AMDs Mantle

[Boo Berry]

For an APU, that's actually pretty decent numbers.

[+Zlip792 MVC]

AMD CES 2014 Battlefield 4 running on Mantle API...

[Andre S. Veteran]

Some leaked performance claims for Mantle and Kaveri processors, probably being readied for the launch or CES.

 

http://wccftech.com/amd-kaveri-apu-a107850k-gaming-general-performance-unveiled-mantle-45-faster-directx/

More than 3 times the speed of DirectX  :| I was expecting good results for Kaveri but not of that order of magnitude... If this actually materializes it could be excellent news for mobile gaming. No need for discrete cards to get decent framerates at decent settings in the latest AAA shooters is something we couldn't even dream of a few years ago.

[MillionVoltss]

Maybe its Microsoft's lack of interest in the PC gaming market, Its eyes are set on the Xbone now days, which is competition. Directx probably needs quite a team to keep it upto date and would be lots of cooperation between Microsoft, Nvidia and AMD, teams that may not always whish to share hardware information.

[spaceelf]

Maybe its Microsoft's lack of interest in the PC gaming market

As little developer interest as there was in DX10/11 for a long time I can't really blame them.  I know the reasons, but it still confuses the crap out of me.

 

Now devs are finally interested in the high end again, so there's plenty of reason to move forward.  I suspect MS will have a proper response within a few years, but it might well be too late for it to matter.

[Island Dog]

Oxide Games have also been showing off the Star Swarm demo. :)

 

[spaceelf]

Oxide Games have also been showing off the Star Swarm demo. :)

Now that I finally watched it, nice.  Can't wait for this stuff to show up. :)

[Athernar]

Figure I would post this here since it's along the same lines, the slides from nVidia's OpenGL talk at Steam Dev Days.

 

http://de.slideshare.net/CassEveritt/beyond-porting

 

Of note, yet more evidence for the pile that OpenGL is indeed faster than DirectX. And the MSFT fans said that Valve was lying. :rolleyes:

[spaceelf]

Figure I would post this here since it's along the same lines, the slides from nVidia's OpenGL talk at Steam Dev Days.

 

http://de.slideshare.net/CassEveritt/beyond-porting

 

Of note, yet more evidence for the pile that OpenGL is indeed faster than DirectX. And the MSFT fans said that Valve was lying. :rolleyes:

I believe a large part of it is talking about future extensions, not current.  And Valve has yet to show evidence that they've even experimented with DX11.

 

I wouldn't say Valve lied, just was very selective about what they showed.

 

Considering how behind the times their games are, go figure.

[Athernar]

Note that a large part of it is talking about future extensions, not current.  And Valve has yet to show evidence that they've even experimented with DX11.

 

I wouldn't say Valve lied, just was very selective about what they showed.

 

Considering how behind the times their games are, go figure.

 

The extensions listed are a part of OpenGL 4.4 - which is current (At least for NVidia, AMD are approaching full support).

 

There is also evidence Valve has experimented with newer DirectX versions, such as the existence of shaderapidx10.dll and mat_dxlevel 100 in dxsupport.cfg in certain games.

 

As the slides indicate however, D3D11 is still slower.

[spaceelf]

The extensions listed are a part of OpenGL 4.4 - which is current (At least for NVidia, AMD are approaching full support).

 

There is also evidence Valve has experimented with newer DirectX versions, such as the existence of shaderapidx10.dll and mat_dxlevel 100 in dxsupport.cfg in certain games.

I did reword that before you posted again heh :)  I wasn't actually sure.

 

As for OGL 4.4, it is much newer than DX11.  It took three years for OGL to reach feature parity with DX11 IIRC.  They are currently ahead from what I know, but there's also very little in the way of DX11.1/11.2 or OpenGL 4+ games.

 

My point about Valve still stands.  It is more than a little suspicious that they chose to benchmark their OpenGL code against their DX9 code and yell LOOK HOW AMAZING THIS IS.

[Athernar]

I did reword that before you posted again heh :)  I wasn't actually sure.

 

As for OGL 4.4, it is much newer than DX11.  It took three years for OGL to reach feature parity with DX11 IIRC.  They are currently ahead from what I know, but there's also very little in the way of DX11.1/11.2 or OpenGL 4+ games.

 

My point about Valve still stands.  It is more than a little suspicious that they chose to benchmark their OpenGL code against their DX9 code and yell LOOK HOW AMAZING THIS IS.

 

OpenGL 4.x has "feature parity" with D3D11, 3.x with 10, 2.x with 9. In this case, 4.4 is better compared to 11.2. (Both added implementations of PRT for instance)

 

It seems you're trying to imply Valve were being disingenuous by comparing an OpenGL 4 codepath vs D3D9, however that was not the case. Given the were using L4D2 as the testbed, the likely comparison was D3D9 vs D3D9 ontop of a GL 2.x layer. Since Source 1 does not natively support OpenGL, unlike Source 2 which is native OpenGL.

 

That, and the notion of "fixing" the comparison doesn't help anyone - if anything it hurts Valve more than MSFT.

