADM GPU windows box. Have an nvidia ? cant play it. Have a onboard intel gpu ? no gaming for you.

I cant see game publishing company going this road as it actually split their potential customer base in pieces. Or the others companies are all going the same road. Say good bye to DirectX and openGL... Even if those have a goal to limit gaming plateform fragmentation so allowing the games companies to reach more clients.

All I can tell you is the gaming world is on fire theses days, and they are all trying to find a way out of Windows 8. Nvidia opensource initiative started a few days ago also show this. As this steanbox hype.
 
be better if they released it as an open API, I dont think it will take off otherwise. To me sounds like it only would benefit those writing apps for console/PC.

the presentation really sounds vague, 9x draw calls?....thats a loaded statement. Draw calls can vary so much, it could be 9 more triangles vs 9 high meshed models.

What would be ideal is if mantle is a start from scratch API that takes lessons learned from D3D and OpenGL and creates a small open API designed for efficiency and portability.
 
be better if they released it as an open API, I dont think it will take off otherwise. To me sounds like it only would benefit those writing apps for console/PC.

the presentation really sounds vague, 9x draw calls?....thats a loaded statement. Draw calls can vary so much, it could be 9 more triangles vs 9 high meshed models.

What would be ideal is if mantle is a start from scratch API that takes lessons learned from D3D and OpenGL and creates a small open API designed for efficiency and portability.
It says in the first link it is an Mantle is an OPEN API and Nvidia could purpose it for their GPUs.
 
This is as revolutionnary as a setback imho.


This is the best way for AMD to try to keep the gaming for themself.


Imagine a world where a game is none only designed to run only on windows, but only for ADM GPU windows box. Have an nvidia ? cant play it. Have a onboard intel gpu ? no gaming for you.


I cant see game publishing company going this road as it actually split their potential customer base in pieces. Or the others companies are all going the same road. Say good bye to DirectX and openGL... Even if those have a goal to limit gaming plateform fragmentation so allowing the games companies to reach more clients.
Not how these things work.

The last company that did something along these lines in this particular area developed something called EAX, and a number of developers adopted it.  This company is called Creative Labs and none of the games that used EAX will handle EAX calls without some kind of compatibility layer, but the games will still run regardless of whether EAX hardware is present.

There's no reason this would be anymore of an issue then technologies like EAX or Physx, _if_ it ends up being an AMD exclusive API.
 
And you know why only "a number of developpers" adopted it ? Writing theses compatibility layer cost money.

Beside, when writing these compatibility layers, you've to create some generic stuff which in the end doesnt allow you to use the new API to it's fullest extend making moot to write it in the first place rather othen.

If it was possible to get these 9x performances improvements in a compatibility layer retaining GL compatibility why would AMD going this road instead of just improving their GL drivers ?

To get a visible effect using theses potential performances impovements, you'll need to create higher-poly count model and better textures. That drive the developpement cost way higher. That wont be easy to get management agree on this.

Or they'll just throw out a few added effects here and there for eye candy without realy changing the game data, which in the end wont improve the visual much... Making this technology moot.
 
The performance increase would have to be really worth the trouble of coding for the amd API as well as openGL (for SteamOS) and directx. I can see a lot of developers deciding that there wasn't much return on investment involved in incorporating yet another API into their code.
 
I see openGL as the future of graphics API for gaming, support by all major platforms (SteamOS, Windows, PS4, XBONE, Android) will ensure it's success. Why would a company waste time creating a version of their game for AMD GPU's only. 
 
AMD: "We can't figure out how to write good OpenGL drivers so we created a new API"

Traditionally draw calls speed was the strongest argument for OpenGL (vs Direct3D) as every OpenGL "driver" is in fact full OpenGL userspace implementation that can optimize call forwarding to the kernel driver however it wants and can include any number of added non-standard functionality (it's even possible to create completely new 3d rendering api atop OpenGL). On the contrary, Direct3D implementation is fixed and written by Microsoft so driver developers can't actually optimize Direct3D implementation to better suite their particular hardware.

