AMD Mantle Support?

[KING5TON]

Let all hold off till December when Battlefield 4 gets the mantle update.
I'm personally looking forward to seeing how much of a difference it will make.

I can already run the game on Ultra at 1080p over 60fps with a few tweaks here and there (mainly all AA settings off but 125% resolution scale). Wonder if I'll be able to play it with the same settings but with 200% resolution scale!

 

[HaiderGill]

Let all hold off till December when Battlefield 4 gets the mantle update.
I'm personally looking forward to seeing how much of a difference it will make.

I can already run the game on Ultra at 1080p over 60fps with a few tweaks here and there (mainly all AA settings off but 125% resolution scale). Wonder if I'll be able to play it with the same settings but with 200% resolution scale!

Agree with you on that the 'proof of the puddings in the eating'. What I will say is with GTA5 it has taken time for the devs to squeeze that level of graphics performance out of a so-called out of date GPU with the Sony lower level libGCM. PS3 RSX GPU is equivalent in hardware terms to a nVidia GeForce 7800 GTX card. You try running any PC game on that (7800 GTX) and try to get the graphical quality of GTA5. It may take sometime for Mantle to be truly harnessed considering that the AMD GCN architecture is new to console developers too. They don't yet have the experience of coding GCN at a lower level. This will come with time, hopefully as my HD 7950 Vapor-X starts ageing...

I am a fan-boy of getting 'the best bang for my pounds'. I don't care whether it's nVidia, AMD or Intel. As far as I am concerned keep the competition healthy and no cartels or monopolies...

 

[Kerrash]

Let all hold off till December when Battlefield 4 gets the mantle update.
I'm personally looking forward to seeing how much of a difference it will make.
...

Ahhh that is interesting, I didn't realise that Mantle was not available in the PC version yet. This update will lead to some interesting comparisons.

I wonder if PCPer will do an article, they are normally well informed reviews.

 

[manifold]

I totally and utterly disagree with you. Direct3D is not wringing out sufficient levels of performance out of our GPUs. You only have to look at the PS4/XB1 GPUs to realise that. There needs to be a lean and mean API. Why should consumers spend their hard earned money to have OS system software screw it up. Look at Sony and what they extract with libGCM. MS aren't really that bothered with PC gaming since the Xbox. SteamOS I will be migrating to as they seem to be more interested in PC gaming and a protest vote. MS I'm gonna give you a bloody nose Windows 8, Surface and phone MS you're gonna have a big bloody nose.

Well, I think the desire to move to a lower level is symptomatic of a misunderstanding of high-level architecture, or a poor implementation thereof. Like I said, PC devs are in rough shape because neither API is what I would consider "good." You're right, neither D3D nor GL expose the raw power of modern GPUs. But moving to a lower level is only the base reaction to that deficiency. It's not a long-term nor is it a smart solution.

The right solution, as it always is, is to build the right high-level architecture. Many people equate high-level with non-optimally-performing. That's only true of the bad high-level architectures. Take c++, for example. While I wouldn't call it a perfect language, I can use it to generate literally thousands of optimal assembly instructions using only a few high-level lines. Devs don't write games in assembly these days because they wouldn't gain a significant performance advantage over a knowledgeable c++ engineer. The fact that such is not the case when it comes to, for example, D3D/GL vs. hardware-specific or console-specific APIs is, again, symptomatic of a design failure.

The way I think of it is in terms of the future. Will devs in the future write hardware-specific, low-level code? Of course not. That's why I call it a step backwards. It may be a temporary "patch" to boost performance, but overall I think it's another failure to design for the future I wish one of the vendors would start trying to build for the future!

Maybe nVidia and AMD should collaborate on a separate high-level API. If this lower level (note I say lower) API can deliver more performance on same hardware why not. From the looks of it is going to be embedded in quite a few engines.

Yes! Collaboration by the "greats" on a new high-level, cross-platform API would be, in my opinion, one of the healthiest things that could happen in the industry right now!! I'm sure politics is what's keeping us from that

 

[SuperG]

Well let see who meeds mantle and who not. From dev stand point. DX and openGL just do a good job. Not on performance but time to market. It takes lot of time to optimise to use lowlevel api to get the most out of it. But it depends on budged and scale of the project.

You got this BF4 on frostbyte3 engine.
But you also got telltale studio using not low level API.

Most devs limited by workforce and time will stick to dx ogl and often with a engine that sufice for the task at hand.
Even a licenced engine sufice.
Or unity3D.