[spaceelf]

OpenGL 4.x has "feature parity" with D3D11, 3.x with 10, 2.x with 9. In this case, 4.4 is better compared to 11.2. (Both added implementations of PRT for instance)

 

It seems you're trying to imply Valve were being disingenuous by comparing an OpenGL 4 codepath vs D3D9, however that was not the case. Given the were using L4D2 as the testbed, the likely comparison was D3D9 vs D3D9 ontop of a GL 2.x layer. Since Source 1 does not natively support OpenGL, unlike Source 2 which is native OpenGL.

 

That, and the notion of "fixing" the comparison doesn't help anyone - if anything it hurts Valve more than MSFT.

OpenGL 4.3 was the first to offer compute shaders, which were a feature in DX11 that extended to DX10 cards.

 

Really?  Making Windows look bad doesn't help SteamOS?  I'm not sure I'll ever believe it wasn't intentional.  Whatever works.

[Athernar]

OpenGL 4.3 was the first to offer compute shaders, which were a feature in DX11 that extended to DX10 cards.

 

Really?  Making Windows look bad doesn't help SteamOS?  I'm not sure I'll ever believe it wasn't intentional.  Whatever works.

 

Linux doesn't factor in, I'm talking purely on Windows.

[psionicinversion]

I hope mantle is a success and is as good as they claim it be but tbh with all talk of will mantle be good or wont it be good etc weve only got wait till BF4 comes out and can look ourselves, i know what framerates i get so ill tell instantly if theres a boost or not and an approximation of how big of a boost.  Im getting thief for free from AMD's game bundle although that comes with mantle out of the box?? so wont be able to tell unless theres an option to use DX or mantle then you could switch between API's and gauge is theres more performance or not.

 

Nvidia does to much proprietary stuff and doesnt give jack away for free.... think some opengl extensions are proprietary to Nvidia so no one could prolly use em without a license fee maybe??

 

With DX 11.2's PRT stuff i think it could be used to generate highly detailed procedurally generated game environments like in a space game when you fly through the atmosphere and come down to ground level and fly around a planet. Think that would be good but cant see something like call of duty needing it or battlefield etc

[Athernar]

Nvidia does to much proprietary stuff and doesnt give jack away for free.... think some opengl extensions are proprietary to Nvidia so no one could prolly use em without a license fee maybe??

 

That would defeat the point of OpenGL if swaths of the spec were proprietary.

 

There are vendor extensions, but those essentially exist so they can be later ratified by the OpenGL ARB.

 

OpenGL 4.4's ARB_sparse_texture extension is an example of this, it's more or less a copy of AMD_sparse_texture with some improvements/adjustments.

[psionicinversion]

ah right ok

[spaceelf]

First Mantle port is in

http://tinyurl.com/mbumu5k

I'd write more about it if I was awake :)

[Matthew S.]

Now the wait for the 14.1 beta drivers...

[spaceelf]

Seems the Star Swarm benchmark is up on Steam as well.  Cool.

[Pupik]

Javascript is not enabled or refresh the page to view.

Click here to view the Tweet

[Pupik]

Any moment now...

https://twitter.com/AMDGaming/status/429288423384301569

[Andre S. Veteran]

Sweet. Dice has measured performance improvements ranging from 14% to 58% (!) - apparently, this is most beneficial to systems with very fast GPUs under graphically heavy workloads. Can't wait for Kaveri numbers and more benchmarks. http://battlelog.battlefield.com/bf4/news/view/bf4-mantle-live/

[FlintyV]

It's going to be another day at least?

 

- http://www.guru3d.com/news_story/update_on_amd_catalyst_14_1_mantle_driver_release.html

[psionicinversion]

Ive benchmarked my DX 11 performance

 

===========================================================

Oxide Games

Star Swarm Benchmark - ?2013

C:\Users\amell_000\Documents\Star Swarm\Benchmark_14_01_30_2210.txt

Version 0.95

01/30/2014 22:10

===========================================================

 

== Hardware Configuration =================================

GPU: AMD Radeon R9 200 Series

CPU: GenuineIntel

Intel® Core i7 CPU         950  @ 3.07GHz

Physical Cores: 4

Logical Cores: 8

Physical Memory: 12875575296

Allocatable Memory: 140737488224256

===========================================================

 

 

== Configuration ==========================================

API: DirectX

Scenario: ScenarioAttract.csv

User Input: Disabled

Resolution: 1920x1080

Fullscreen: True

GameCore Update: 16.6 ms

Bloom Quality: High

PointLight Quality: High

ToneCurve Quality: High

Glare Overdraw: 16

Shading Samples: 64

Shade Quality: Mid

Motion Blur Frame Time: 16

Motion Blur InterFrame Time: 2

Detailed Frame Info: Off

===========================================================

 

 

== Results ================================================

Test Duration: 120 Seconds

Total Frames: 2248

 

Average FPS: 18.73

Average Unit Count: 3624

Maximum Unit Count: 5314

Average Batches/MS: 552.50

Maximum Batches/MS: 913.88

Average Batch Count: 33284

Maximum Batch Count: 131851

===========================================================

 

using MSI 280x gaming edition (pretty much bog standard) will have to try again when i get my sapphy toxic back but least ive got a benchmark to see if mantle improves it :d