There are nowadays two problems for a complete cross-platform 3d rendering api: Microsoft refusal to allow anything other then Direct3D on their platforms (especially consoles) and two competing OpenGL branches: "full" and ES, the latter being actively used in mobile space and Playstation 3 (don't know anything about 4). In the long term I thing these two branches will merge and ES version will be something like core profile with optional "desktop" extensions to support legacy code
 
I have a NVidia card, am I screwed now?  :unsure:

I remember the old days where "GLIDE" ruled over OpenGL and especialy DirectX but also needed compatible GFX cards to work. I had a Voodoo2 and Games like Unreal or Diablo 2 really took advantage of Glide in those days. So it's good for people who have compatible cards but bad for people who havn't. I hope NVidia can keep up with that somehow. I'm actualy always for optimized stuff that bring the software closer to the hardware and resulting in better and faster gaming.
 
And you know why only "a number of developpers" adopted it ? Writing theses compatibility layer cost money.
Optimizing code takes more. 

The implications around this are that Mantle is the low level XBox One API that pairs with the high level DirectX 11.2 API the XBox One will use.  If that analysis is true then all high performance XBox One games will use the Mantle API by default, and the associated quickly thrown together PC ports will by default.  The Industry norms related to PC porting in turn mean that unoptimized PC ports are considered acceptable.

I don't see a basis for your case in context related to Mantle.  For PhysX, yes, but we're not talking about the vendor that produces card killer drivers, shuns GNU/Linux, and was deemed most responsible for Windows Vista being nonfunctional at launch.

If it was possible to get these 9x performances improvements in a compatibility layer retaining GL compatibility why would AMD going this road instead of just improving their GL drivers ?
Because Mantle is a Low Level API and OpenGL is a High Level API?
 
Last edited by a moderator:
Looks like AMD is throwing some serious punches lately.

First they win the contracts on the consoles both Xbox and Playstation. Rumor had it that it wasn't very profitable for them which is why Nvidia didn't bother.  They have also been contributing a stack to the Linux Kernel lately and Steam OS is just around the horizon so who knows how that will turn out.

Now they released the mantle api. Granted if it was a pc only thing I could understand everyone's hesitation about it's success but to me I imagine it would translate across the platforms easily. Why wouldn't they make it easy?  It would be a major blow to nvidia who has always had the edge on gaming performance due to their close ties with gaming companies.  Now that doesn't mean nearly as much because if developers optimize for the consoles it could benefit the pc as well.  

Not only that the new 290x card they are releasing looks like it trades blows with the titan (Which it should given the time difference) However the price they are supposedly releasing it at is going to curb stomp the titan. (Granted prices will drop but I can't see by that much)  I think this is when AMD really starts to shine.
 
The implications around this are that Mantle is the low level XBox One API that pairs with the high level DirectX 11.2 API the XBox One will use.  If that analysis is true then all high performance XBox One games will use the Mantle API by default, and the associated quickly thrown together PC ports will by default.  The Industry norms related to PC porting in turn mean that unoptimized PC ports are considered acceptable.
http://attackofthefanboy.com/news/amd-mantle-implications-xbox-ps4/
If it was possible to get these 9x performances improvements in a compatibility layer retaining GL compatibility why would AMD going this road instead of just improving their GL drivers ?
Because Mantle is a Low Level API and OpenGL is a High Level API?
You read me wrong here.If these 9x performance improvement are real and general, then I dont understand how ADM couldnt use that low-level API to create a faster GL drivers (reducing the gap. That 9x is a huge number).

Or these 9x improvement are just some corner case benched to get that huge number and the improvement aren't that great.

Still quoting Carmack :

If you just port code that is efficient on GL or DX to a nativeish api, the per gain will be small... 9x draw calls is credible over stock D3D, but Nvidia OpenGL extensions can give similar improvements.
All in all, I would be in 'awe' if that realy take off and realy offer visible improvements, but I dont bet on it...
 
Back
Top