There are two direct
From small highlevel programming vs complex down to the metal programming.
From smal project with low needs to large scale heavy weight games that pushes the hardware.

There is room for gamemaker to unity3D to c4 engine to unreal and source to dx ogl to mantle and nV_API.

Mantle isnt a replacement for DX it depend if you need to go that deep into to the hardware.

What this means. DX and OGL are realy great things. Hardware abstraction is realy good thing for a large part of game industry.

Mantle to for those that want to get more out of the hardware if the got the need and dev resource to go that far.
While unity3D is great thing just like XNA to bring game development to the masses.

Dice got a engine team the resource the budged to go that extra mile also a game that a heavy one who could gain a lot from direct interacting with the hardware in this case the gpu.

But not all games need it DX wil be fine.

With that it depens on what project and level you are producing a game.

 

[rab]

I really hope mantle takes off acordign to AMD Devs have been asking for something like this for a long time and there are already a few big studios on board and in theory from what I have heard there is no reason why it shouldn't work on Nvidia hardware if they choose to add support for it... big if I know I think even if it blows DX performance out of the water Nvidia would probably not use it out of spite but we can hope lol.

 

[SuperG]

I dunno but mantle support formost the latest AMD CNG architekture. Its a direct API to gpu cores. nV cuda cores have a lot of features in common but there is difference. Also nV has there own nVAPI. Wich does somewhat the same.
The merrit seams that if devs do optimise heavely for console APU trough mantle. PC AMD ports could benefit from a port who takes these optimisation over.

Altho MS states it doesn't support mantle on xbox one.

But then again the dx implimentation in the console OS could be much more direct with less abstraction and less overhead. Because one fixed type of hardware one driver. Wich could be intergral.

So I am very currious what mantle brings oerformance wise.

Well those COM based interfaces uses lplp function calls wich have heavy overhead. So the batch up things in one call with a burst of data.

So dev already optimise to tacle the high cost of function call overhead penalty. Wich means the gain from mantle might be a worse case scenario in situation the engine can not use buffers optimal. As for nV mantle support only the CNG architektures this means older different architectures of ATI AMD also not. So it would be that due to different implementation of the cuda architecture this API is incompatble with cuda cores.

 

[Wulf]

Mantle and G-sync

I'm wondering if ED is going to support AMD's Mantle or Nvidia's G-Sync? Both these have the potential to greatly boost PC graphics for little cost.

I admit that my interest mainly lies with Mantle as if offers better performance, where G-Sync only makes slower frame rates look more fluid. AMD is also introducing a dynamic frame rate technology called Free Sync to compete with Nvidia. I suspect G-Sync will be the better than Free Sync, but requires specialist monitors unlike the latter.

 

[Mazhurg]

I'm wondering if ED is going to support AMD's Mantle or Nvidia's G-Sync? Both these have the potential to greatly boost PC graphics for little cost.

I admit that my interest mainly lies with Mantle as if offers better performance, where G-Sync only makes slower frame rates look more fluid. AMD is also introducing a dynamic frame rate technology called Free Sync to compete with Nvidia. I suspect G-Sync will be the better than Free Sync, but requires specialist monitors unlike the latter.

I suspect that Mantle (and other gizmos) will probably be part ongoing Cobra engine refreshes; not specifically for Elite Dangerous. So, I would not expect to see the technology included soon(tm), perhaps soonish or even eventually.

 

[Wulf]

Soon.

 

[Rog]

Hopefully Mantle will be incorporated, but G-sync is just dependant on GPU drivers I think, it isn't game specific as far as I recall.

 

[Cody]

G-sync is just dependant on GPU drivers I think, it isn't game specific as far as I recall.

<nods> Pretty much - Nvidia say there are a few games that might not work with it. It does require a GTX 650 Ti Boost or above Kepler-based GPU. There was mention of an adaptor for normal monitors, but I ain't too sure about that.

 

[Rog]

Yeah, the adaptor was for Asus monitors (possibly only one model). But all the other monitor manufacturers are onboard, and will be launching G-sync equipped monitors very shortly.

They shouldn't be much more expensive either, as I believe the G-sync hardware is just basically an adaptation of the existing controller hardware already present in monitors.

 

[Cody]

They shouldn't be much more expensive either...

The early models may well be a trifle expensive - the first Asus may be $399. Very nice specs on that model, btw.

 

[KING5TON]

So Mantle got released this weekend for BF4 and the Starswarm demo on steam.

The performance increases are a bit of a mixed bag at the moment.
You'll only see gains on certain cards (HD 7000/HD 8000 Series GPUs and R9 280X and R9 270X GPUs aren't optimized currently) and the gains you do receive are based on how much your CPU is currently bottlenecking your GPU(s)

A rough summary is

A slow CPU and Fast GPU - Excellent performance increases
A slow CPU and Slow GPU - Some performance increase
A fast CPU and Slow GPU - Small performance increase
A fast CPU and Fast GPU - Small performance increase
A fast CPU and Multiple GPUs - Excellent performance increase although there is an issue with stuttering on xfire setups at the moment that should be ironed out in a subsequent release of the driver

On my system (3570K OC'd to 4.4Ghz and a 7970 OC'd to 1Ghz) I don't get much in the way of FPS increase but I do get what feels like a smoother experience in BF4.
My GPU isn't one of the optimized ones so I'll have to wait for a driver update that will hopefully increase my FPS further.

I've seen benchmarks that shows Starswarm in Mantle to get a 319% increase in performance over DirectX!
That's a boost not to be sniffed at

http://www.fudzilla.com/home/item/33803-amd-releases-mantle-shares-performance-numbers

 

[Cmdr Eagleboy]

Yes! Collaboration by the "greats" on a new high-level, cross-platform API would be, in my opinion, one of the healthiest things that could happen in the industry right now!! I'm sure politics is what's keeping us from that

You mean like OpenGL?

 

[xyphic]

You mean like OpenGL?

The trouble with OpenGL as I understand it is that it moves really slowly in comparison to its competitors. I'd say it's likely that you'll see similar features appear in DirectX long before OpenGL.

Regarding whether NVIDIA take on the Mantle API, it will depend on how much of a performance benefit it brings and how many of the big game development studios adopt it. If AMD GPUs deliver much higher performance at the same price point as NVIDIA ones simply due to an implementation of an API, you can bet that they'll get support for that API implemented as soon as possible. Or if not, they'll be working to create their own or get something similar into one of the 'standard' APIs.

 

[Cmdr Eagleboy]

The trouble with OpenGL as I understand it is that it moves really slowly in comparison to its competitors. I'd say it's likely that you'll see similar features appear in DirectX long before OpenGL.

That was true five years ago. At this point OpenGL 4.3 has reached feature parity with DirectX 11. Combined with the fact it is supported in more platforms than Windows, it says something.

My problem with Mantle is that it is strange hybrid API/direct hardware access system. It could be good for boosting performance let's say on console, or one solely system like Windows. It is not a long term solution. OpenGL maybe not perfect, but it is expandable, and it works if companies are really interested in improving it (SteamOS, Mac OS X, etc.).

 

[Squicker]

The right solution, as it always is, is to build the right high-level architecture. Many people equate high-level with non-optimally-performing. That's only true of the bad high-level architectures.

This is true. we have to remember how DirectX came about. When MS released Windows NT they prevented all direct access to hardware. Why? Because previous editions of their OS which did not do this suffered reliability and security issues. So MS released Windows NT with the Hardware Abstraction Layer (HAL) and all of a sudden no-one could play games on Windows anymore as the hardware was not accessible.

Various vendors developed their own solutions but the vast array of options stunted game development for Windows NT variants significantly. People had to dual boot to DOS!

So MS introduced DirectX. Games spoke to DirectX which was allowed to speak to the HAL which went onto speak to hardware via approved drivers. Of course, these multiple layers of abstraction slow us down but they are there for a reason.

Is the whole model due an overhaul in which it is leaned off and made more performant? Very probably. Is it wrong to have a HAL and API style model? Most likely not, it was developed for very sound reasons.

It's amazing how everything goes full circle in IT if you stick around long enough!

 

[Dirk Braben]

My problem with Mantle is that it is strange hybrid API/direct hardware access system. It could be good for boosting performance let's say on console, or one solely system like Windows. It is not a long term solution. OpenGL maybe not perfect, but it is expandable, and it works if companies are really interested in improving it (SteamOS, Mac OS X, etc.).

Then I am glad to report:

There's been comments early on out of AMD that they would like to see the Mantle API supported on other platforms and it was reiterated this week during AMD's APU13 Developer Summit. AMD would like Mantle on Linux and Apple OS X as it would be significantly easier to do an efficient renderer with Mantle than OpenGL, according to AMD. Mantle also has uses beyond game-engines, reportedly in workstations and R&D too. One of the slides shared during the APU13 summit stated "Mantle + SteamOS = powerful combination!"

(highlighting by me)

http://www.phoronix.com/scan.php?page=news_item&px=MTUxMzY